May 4, 2026 • 65 mins
Most organizations have rolled out AI tools and called it a strategy. They've issued logins, run compliance training, and watched adoption numbers tick up — while the actual quality of the work stayed flat.
The problem isn't the tool. It's the mental model. If your people are treating AI like a search bar, they're only accessing a fraction of what's possible.
Geoff Gibbins has spent close to 20 years helping organizations figure out what actually works at the intersection of strategy, technology, and human behavior. In this episode, he makes a sharp and practical case for shifting from AI as a tool to AI as a collaborator — and walks through a four-mode framework that gives leaders and their teams a real working vocabulary for that shift.
You'll hear why a study of 450 people found that managers consistently scored lower than individual contributors on AI collaboration quality — and what that means for how you're modeling behavior on your own team. You'll also hear why workers in their fifties outperformed workers in their twenties, and what that data suggests about the habits we need to build deliberately.
Geoff also introduces a three-part measurement system — Results, Relationship, and Resilience — that gives mid-market leaders a practical way to assess whether their AI investment is actually paying off beyond license counts and time-saved metrics.
Geoff Gibbins is the founder of Human Machines and the author of Critical Intelligence, a book on strengthening human thinking in the age of AI. He previously served as a partner at Accenture and has worked with organizations including Walmart, Nestlé, Coca-Cola, and Vanguard.
This episode gives you a concrete framework for evaluating and improving how your organization actually collaborates with AI — and three specific moves you can make starting Monday.
AI For the C Suite™ podcast keeps C-Suite leaders informed and engaged in the world of AI for business. If you're a CEO, President, Owner, or C-suite leader looking to understand how AI will impact your organization, you've found the right podcast. Generative AI is the most disruptive General Purpose Technology we have seen in the last 45 years. It holds tremendous promise when leveraged properly and tremendous peril for those who disregard its existence. AI for the C-Suite™ is a continuous learning and application experience that exists to unite, elevate, and equip leaders to navigate the Exponential Age.
Watch the full episode: https://youtu.be/qeQ-ImaG-dI
Join the AI for the C Suite community today: https://aiforthecsuite.com/ #chadharvey #aiforthecsuite #aic
Listen
Watch
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(01:57):
I'm Chad Harvey, and this is AI for the C-suite.
The show for senior leaders who know AI matters and need to figure out what to do aboutit.
Each episode, we dig into what it all actually means for middle market organizations, howit's changing decisions, strategy.
leadership and the nature of work.
Let's get into it.
(02:17):
You know what I love when someone walks in and says most of what you're hearing about AIadoption is wrong.
And that is basically what today's guest does for a living.
Jeff Gibbons is the founder of human machines.
He spent close to 20 years at the intersection of strategy, tech and innovation, workingwith organizations like Walmart, Nestle, Coca-Cola and Vanguard to figure out what
(02:41):
actually works
when humans and AI sit down at the same table.
Before his current venture, Jeff was a partner at Accenture, where he launched entirelynew businesses across multiple sectors.
He's also the author of Critical Intelligence, a book about strengthening human thinkingin the age of AI, and he holds a master's degrees from Oxford and Imperial College of
(03:04):
London.
What I find refreshing about Jeff is that he blends real optimism with real skepticism.
And he's got a framework for how your people should actually be collaborating with AI thatI think is going to change the way you think about this.
Jeff, welcome to the show.
Chad, great to be here and look forward to the conversation.
Thank you.
(03:25):
You have got quite an interesting background.
And as you and I talked about uh prior to recording this conversation, there is a realinteresting shift that I think a lot of leaders need to go through.
And you talk about that shift in terms of thinking of AI as a tool that you use tothinking of it as a collaborative partner.
(03:46):
And then of course, there are some implications to that.
I'm interested in starting with what that mental model shift
looks like for a leader that's used to managing and leading people instead of AI.
Yeah, great question.
So I think for me, you know, I think it's easy to think of AI using the mental, mentalmodel of a tool, right?
(04:07):
So sort of like a search bar, right?
So I can put in a question, you know, what's the best X or how do I do Y and get an answerout of it.
And I think for a lot of people, when we first started using AI tools, whether that was ayear ago or three years ago or whenever it was, that was kind of the first frame.
(04:28):
of how you use it, right?
It's sort of like a better search bar.
And that leads you to think, well, how do I get the perfect prompt?
How do I construct it right?
So I'm going to get the right output.
And then in terms of what I get out of the other end, I'm thinking about it in terms of,how good is that?
Are there errors in that?
I'm primarily thinking about how good is the tool?
(04:49):
I think what's interesting is that
I think a lot of people are starting to see that AI can actually do so much more and theycan achieve so much more in their jobs when they think of it as a collaborator.
So rather than thinking of it as a one-time quick hit, how do I put in a question and getan answer?
It's about how do I do my job differently by imagining there was somebody sat next to meevery day who said, Hey Chad, I am right next to you.
(05:17):
I am very skilled in all of these ways.
have nothing to do today.
Ask me for some help on these tasks and let's collaborate.
And not only will you Chad, like give feedback to me, but I'll give you feedback too.
I'll help you to be better.
Right.
And so thinking about AI as a collaborator who can actually help you be better, who cangive you feedback on your work and help you understand biases and blind spots in your
(05:44):
work, I think gives you a different frame of how you can work with AI and actually helpsyou to achieve more too.
It's so interesting when this journey started in the the wider world uh and everybodyloves to talk about chat GPT but as you and I know AI is much more than that but that was
kind of the uh origination point for a lot of conversations with folks.
(06:06):
I used to say something along the lines of it's kind of like working with an absolutelybrilliant you know college junior that has access to all the information in the world but
lacks all common sense and real-world experience and
I feel like we've evolved significantly beyond that to the point where it's now maybesomeone that's got a master's or a PhD level and maybe a few years of experience in your
(06:29):
business.
Again, some of those same limitations, uh but the ball has significantly moved or uh thecapabilities have advanced.
And so when we think about that, I love your idea of this being somebody sitting rightbeside you that is just saying, hey, give me something to do or bounce some ideas off of
me.
(06:49):
So when you walk into organizations and you work with executives and leadership teams,what do you find their biggest limitation is in terms of thinking about these tools as
collaborators versus that standard search box?
Yeah.
So I think, I think what I often see in organizations is that there will be a desire, youknow, one intention, good desire to think about, well, there's this new thing we should,
(07:15):
we should, we should do that too.
We should roll out some kind of internal copilot system or enterprise chat, GBT or Claudeor whatever it might be.
And we think that will be a good way of helping us to make our people able to do theirjobs more efficiently.
you know, save some time, all good ideas and totally true.
(07:38):
And that's, that's possible.
Typically there'll be a rollout.
People will get some kind of compliance focused training just to make sure they don'tupload anything that's particularly sensitive or, know, kind of the general rules of the
road, mostly from like a legal and confidentiality standpoint, although, you know, a lotof that is really covered now by the, the sort of frameworks and contracts that you sign
(08:00):
with an LLM from an enterprise standpoint.
So.
People will get that and then they just kind of get often left to figure it outthemselves.
And I think people default to how do they use other tools, right?
And it's easy to understand why, right?
So you can think of it as a search bar, right?
You can ask it questions.
You can think about, well, this actually can do some writing.
(08:23):
And so, you know, I can improve my emails.
can get it to draft something.
And so you sort of think about, what are the simple tasks I can get it to do?
What I think...
actually becomes a bit of an unlock for people is thinking about there are lots ofdifferent modes in which you can work with AI.
And so that's one of things I try to unpack in the book, Crystal Intelligence, and in mywork with companies is to think about how do you understand what's the right mode that you
(08:50):
could be working with AI in this?
And how do you think about the right mode for different types of work?
I think the, so there are four modes generally that I talk about.
So the first mode is consulting mode.
And so this is kind of like treating it like a search bar.
So as a human, you're kind of driving the task.
You're thinking about what am going to do?
But you're asking AI for some input.
(09:12):
So you might be saying, what do you think of this email?
What's the best X for Y?
And you get an answer as a result.
The second mode I think of is where you're in approving mode.
So you are going to give it a task.
please come up with three or four different ideas for X.
(09:34):
And then you approve that and you say, okay, great, let's go with this one.
m And I think that's a deeper form of collaboration, but where it gets more interesting iswhen you actually start to then enter into a dialogue with AI and actually say, okay, so
let's take these three or four and let's build them further.
(09:56):
Let me...
give you some constraints and some more context so we can make it better.
And then let's work as partners to do that.
The third mode is what I call supervising mode.
So this is where you're actually giving more of the involvement in the work to AI itself.
So you might say, give me three or four ideas for this, and then actually flesh out a riskmitigation plan for each of them.
(10:22):
And then figure out what's an implementation plan.
And then
give me feedback, and then I'll give you feedback on all of those things.
And then let's improve that plan together.
And so you actually work, you're kind of acting as a supervisor or a director of the workthat AI is doing.
And then the last mode is what I call delegating.
So this is where really the AI is going to do things for you.
(10:47):
And you're going to set the parameters and review its performance over time.
But this is really useful for situations where
It's fairly low risk and the AI is going to be better at doing it than you.
So let's say for example, this might be, Hey AI, I want to be on top of trends in thiscertain area.
(11:09):
Please every morning look for any news around this and give me an update and tell me thethree things I should learn about this.
That will be useful for my job.
Right.
And so you're actually kind of delegating the work and saying,
You go and do it.
Give me a bit of a feedback report on what's going on, but you just kind of go and dothat.
(11:29):
Or it could be that you're actually letting it to letting it do work for you.
could be that you're going to say, can you please send these three or four emails a day toother people?
It could be that you're going to say, here's my, you know, here's my household budget.
Please create a grocery list for me every day.
Like, you know, outside of the world of work, you're basically letting it do tasks.
(11:54):
and just give you an update on how they're going.
I want to dig into these four uh distinct modes of AI collaboration a little bit more.
But before we go right into that, I want to go back to where we started with the issue ofthe leader or the leaders of an organization.
Because I feel like a lot of leaders of organizations are very hot on AI for theirorganization, but they're not really uh leveraging any of these four different modes uh in
(12:24):
their personal work.
let alone thinking about how they do their work so that they can model it for their team.
Agree?
Disagree?
Thoughts on that?
Yeah, I think there's a lot of leaders today who are thinking, number one, I don't want tobe left behind by this.
And so I think there's a lot of FOMO out there and people are thinking about how do Ifuture proof my own career and what does that mean to me?
(12:47):
And I think in working with leaders, one the things I like to focus on is that I think alot of people in developing their careers have become really good at
answering questions and being known as the person who has the right answer.
Right.
That's why you get ahead.
That's where you get promoted.
And I think what's challenging for some people is to think actually maybe the AI can dothat just as well as me, right.
(13:12):
And maybe even better in some ways, right.
And actually, especially if you think about it in terms of AI being able to answerquestions that are outside of your domain expertise that you might not know about.
Like if, you know, if you're a finance executive,
AI might know more about R &D than you, right?
And so there's a sense of, well, I mean, what do I really excel at if AI can give betteranswers than me?
(13:36):
And so what I try to focus on with people is you need to focus on what are the rightquestions that you should be asking, right?
So in a world where you can get an answer to any question in seconds or minutes, reallywhat matters is what are the questions you ask?
And then the second thing is
Once you've asked that question, how do you actually work with AI collaboratively to getto an output that you couldn't have ever done before?
(14:01):
Right.
So how do you actually understand the best way to get the best output together?
Um, by using the best of your skills and your domain expertise and the analytical skillsand research skills of an uh AI system, which are able to surpass what you can do as a
human.
And so how do you actually use the best of those things?
(14:24):
And so that's one of the things I've been working on is how do you actually train and helppeople develop and measure their skills around this principle of collaboration rather than
just prompting or getting a good answer.
Right.
So it comes from how do you actually identify what's the right question to ask?
How do you provide the right context?
(14:44):
How do you provide constraints?
How do you understand how you can work with AI to figure out, is this a good answer?
Could we do better?
Can we analyze the options here?
Can you, AI, understand my blind spots around this and help me make a better decision?
So it's all about how you can work with AI to amplify what you can do as a human, not justget a good answer.
(15:09):
Do you find that a lot of leaders want to put that prompt in or give the task and they areassigning it in a way that does not allow the AI to ask questions and to clarify the
intent?
Cause I see that sometimes with folks, are used to, I don't want to say barking orders,but they're used to dispensing direction.
(15:36):
And I think that that leaves a lot of opportunity on the table per your point aboutworking with these tools.
Yeah, no, absolutely.
think that can happen a lot.
And I think it comes down to two things.
One, I think is it's partly the mental model, right?
If you're not thinking of this as a collaborate, so you're just thinking, basically, howcan I get what I want from this thing?
(16:00):
Right?
You're just thinking, what can I get out of this?
And I'll be the judge of whether or not it's good.
Right?
ah I'll use my best judgment.
You're not thinking about how could it actually make me better?
you know, as a humility to thinking about it in that way, right?
The second thing actually is that, so that happens a lot.
And I think the other reason for that is it's actually about the level of context that youprovide.
(16:25):
So one of the things I've been doing over the last month is actually developing a systemfor measuring and helping people understand how well they actually collaborate with AI
systems.
And I ran a study actually.
So, so about a month ago, I ran a study with about 450 people roughly who were eachworking with an AI tool to define a problem.
(16:49):
So real people working with an AI tool to define their own problem and then work through aseries of conversations with that AI to solve the problem and get to a good solution.
Right.
And so the AI system was instructed to
You know, dig in and understand the question, give suggestions, but then also flip thescript a bit and work with the user in a different way to ask them for suggestions, to
(17:16):
give them feedback and see how the person would actually react to different styles ofcollaboration.
Like those, those formats that I talked about.
So using this, uh, this tool and this study, so there are about 450 people ran throughthis and they had a conversation about 40 minutes in length.
I was able to understand like really what was going on with these people.
So very rich data set with thousands of messages.
(17:38):
And one of things that was really interesting was that managers versus individualcontributors behave differently.
So managers generally provided instructions and context that was a little bit terse andmore sparing as you described.
Right.
So it's actually about the, you you might just say like, do this or please give me threeof these.
(18:01):
Right.
And.
versus individual contributors were more likely, I guess, they were, you know, potentiallybecause they're more used to having to kind of explain themselves and provide context even
for a question, they were more likely to provide better context to the AI.
And they actually scored overall better in terms of the quality of the output that theygot to working with the AI.
(18:27):
Very interesting.
A recent episode where I chatted with Dr.
Sam about some of her methodologies, we got into this distinction between managementversus leadership.
I purposely used the word management during one question as opposed to leadership to seehow she'd respond.
But we really didn't unpack this issue of collaboration.
And I think those are the three um areas here in terms of how we're working.
(18:53):
with these tools and within organizations that I think are worthy of focus.
The idea of management versus leadership versus collaboration.
And I think there's some spillover between those, but you raise a really interestingpoint.
And one of my thoughts is that we're rapidly moving into an era where everyone is a leaderof something in the organization, whether they're leading other humans, whether they're
(19:15):
leading AIs or both.
And many of those...
uh
think relationships, for lack of a better word, are going to involve the type ofcollaboration you're talking about.
Yes, absolutely.
And I think, you know, right now for most people, the relationship with AI is sort of likea one-to-one relationship, right?
I'm just chatting with my chat bot thing.
(19:35):
think that's starting to change in that now you, know, if I'm working with Claude, it willcreate a little team of agents.
And so even just in the language it uses in user experience, I'm thinking about itdifferently.
I'm thinking about it as.
you know, a team of little agents that's going off and doing different things and comingback to me.
So I'm starting to conceptualize it a bit differently, even though it is just a chat botthing.
(19:59):
It's like, it's actually like little entities doing their own thing.
Uh, but I think the shift really will come from when the user interfaces start to lookdifferent in a way that you and I could both be collaborating in a chat environment.
with an AI or multiple AI agents at the same time.
(20:19):
I don't think we've really seen that emerge, but I think that will be the driver of thatmindset shift.
And I think it's a question of how comfortable will people feel with that and how do youactually make that something that gives you better amplified outcomes versus you just
having your own little chat.
I couldn't agree more and I think uh from my personal and professional experience, I havebeen very wary over the last several years of especially with client data exposing too
(20:51):
much in the public models even if they are secure.
So we spun up our own secure uh environment and we've just recently enabled the kind ofpervasive memory that allows it to understand the individual users.
So you're getting
what you've had in the public models in a very private setting and I haven't had thatbefore with confidential client work, but I've had it in the personal setting when I don't
(21:13):
care whether you know something gets leaked out and I'm saying this and laying the tablehere because I think what you're starting to uh drive towards here, which is fascinating
towards me, is this idea of a pervasive kind of larger entity with the subagents thatunderstands all the different projects that I'm working on.
(21:33):
Maybe not thoroughly, but it has knowledge that can be applicable.
So when I'm asking for research, yeah, yeah, yeah, yeah, exactly.
And, I think you're onto something here, Jeff, that I don't think a lot of people arefully appreciating yet about where this game is going.
Um, and so I don't know that I've got a question here necessarily, but what you're sayingreally resonates with me.
(21:55):
And I feel like this is the type of maybe not bleeding edge, uh, usage with these tools,but leading edge usage.
that more folks need to start to wrap their heads around.
I think it has some really interesting human and organizational implications too.
So if you think about it, let's use an analogy, right?
So people working in a job, they have their own professional networks, which are theirown, right?
(22:20):
They don't belong to the company.
And in the same way you, know, connections, whether that's LinkedIn connections or emailaddresses are just people that you know, and trust that you have with other people in the
same way, people are building very sophisticated.
memories in AI systems about who they are, what they know, that, know, the AI systems areunderstanding this tacit knowledge.
(22:44):
Sometimes that's in their own personal AI accounts and they're using that for kind of, youknow, for work things as well.
This kind of shadow AI usage that you, that you've probably heard of a lot.
And then also then they're doing that in their own company systems and the tacit knowledgethat these people have.
(23:05):
And their understanding and their expertise is really valuable.
And so then I think it becomes a question of how do I actually save that for myself?
How do I understand that myself?
Who owns that?
Like, how does the company benefit from that?
There's like an interesting thing to understand there.
And all of that's just talking about an individual.
When you think about the level of teams, right?
(23:27):
So if somebody was to retire for a company or leave a company, how do you think about
carrying over that tacit knowledge and expertise from that person.
That's different.
Like, you know, currently people think about it in terms of like, well, how do I get theirfiles or their contacts from their Outlook calendar?
It's a very different thing to think about how do I actually store and codify and protectsome of the tacit knowledge and workflows and things that they've been doing for years,
(23:58):
which we could benefit from.
And then how do I, an individual think about how I can maintain that to myself?
Yeah, no, this this gets directly to the heart of something that I've been talking aboutin a lot of workshops and in public speaking over the last couple of years, which is when
you've got a general purpose technology, it reshapes everything, uh including law.
(24:18):
And my head as uh a as a classically trained attorney that does not practice, don't callme for advice, uh goes immediately to what is the right of the individual worker like you
laid out with respect to that tacit knowledge?
and where is the line between what I know and what I bring to the organization versus whatthey're now able to capture ah and begin to deploy.
(24:41):
Yeah, absolutely.
It's actually one of the things I've been building.
So based on this study, I've been building a tool that measures your AI collaboration.
So it's called Corex and I can, I can share with you the, the details of the, of the tooland the assessment methodology.
But really what that does is it works with you.
So you can work with it in a live chat assessment, as I described with those sort of 30minute conversations.
(25:04):
You can also actually use it to analyze your past conversations to unpack and
and go through and analyze, well, what's the quality of the collaboration?
How is it changing over time?
Is there evidence that you're improving or learning or taking on new concepts?
And in the design of that tool, it's been really important to think about what does theindividual keep and what's visible to a manager, a training company that might be working
(25:31):
with them on it?
What's the boundaries?
I think from the work I've been doing so far, I think what's really important is that the
The conversations themselves seem to be something that are something that needs to beprotected and private to the individual and the employer too.
But the overall kind of signals and implications of it are something that could be shared,but never at the granular level of like, here's the specific thing that Chad was working
(26:03):
on or doing, but here's the kind of general strengths.
and opportunities that Chad has within this.
And then, you from that, I've been developing a dashboard that teams can look at to see,how well are my people able to collaborate with AI, but that's aggregated and anonymized,
right?
So you can't see how well does Chad do versus how well does Jeff do.
(26:23):
You can see at the aggregate level, how is my team doing.
Well, this, you you gave the example earlier in terms of uh when, I can't remember if youmentioned when someone leaves or when someone starts, but basically talking about how we
think about that traditionally.
And what I'm wondering about here is whether any organizations are starting to grapplewith the idea of what information from our system that the human has been collaborating
(26:50):
with when the human leaves, do we need to provide some type of portability or export orpurge?
uh in a more extreme example from the system.
Like what do I have the right to as uh that individual employee uh to know about myselfand the way I work and is there a way to take some of that with me to my next job, maybe
(27:11):
to see my own agents.
Yeah.
So I think, um, two things I can say.
One is I, I've seen companies start to look at that at the enterprise level, right?
And this is actually something I've been doing research into myself as well is how do youactually understand, um, what's the, what's the kind of decision-making culture of an
(27:32):
organization and how can you make sure that your AI systems are aligned with that and makesure that your
fine tuning and instructing the AI to work within the parameters of that.
But that's more at the enterprise level, at the company level.
What I have also seen, um even actually like two, two and a half years ago at my previouscompany, somebody actually put into their employment contracts a term, like a term in the
(28:05):
employment contract that they couldn't...
The company wasn't allowed to recreate an AI version of that.
So this was like two, two and a half years ago, right?
So I think it's probably coming.
That was a very early example of that.
We were quite surprised by that and we agreed to it.
But I think it's a really interesting emerging area.
(28:26):
I haven't seen companies grapple with it at much scale.
Yeah.
I think the tool that you've talked about establishing in terms of measurement andcollaboration is really interesting.
If I'm a mid-market CEO and maybe I don't have the type of resources of an enterpriselevel organization, and I came to you with a question because I'd heard this episode and I
(28:50):
said something like, I need to know if my investment AI is paying off just beyond licensecounts.
What would you encourage them to track?
Is there a way for them to, without a more sophisticated system, begin to gauge the levelof collaboration that's going on as well?
Yeah, sure.
So the Corex tool is designed exactly for that.
(29:13):
So the goal of it is that there are two ways you can analyze your skills.
The first way, as I mentioned, is actually having a live chat assessment.
And so that's for an individual to have a conversation with.
I think the main applications I'm thinking about for that are you could use it to assesspeople you might be looking to hire.
You could...
(29:35):
use it at the outset of a training program or just even to design a training program andunderstand what are my needs, as you said.
So that gives you sort of a real problem solving challenge and then you work through andget advice on how well you do.
And that's something you can run across your team.
And I've started to see there are similar types of efforts going on in hiring assessmentsacross other companies too.
(30:01):
The second, which is, and so that takes about 30, 40 minutes for an individual to do.
The second form of it, which is a little, little faster and easier too, is that you canbasically log in and you get a really long prompt, which instructs your AI tool of choice,
whether that's, you know, chat, GBT or copilot or Claude to analyze your pastconversations through the lens of a range of different metrics.
(30:29):
I can tell you what they, what they are.
and then give a report back to the tool and then it generates some analysis.
And in doing that, it strips out the actual content of your conversations and anyconfidential information and just gives you an analysis of, when you're doing X, Y, Z
things, here's what you can improve, here's what you can do better at, and here's thetrend over time over the last X weeks or months of you using it.
(30:55):
And so what it looks at is,
Overall three main areas.
the, so the three Rs results, relationship and resilience.
So the first area results is I'm actually getting good results from working with AI.
Like is the output actually really good?
And is it better than a human could actually do by themselves or an AI by itself?
(31:16):
So when you're actually having this live chat conversation, one thing it does in theanalysis is it looks at different branching points in the conversation.
And then it simulates out if the conversation had stopped there, the human had stopped andthe AI had just continued talking with itself, what would the output have gotten to?
(31:37):
And then it compares where you actually got to.
And so it uses that to understand what's the value of the human judgment that you've addedin the conversation.
And it does that at different points in the conversation and figures out not only what'sthe value that you've added as a person, but also when and where did you actually add
particular value?
So then you get an analysis showing, yeah, actually when you, when you provided thatconstraint or when you add the extra content chat, you know, this is when you actually
(32:06):
added a lot of extra value.
And so you can use that to get a real feedback on, your conversation style and, where you,where you sort of did well and didn't do well.
So that's sort of the results.
The second piece is the relationship.
So what's the quality of the dialogue?
So are you pushing back enough?
Are you actually providing.
the right kind of context.
(32:26):
So one of things that was really interesting from the study with about 450 people was thatthere was a general trend.
People actually in their fifties were the highest scoring age group.
People in their twenties were the lowest going age group.
And the reason for that was that people in their twenties were much more likely to justpassively accept the output of the AI and be like, yep, right.
(32:51):
Sounds good.
Okay.
Tell me more.
Or let's go with that versus people who were more in their fifties, they were more likelyto engage, provide constraints, say, I'm not sure that actually works.
Here's a nuance on that.
And so this is like a behavioral thing.
It's very trainable, but had a noticeable impact on the quality of the output.
(33:13):
Can I pause you there before we get to the three Rs?
That distinction is fascinating because one of the things I keep hearing over and overfrom folks is, well, I just need to hire somebody that's younger that can absolutely be
brilliant with our AI tools.
And what I'm wondering here is whether that contrast in the generations is a issue ofexperience and seasoning, or is it a true cultural generational change uh in that?
(33:39):
the 20-somethings are more digitally native uh and are used to a different set of, uh Iguess I'd call it, and usage with uh digital tools than folks in their 50s.
I'm interested in what your thoughts are on that.
Yeah, I I think it's probably a little bit of both Chad, but I think maybe more thelatter.
So I think it's partly about having more experience, being more seasoned and maybe havingmore context.
(34:04):
And I think that's often what people say.
Well, yeah, people, people with more experience are going to be able to provide moredomain expertise.
But I think actually the data showed it was primarily more that, you know, there werepeople in their twenties who were able to do that, but the, it's more of a sort of like
the cultural usage, which is right.
So I think.
And this is, you know, this is a hypothesis.
(34:26):
This is, it's hard to know the exact reason, but my assumption would be that there's justa general experience for people who are younger, more digitally native, who have grown up
more with AI tools, whereby they are more used to the sort of passive response andaccepting of the output.
(34:47):
Uh, whereas people who were generally older, they, they just
work with it a little bit more like you would in a conversation as opposed to working witha tool.
So it was more that kind of culture of, well, let's actually like work with this thing.
Let's collaborate more actively rather than just using it as a thing you can do, you know,have to run a task and say, great, now do the next thing.
(35:13):
Thank you for pausing and interrupting your flow there, because I think that particularexample goes directly to the heart of a lot of conversations we need to be having about
the nature of collaboration in the new work environment and also our personal environmentas well.
mean, you know, there's that old joke on the Simpsons.
uh I, for one, welcome our new robot overlords, right?
We need to figure out how to work with these systems.
(35:36):
And if there are very real generational differences,
in output and the way we're interacting with them, they're showing up.
I just thought that was a very interesting data point.
So thank you for that.
uh Your third R, we talked about results and relationships.
Yeah.
So the third R is resilience.
So this looks at a few different things.
overall it looks at, you building skills?
(35:56):
Are you building?
And so as an example, in the live chat assessment, what it will look for is, is thereevidence that I'm actually taking on an understanding and building upon what the AI has
told me or what solution we've come to together?
Or am I just kind of like, like, or is that like, I'm actually understanding this?
(36:16):
Am I developing new?
skills and understanding as a result.
Also looking at your past conversations over time, it looks at, you building skills?
Are you getting better at your AI collaboration over time?
Are you understanding new concepts over time?
Right?
Is there a sign that in working with AI, you are picking up new things, developing morecomplex framing of issues over time?
(36:42):
This is something that I think is really important because we talked a little bit aboutearlier.
You could be a leader, let's say in finance, right?
Like you have very deep functional expertise.
One of the things that's really important is that AI can actually help you get better atthings outside of that domain of expertise.
Right?
So like R &D was the example I used earlier.
(37:02):
Right?
So what's really interesting is to look at, you know, for a finance leader, are theybetter able to understand and grasp and grapple with concepts outside of their
area of expertise and is that growing over time?
Because that's really one of the most beneficial things is that you become less siloed asan expert and more of a, I don't want to say a generalist, like you just have broader in
(37:26):
your expertise.
Mm-hmm.
I think that out of all of those three Rs, I don't want to say one is more important thanthe next, but that one resonates, the issue of resiliency resonates with me the most
because I think there's a temptation for us to just want to lean into our strengths anduse some of the AI tools to get even better in that zone.
(37:48):
But to your point, knowledge and the value of knowledge and expertise is rapidly gettingcommoditized.
My words, not yours.
I think that if you are able to understand that and to become more of that generalist andto build that resiliency, then perhaps you can begin to work with these tools to help
(38:10):
assess some of these different domains through your particular lens, right?
Again, going back to that idea of the smaller agents coordinated by the super agent andunderstanding who we are, what we do.
I think there are ways to work with these tools in a collaborative environment that domake you more of a powerful generalist.
and add more value to the team as at the same time, the value of traditional uh knowledgesets is perhaps decreasing.
(38:37):
Yeah, there's a huge issue with the sort of the cognitive atrophy.
And also I think just really an understanding the work that you're delivering.
It's so possible now to produce, you know, this would be the same in educational contextsand work contexts to produce a beautiful presentation or a great essay, but not actually
(38:58):
understand this.
And I've seen this myself, actually.
I was at a conference a few months ago, about five months ago, and I saw
a very seasoned executive deliver a presentation and it was clear he did not understandactually what the presentation was.
It was clear that the AI had written the presentation and it was actually great content onthe slides, but it was clear he didn't understand it and didn't have the fluidity to be
(39:23):
able to answer questions about it as a result.
I haven't seen anything that extreme, but I don't have difficulty envisioning that at all.
It was really bad.
It was really embarrassing, actually.
Oh, just because you've got uh the world at your fingertips and the ability to dosomething like that doesn't mean that you should, right?
(39:44):
Yeah.
And it's back to this issue of like, what happens if you don't have the tool, right?
Like if someone asks you a question outside of the meeting, you need to be able to answerthat, right?
You can't just be like, wait, hold on second.
Let me go and ask Chachi if you see the question, the answer to that, right?
You need to actually have the understanding yourself, not just be able to get the thing tospit out the thing.
(40:05):
And so that's the important thing of how you building skills and knowledge through this,not just getting out.
We've been talking an awful lot here about knowledge work and I know because I haveprivate clients in this space, but also we get a lot of email from, from listeners that
are in more tangible uh trades, manufacturing or uh construction, things like that.
(40:26):
So let, let's shift gears a little bit.
know you did some work with Walmart and I believe the system you worked on with them wascalled trend to product that changed.
Yeah.
Okay.
Good.
I got that right.
Changed the way that they design and ship products.
So I'm interested for our folks that work in more tangible areas, not just knowledge work,because there's an intersection here between the knowledge and the tangible.
(40:48):
Can you walk us through what trend to product looked like and what was the starting point?
How did humans and AI divide the work?
What changed in terms of outcomes?
I know I'm throwing a lot up there, but I'm fascinated by this intersection between the AIin knowledge work and then the tangible world that a lot of our listeners live in.
(41:09):
Yeah, sure.
So, sorry.
So, so the work, so the work really started over two, two and a half years ago, looking athow would you actually introduce AI into the product design process?
So Walmart, they produce their own private label brands and products across a range ofdifferent categories.
(41:30):
And the conversation started with where would we apply this and what would be the rightcategory to start with?
So they.
You know, they produce everything from clothes to food to furniture to everything, youknow, across all of the realms of everything you can think about buying in Walmart.
And so we started out with the idea of working in fashion and clothing, partly because itwas a very difficult area.
(41:58):
So actually designing a garment is not easy, right?
There are lots of different parts to it.
You have different suppliers, you have
different components, and just the process of designing something that actually isconstructable is technically feasible is actually quite difficult.
(42:18):
And so we started with that as the area to start with.
And what we looked at first of all was understanding what's the current process for, Iwant to do a blue dress to that blue dress being on a shelf that somebody can actually
pick up or rack in this instance.
And so that process is quite arduous, involves lots of different steps, lots of differentteams, different suppliers, and about a year actually.
(42:47):
So working as a designer, you're having to forecast and use a crystal ball to understand,what could I do that people will want a year from now, or maybe nine months, but basically
a year, depending on the type of garment you're looking at.
And so we mapped out.
what were all the different tasks in the current process?
(43:07):
But instead of then looking at, what are the point solutions to automate all of thosesteps?
We looked at, what were the real outcomes that we're looking to achieve here?
So we're looking at, we want to achieve cohesion across a line.
We want to make sure that we're being efficient.
We want to make sure that we're only producing 20 different garments rather than 30.
(43:29):
Right?
You know, like let's satisfy as many needs as we can with the fewest number of items.
We want to make sure we're being efficient in terms of the use of suppliers and the use ofdifferent fabrics, but also staying on trend.
And so what we identified was a new working system whereby the tool would be picking up onthings going on on TikTok on the runway.
(43:54):
would then generate mood boards and inspiration, you know, know, a
broad range of design inspiration for the teams to analyze.
There would also be a team of agents that would be supporting in that in terms of acreative director agent that would assess, how well was this in line with how a different
(44:14):
brand worked and all sorts of different parts along the process.
But the key thing was actually, how do you design the system that gets to a workable techpack, as it's called, that is
something that can be translated from a computer screen to something that a factory canproduce and do that at scale.
(44:35):
And so that required a lot of work with suppliers, with merchants, with designers workingtogether to redesign what their job would look like.
And so was less about how do we automate these things, it's more about how do we redesignthe workflow.
And now they have managed to successfully launch that, they're shipping clothes every daywith
(44:56):
using this new system called Trend to Product.
And as a result, they have built a capability and an intelligent system to understand andbring back feedback from the market into the design process.
And so that's the real innovation is that they own all of the intellectual property aroundthe garments, but also understand from the system, how to feedback design and demand
(45:20):
information into the next design and the next design and the next design.
That is absolutely fascinating.
mean what you're talking about here is Pattern mining and trend spotting over a large anddiverse set of inputs I mean you mentioned tick-tock and some other things there coming
back in Yeah, that is absolutely wild and I'd love to know Clearly we don't have I guessthe the data or the timeline yet.
(45:48):
I'd love to know how an approach like that contrasts with um
the accuracy of more traditional approaches because my sense of fashion is that somebodykind of goes like this ah and you know senses which way the winds are blowing and then
they make their bets in terms of the production.
Yeah, so I think that, yeah, that is true a lot of the time.
(46:10):
And I think, unfortunately, in the fashion industry, about 20 % of all clothes that areproduced end up in landfill or burned.
So this is not like 20 % ends up in TJ Maxx or being sold after the fact.
It's burned or put in landfill, just because it's hard to predict.
So this was one of the reasons to select that as a category.
(46:32):
um
you're never going to have a hundred percent hit rate.
But the key thing was how do you design enough feedback loops early on in the process inso that you could actually improve the predictability and also shorten the time between
the prediction and actually production, right?
(46:53):
How do you shorten that time so that you can at least have a better view of what's goingto be trending and then also where is it trending, right?
So a certain trend in culture
might manifest itself in different parts of the US, you don't want to produce it in Oregonif it's really only a thing in Virginia, right?
(47:14):
So there's things like that you need to understand too.
Wow.
what role, uh no, that's probably the wrong question.
I'm interested in the role that human designers play in the new system and how their roleschanged and where that, you know, you hear people talk an awful lot about human in the
loop.
I'm interested which parts of that loop or which part of that process do the humandesigners play a role in and how does that compare to what they were doing previously?
(47:41):
Yeah, so I think one thing that was really important to understand was that the humanswere involved in the redesign process.
And I think it was framed as, let's not design a tool, let's redesign what your job lookslike.
And so it's a different framing.
so for them, actually, they were very involved in helping the team understand, what dothey do today?
(48:06):
What are all the things they find a pain, frankly?
what are all of the monotonous tasks and the steps that they wish they could skip andunderstanding all of that and then using them as really testers of the new system, both
the actual tool itself, but also the way of working, right?
And so they were critical in that.
(48:27):
And they were very engaged in that because I think the framing of how do we redesign yourjob to do things better, faster.
was more appealing than how do we just automate your job, right?
That's a different framing and they were actively engaged and I think enjoyed thecollaboration as part of that because they could see how it was actually gonna make their
(48:51):
job more enjoyable too.
And I think there's a factor in that too when you're working in an area where people arefeeling like that risk of automation is there, there's a desire to future-proof yourself.
be part of the leading wave of how things are going to change, both in terms of shapingwhat your job looks like, but also being somebody who's ahead of the curve.
(49:17):
There's a future-proofing element that strengthens your resume as well.
That future proofing piece, I see an awful lot of organizations out there uh throwingaround some version of the idea that we are going AI first, right?
We're an AI leader, we're AI first.
And I know that you are very focused on outcomes first, AI second.
(49:41):
So I'm interested, have you ever gotten some calls?
I know I have, but have you ever gotten calls from organizations that desperately want toadd AI to an existing process?
And when you get in there, you realize that the entire workflow needs to be redesigned.
Yes.
I've also, think I've also myself been guilty of that as well.
(50:04):
Right.
And so I can think of, I can think of projects I've done with clients where that wasreally what we did is we automated an old process rather than fundamentally redesigning
it.
And so that's why I I've taken this view that you need to really think about what's theoutcome, not how do you automate the tasks?
Cause I, there are projects I've done in the last
(50:27):
three years, probably sort of, you know, in the sort of two to three years ago timeframe,when a lot of this technology was newer, where that was really what we did.
We actually automated the existing workflows meticulously and to a high level of quality.
But I think it made me realize how much we were missing out on by not redesigning the waythat the work could be done.
(50:52):
And I also would imagine you've walked into a number of organizations that want to deploysome type of AI and found that more traditional solutions are actually in their better
interest.
Yeah, well, absolutely.
I think, um, yeah, there's another, another project I'm thinking of with a client where webuilt this amazing AI tool, but it was, it was something that people didn't really need,
(51:13):
right.
And it didn't get used and the existing ways of working were clunky, but fine.
And there wasn't the need to change that there wasn't sufficient need to change.
Yeah.
I think that's a good hinge point to pivot into a related area, which is common mistakes.
(51:36):
So organizations that try to roll out AI collaboration at scale, you know, they've donethe pilots, the leadership's bought in, they've done some training, but then something
stalls out.
In your experience, where do you see this break?
Yeah, I think it can break it a few stages.
So think the first thing is just releasing access to the tools without active guidance onhow to get the best out of them.
(52:03):
Right.
So what kind of tasks could I be doing with this?
You know, beyond just getting a good prompt out there, like how do I actually bestcollaborate with this?
Some of the things we talked about, I think the second layer is it's great to have peopleusing and experimenting, but
how do you actually identify, what are some real opportunities to reinvent workflows orreinvent how we work that require some active sponsorship and saying, okay, this team,
(52:33):
this team, this team, we're gonna do something together to reinvent how we work together.
And it's gonna require some real effort, but it's gonna be transformative for us.
And so identifying some of those, it doesn't have to be everywhere in the business all atonce, but some of those that you think,
have particular interest for the business.
And you could say the work with Walmart was like that.
(52:54):
They identified that product design was something that was really important.
And to do that, they had to get a few different functions working together differently,right?
So you couldn't just leave it to the teams to experiment with AI and get to a goodoutcome.
And then the third piece I would say is how do you actually provide kind of theinfrastructure and graduating path to go from
(53:19):
good pilots and experiments to actually, how do you reorganize how the business works?
So this gets into talent, right?
How do you actually hire people?
How do you promote people?
Like, do you promote people because they're good at working with AI or not?
Like you're starting to see some consulting firms, like I think it was Accenture now,they're looking at integrating your usage of AI into whether or not you can get promoted.
(53:46):
Right?
there's all sorts of how do you rewire how the organization works?
Once you've started the usage, how does it redesign the way you look at hiring talent?
As well as how do you make sure that once you have deployed something, you stay fresh andyou use the new tools and the latest infrastructure so that people aren't tempted to go
(54:09):
outside of the company walls and use something that's kind of better.
for their purposes that's outside of the system, right?
And so this is really important because you don't want to get into situation where peopleare doing all of their experimentation outside of the company's both from a
confidentiality standpoint, compliance standpoint, as well as, we talked about earlier,all of that memory and that knowledge is lost to the company's systems in terms of the
(54:39):
tacit expertise that's captured by it.
and the way that the work could be reinvented, like all that kind of seeps out of thecompany's purview.
You just touched on something that I think is critically uh important and verymisunderstood.
So I work with a group of CTOs, a chair group that meets every other month and the groupdoes nothing except talk about AI in terms of the impact on their organizations, what
(55:04):
they've done, where they are, where they're going.
And one of the recurrent themes, because we just had a meeting the other week, was thatthey've got a lot of folks in the organization uh that are doing the whole vibe coding
thing, right?
And they're
They're getting minimum viable product and in many instances it's quite good, but they'recoming uh to the IT department, right?
(55:24):
Because that is where everything gets funneled right now for right or wrong in mostorganizations and they're getting very frustrated then because IT has got three, six, nine
month type uh lead times where they need to secure, harden, right?
So there's a traditional path there that is bumping directly up against what a lot of thefrontline employees want to do.
(55:46):
And so I think there's a challenge there for organizations to figure out how do you havethe security and the stability that you've always needed that centralized IT provides
while accelerating and not chilling the enthusiasm of the folks uh that uh are working onthese different ideas uh so that they then don't go, in your words, outside the system.
(56:06):
Yeah, that is, that is, that is a, that is a huge tension.
And I think, you know, what I have seen is, sometimes larger companies have been furtherahead of this.
Actually Walmart actually is a really good example of this.
So they actually have even, uh, like earlier last year, they were implementing effectivelylike a, internal vibe coding infrastructure so that whoever it was, if they had a really
(56:31):
good idea, they could experiment with the latest stuff, but it would be compliant.
And not just compliant, but it would actually be able to pull from the data that they hadand the systems that they have.
Right.
So it's actually a better thing you can create because you create something that'sactually connected to your systems and has the context of your business rather than just
(56:52):
being developed on the outside.
And they are trying to figure out how do I bring this in and make it of any use to mybusiness.
And so I think that's creating that kind of sandbox for experimentation is important.
but it also has to stay outdated.
And that's the real challenge is that you could do it one time and then some new thingcomes out and no one's using Claude code anymore because it looks outdated compared to
(57:18):
some new thing.
And you need to make sure that you can stay at the cutting edge, at least for some of yourpeople too.
Now I think that's a really uh good set of points, Jeff.
All right, so our time always goes very quickly on these calls.
So as we look to wrap up here in the next six or seven minutes, uh I'm interested in twothings.
(57:42):
Number one, I'm interested in middle market organizations that let's say they've gotanywhere from 10 or 20 million to a billion and a half in revenue.
How does the playbook for them change versus some of the enterprise level organizations?
What can they actually do with fewer resources?
Where's a good starting point there?
(58:02):
Yeah, so I think for a mid-market organization, the questions are the same in terms of howdo we actually figure out what we're trying to achieve?
What's the right mix of types of solutions that we might look at?
(58:22):
But the way you might get there is different.
So I think...
Overall, the vision of understanding, what are we trying to achieve with AI is important.
And I think it's important to pass apart.
Well, we might want to have some efficiency gains, but also what are our hypotheses on howwe could use this thing and work with this thing in a different way in order to do things
(58:46):
that we couldn't have done before?
And what do we think that looks like for us as a business?
So beyond like, okay, we could be more efficient at these tasks.
What are the new types of outcomes, the new types of services that we couldn't haveachieved before?
And what do we want to aim towards?
So I think that that's an exercise that every company should do no matter their size.
And I think that that, yeah, that's the first step.
(59:10):
The second step then is thinking about, well, what's the portfolio of different effortsthat I might need.
And so that can look, that can work at different levels, kind of as I've described.
So firstly, well, how do we kind of deploy some tools?
and give people enough infrastructure and knowledge to be able to use that in theirday-to-day work.
The second thing is how do we actually identify some select areas where we want to reshapeand redesign how we work, redesign some workflows, not just automate them, actually like
(59:40):
rethink what's possible.
And then the third piece is what are new value streams that we might be able to create?
And not all organizations are going to be there yet, but
at least having a hypothesis as to, could we use this to create new types of services?
Could we use this to offer a different level of value to our customers?
(01:00:02):
Is there something that could reshape the type of business that we become based on thedata that we have and based on the problems that our customers or clients have?
The how you get there though is different.
So I think
The first thing to understand is that for a large enterprise, in the five upwards billionin revenue, they will be hounded by large consulting firms all the time.
(01:00:30):
And so they'll have no shortage of people pitching them solutions, and they can get freeadvice from them as well, which is good and bad.
It's always going to come with some kind of attachment to some kind of solution, usually,in mind.
Um, so I think for a mid-market firm, the big challenge is what's the extent to which youshould think about using off the shelf tools and systems and what should you actually
(01:00:59):
build and maintain yourselves?
And I think that's the key thing that really is important to understand and where are theareas you want to develop a focused team or effort to have that capability and that
expertise in house.
to understand what are the new tools that you should be keeping up on so that you preventpeople from using outside.
(01:01:21):
And that needs to be within the business, not just within IT.
And then how do you actually start to cultivate some kind of system for looking atidentifying efforts and funding them and making sure that you're making the most out of
them rather than just...
(01:01:41):
you know, rather than just kind of like letting people experiment.
And so for a lot of mid-market firms that probably won't look like creating a complexcenter of excellence and having teams in different functions and different business units.
But you still need to make sure that you have enough of that expertise in house so thatyou can effectively evaluate vendors and also evaluate consultancies and advisors to make
(01:02:05):
sure that your m
really maintaining the value for your company too.
Because I think the risk is that as a mid-market company, you end up just becomingsubsumed by the roadmaps and the decisions made by SaaS vendors rather than shaping your
own destiny.
(01:02:27):
And many of those vendors may not be here in another few years.
Yes.
All right, this has been thoroughly enjoyable for me, Jeff.
As we conclude here, let's bring this back to where we started and let's just uh hit onelast question here about collaboration.
So if I'm a leader, I've been listening to this episode, again, bringing us back to wherewe started here today.
(01:02:51):
If I want to start shifting my organization from using AI to actually collaborating withit, what are maybe the first two or three things that I should do uh next Monday when I
walk back in?
Yeah, so I think as a leader, would think about three things.
One is how do I shift my mental model?
How do I think about if there was an expert form of intelligence sat next to me every day,how would I do my day differently?
(01:03:19):
What does that mean for me personally?
And how do I look for that as a way to improve who I am as a leader, not just get stuffdone?
The second thing I would look at is how do I find an area of my work that I want to get myhands dirty and do something.
Right.
So a little bit about, you know, people might be vibe coding using Claude code.
(01:03:44):
I think it's really important.
The best way to actually understand how to work with AI is to work with AI to do somethingand build it.
So it's not about like taking a prompting class.
Well, finding a problem that you care about an area of your job that really sucks.
Mm-hmm.
would agree needs to change and start doing something to see how you can change that.
(01:04:04):
Right.
And it might be that the outcome you get to is not worth scaling or not worthimplementing, but you will actually learn what it's like to work with this different form
of intelligence, how you can get good results, how you can get bad results and what youneed to do differently to steer more towards the good results rather than the bad results.
(01:04:26):
And then the third thing is to think about
what that means for how you show up and lead, right?
And so this is about your day-to-day interactions with folks.
It's about sharing some of your learnings with humility and making it comfortable forpeople to say, well, yeah, I'm trying to figure this out too.
As well as actually helping to embody that questioning mindset and that critical thinkingmindset in your day-to-day conversations with your team and with your peers.
(01:04:56):
so that you're cultivating that kind critical intelligence in your day-to-day work.
Wonderful.
think that is a great place for us to conclude today as we continue on our journey with AIfrom just treating it like a tool that we search for things to collaborating then to some
(01:05:16):
hopefully semi-autonomous and autonomous features here in the near future.
uh Jeff, I really appreciate your time and uh everything that you brought to ourconversation today.
for our listeners that want to learn more about you, your work, maybe they want to buyyour book, maybe they want to connect with you.
What are the ways and the paths for them to find you online?
(01:05:37):
Yeah, sure.
So my company website is human-machines.com.
And then I'm also on LinkedIn, and I have a weekly newsletter.
And people can reach out to me by any of those two means.
Lovely.
Jeff Gibbons, thank you very much for your time today.
Wonderful, great, thank you, Jan.
And for our listeners, thank you for listening once again to AI for the C-suite.
(01:06:00):
If this episode was useful for you, subscribe wherever you get your podcasts, follow us onLinkedIn and check out AI for the C-suite.com.
So until next time, keep your algorithms running, your leadership evolving and your AI incheck.
Take care everybody.
I'm Chad Harvey, and this is AI for the C-suite.
The show for senior leaders who know AI matters and need to figure out what to do aboutit.
Each episode, we dig into what it all actually means for middle market organizations, howit's changing decisions, strategy.
leadership and the nature of work.
Let's get into it.
(02:17):
You know what I love when someone walks in and says most of what you're hearing about AIadoption is wrong.
And that is basically what today's guest does for a living.
Jeff Gibbons is the founder of human machines.
He spent close to 20 years at the intersection of strategy, tech and innovation, workingwith organizations like Walmart, Nestle, Coca-Cola and Vanguard to figure out what
(02:41):
actually works
when humans and AI sit down at the same table.
Before his current venture, Jeff was a partner at Accenture, where he launched entirelynew businesses across multiple sectors.
He's also the author of Critical Intelligence, a book about strengthening human thinkingin the age of AI, and he holds a master's degrees from Oxford and Imperial College of
(03:04):
London.
What I find refreshing about Jeff is that he blends real optimism with real skepticism.
And he's got a framework for how your people should actually be collaborating with AI thatI think is going to change the way you think about this.
Jeff, welcome to the show.
Chad, great to be here and look forward to the conversation.
Thank you.
(03:25):
You have got quite an interesting background.
And as you and I talked about uh prior to recording this conversation, there is a realinteresting shift that I think a lot of leaders need to go through.
And you talk about that shift in terms of thinking of AI as a tool that you use tothinking of it as a collaborative partner.
(03:46):
And then of course, there are some implications to that.
I'm interested in starting with what that mental model shift
looks like for a leader that's used to managing and leading people instead of AI.
Yeah, great question.
So I think for me, you know, I think it's easy to think of AI using the mental, mentalmodel of a tool, right?
(04:07):
So sort of like a search bar, right?
So I can put in a question, you know, what's the best X or how do I do Y and get an answerout of it.
And I think for a lot of people, when we first started using AI tools, whether that was ayear ago or three years ago or whenever it was, that was kind of the first frame.
(04:28):
of how you use it, right?
It's sort of like a better search bar.
And that leads you to think, well, how do I get the perfect prompt?
How do I construct it right?
So I'm going to get the right output.
And then in terms of what I get out of the other end, I'm thinking about it in terms of,how good is that?
Are there errors in that?
I'm primarily thinking about how good is the tool?
(04:49):
I think what's interesting is that
I think a lot of people are starting to see that AI can actually do so much more and theycan achieve so much more in their jobs when they think of it as a collaborator.
So rather than thinking of it as a one-time quick hit, how do I put in a question and getan answer?
It's about how do I do my job differently by imagining there was somebody sat next to meevery day who said, Hey Chad, I am right next to you.
(05:17):
I am very skilled in all of these ways.
have nothing to do today.
Ask me for some help on these tasks and let's collaborate.
And not only will you Chad, like give feedback to me, but I'll give you feedback too.
I'll help you to be better.
Right.
And so thinking about AI as a collaborator who can actually help you be better, who cangive you feedback on your work and help you understand biases and blind spots in your
(05:44):
work, I think gives you a different frame of how you can work with AI and actually helpsyou to achieve more too.
It's so interesting when this journey started in the the wider world uh and everybodyloves to talk about chat GPT but as you and I know AI is much more than that but that was
kind of the uh origination point for a lot of conversations with folks.
(06:06):
I used to say something along the lines of it's kind of like working with an absolutelybrilliant you know college junior that has access to all the information in the world but
lacks all common sense and real-world experience and
I feel like we've evolved significantly beyond that to the point where it's now maybesomeone that's got a master's or a PhD level and maybe a few years of experience in your
(06:29):
business.
Again, some of those same limitations, uh but the ball has significantly moved or uh thecapabilities have advanced.
And so when we think about that, I love your idea of this being somebody sitting rightbeside you that is just saying, hey, give me something to do or bounce some ideas off of
me.
(06:49):
So when you walk into organizations and you work with executives and leadership teams,what do you find their biggest limitation is in terms of thinking about these tools as
collaborators versus that standard search box?
Yeah.
So I think, I think what I often see in organizations is that there will be a desire, youknow, one intention, good desire to think about, well, there's this new thing we should,
(07:15):
we should, we should do that too.
We should roll out some kind of internal copilot system or enterprise chat, GBT or Claudeor whatever it might be.
And we think that will be a good way of helping us to make our people able to do theirjobs more efficiently.
you know, save some time, all good ideas and totally true.
(07:38):
And that's, that's possible.
Typically there'll be a rollout.
People will get some kind of compliance focused training just to make sure they don'tupload anything that's particularly sensitive or, know, kind of the general rules of the
road, mostly from like a legal and confidentiality standpoint, although, you know, a lotof that is really covered now by the, the sort of frameworks and contracts that you sign
(08:00):
with an LLM from an enterprise standpoint.
So.
People will get that and then they just kind of get often left to figure it outthemselves.
And I think people default to how do they use other tools, right?
And it's easy to understand why, right?
So you can think of it as a search bar, right?
You can ask it questions.
You can think about, well, this actually can do some writing.
(08:23):
And so, you know, I can improve my emails.
can get it to draft something.
And so you sort of think about, what are the simple tasks I can get it to do?
What I think...
actually becomes a bit of an unlock for people is thinking about there are lots ofdifferent modes in which you can work with AI.
And so that's one of things I try to unpack in the book, Crystal Intelligence, and in mywork with companies is to think about how do you understand what's the right mode that you
(08:50):
could be working with AI in this?
And how do you think about the right mode for different types of work?
I think the, so there are four modes generally that I talk about.
So the first mode is consulting mode.
And so this is kind of like treating it like a search bar.
So as a human, you're kind of driving the task.
You're thinking about what am going to do?
But you're asking AI for some input.
(09:12):
So you might be saying, what do you think of this email?
What's the best X for Y?
And you get an answer as a result.
The second mode I think of is where you're in approving mode.
So you are going to give it a task.
please come up with three or four different ideas for X.
(09:34):
And then you approve that and you say, okay, great, let's go with this one.
m And I think that's a deeper form of collaboration, but where it gets more interesting iswhen you actually start to then enter into a dialogue with AI and actually say, okay, so
let's take these three or four and let's build them further.
(09:56):
Let me...
give you some constraints and some more context so we can make it better.
And then let's work as partners to do that.
The third mode is what I call supervising mode.
So this is where you're actually giving more of the involvement in the work to AI itself.
So you might say, give me three or four ideas for this, and then actually flesh out a riskmitigation plan for each of them.
(10:22):
And then figure out what's an implementation plan.
And then
give me feedback, and then I'll give you feedback on all of those things.
And then let's improve that plan together.
And so you actually work, you're kind of acting as a supervisor or a director of the workthat AI is doing.
And then the last mode is what I call delegating.
So this is where really the AI is going to do things for you.
(10:47):
And you're going to set the parameters and review its performance over time.
But this is really useful for situations where
It's fairly low risk and the AI is going to be better at doing it than you.
So let's say for example, this might be, Hey AI, I want to be on top of trends in thiscertain area.
(11:09):
Please every morning look for any news around this and give me an update and tell me thethree things I should learn about this.
That will be useful for my job.
Right.
And so you're actually kind of delegating the work and saying,
You go and do it.
Give me a bit of a feedback report on what's going on, but you just kind of go and dothat.
(11:29):
Or it could be that you're actually letting it to letting it do work for you.
could be that you're going to say, can you please send these three or four emails a day toother people?
It could be that you're going to say, here's my, you know, here's my household budget.
Please create a grocery list for me every day.
Like, you know, outside of the world of work, you're basically letting it do tasks.
(11:54):
and just give you an update on how they're going.
I want to dig into these four uh distinct modes of AI collaboration a little bit more.
But before we go right into that, I want to go back to where we started with the issue ofthe leader or the leaders of an organization.
Because I feel like a lot of leaders of organizations are very hot on AI for theirorganization, but they're not really uh leveraging any of these four different modes uh in
(12:24):
their personal work.
let alone thinking about how they do their work so that they can model it for their team.
Agree?
Disagree?
Thoughts on that?
Yeah, I think there's a lot of leaders today who are thinking, number one, I don't want tobe left behind by this.
And so I think there's a lot of FOMO out there and people are thinking about how do Ifuture proof my own career and what does that mean to me?
(12:47):
And I think in working with leaders, one the things I like to focus on is that I think alot of people in developing their careers have become really good at
answering questions and being known as the person who has the right answer.
Right.
That's why you get ahead.
That's where you get promoted.
And I think what's challenging for some people is to think actually maybe the AI can dothat just as well as me, right.
(13:12):
And maybe even better in some ways, right.
And actually, especially if you think about it in terms of AI being able to answerquestions that are outside of your domain expertise that you might not know about.
Like if, you know, if you're a finance executive,
AI might know more about R &D than you, right?
And so there's a sense of, well, I mean, what do I really excel at if AI can give betteranswers than me?
(13:36):
And so what I try to focus on with people is you need to focus on what are the rightquestions that you should be asking, right?
So in a world where you can get an answer to any question in seconds or minutes, reallywhat matters is what are the questions you ask?
And then the second thing is
Once you've asked that question, how do you actually work with AI collaboratively to getto an output that you couldn't have ever done before?
(14:01):
Right.
So how do you actually understand the best way to get the best output together?
Um, by using the best of your skills and your domain expertise and the analytical skillsand research skills of an uh AI system, which are able to surpass what you can do as a
human.
And so how do you actually use the best of those things?
(14:24):
And so that's one of the things I've been working on is how do you actually train and helppeople develop and measure their skills around this principle of collaboration rather than
just prompting or getting a good answer.
Right.
So it comes from how do you actually identify what's the right question to ask?
How do you provide the right context?
(14:44):
How do you provide constraints?
How do you understand how you can work with AI to figure out, is this a good answer?
Could we do better?
Can we analyze the options here?
Can you, AI, understand my blind spots around this and help me make a better decision?
So it's all about how you can work with AI to amplify what you can do as a human, not justget a good answer.
(15:09):
Do you find that a lot of leaders want to put that prompt in or give the task and they areassigning it in a way that does not allow the AI to ask questions and to clarify the
intent?
Cause I see that sometimes with folks, are used to, I don't want to say barking orders,but they're used to dispensing direction.
(15:36):
And I think that that leaves a lot of opportunity on the table per your point aboutworking with these tools.
Yeah, no, absolutely.
think that can happen a lot.
And I think it comes down to two things.
One, I think is it's partly the mental model, right?
If you're not thinking of this as a collaborate, so you're just thinking, basically, howcan I get what I want from this thing?
(16:00):
Right?
You're just thinking, what can I get out of this?
And I'll be the judge of whether or not it's good.
Right?
ah I'll use my best judgment.
You're not thinking about how could it actually make me better?
you know, as a humility to thinking about it in that way, right?
The second thing actually is that, so that happens a lot.
And I think the other reason for that is it's actually about the level of context that youprovide.
(16:25):
So one of the things I've been doing over the last month is actually developing a systemfor measuring and helping people understand how well they actually collaborate with AI
systems.
And I ran a study actually.
So, so about a month ago, I ran a study with about 450 people roughly who were eachworking with an AI tool to define a problem.
(16:49):
So real people working with an AI tool to define their own problem and then work through aseries of conversations with that AI to solve the problem and get to a good solution.
Right.
And so the AI system was instructed to
You know, dig in and understand the question, give suggestions, but then also flip thescript a bit and work with the user in a different way to ask them for suggestions, to
(17:16):
give them feedback and see how the person would actually react to different styles ofcollaboration.
Like those, those formats that I talked about.
So using this, uh, this tool and this study, so there are about 450 people ran throughthis and they had a conversation about 40 minutes in length.
I was able to understand like really what was going on with these people.
So very rich data set with thousands of messages.
(17:38):
And one of things that was really interesting was that managers versus individualcontributors behave differently.
So managers generally provided instructions and context that was a little bit terse andmore sparing as you described.
Right.
So it's actually about the, you you might just say like, do this or please give me threeof these.
(18:01):
Right.
And.
versus individual contributors were more likely, I guess, they were, you know, potentiallybecause they're more used to having to kind of explain themselves and provide context even
for a question, they were more likely to provide better context to the AI.
And they actually scored overall better in terms of the quality of the output that theygot to working with the AI.
(18:27):
Very interesting.
A recent episode where I chatted with Dr.
Sam about some of her methodologies, we got into this distinction between managementversus leadership.
I purposely used the word management during one question as opposed to leadership to seehow she'd respond.
But we really didn't unpack this issue of collaboration.
And I think those are the three um areas here in terms of how we're working.
(18:53):
with these tools and within organizations that I think are worthy of focus.
The idea of management versus leadership versus collaboration.
And I think there's some spillover between those, but you raise a really interestingpoint.
And one of my thoughts is that we're rapidly moving into an era where everyone is a leaderof something in the organization, whether they're leading other humans, whether they're
(19:15):
leading AIs or both.
And many of those...
uh
think relationships, for lack of a better word, are going to involve the type ofcollaboration you're talking about.
Yes, absolutely.
And I think, you know, right now for most people, the relationship with AI is sort of likea one-to-one relationship, right?
I'm just chatting with my chat bot thing.
(19:35):
think that's starting to change in that now you, know, if I'm working with Claude, it willcreate a little team of agents.
And so even just in the language it uses in user experience, I'm thinking about itdifferently.
I'm thinking about it as.
you know, a team of little agents that's going off and doing different things and comingback to me.
So I'm starting to conceptualize it a bit differently, even though it is just a chat botthing.
(19:59):
It's like, it's actually like little entities doing their own thing.
Uh, but I think the shift really will come from when the user interfaces start to lookdifferent in a way that you and I could both be collaborating in a chat environment.
with an AI or multiple AI agents at the same time.
(20:19):
I don't think we've really seen that emerge, but I think that will be the driver of thatmindset shift.
And I think it's a question of how comfortable will people feel with that and how do youactually make that something that gives you better amplified outcomes versus you just
having your own little chat.
I couldn't agree more and I think uh from my personal and professional experience, I havebeen very wary over the last several years of especially with client data exposing too
(20:51):
much in the public models even if they are secure.
So we spun up our own secure uh environment and we've just recently enabled the kind ofpervasive memory that allows it to understand the individual users.
So you're getting
what you've had in the public models in a very private setting and I haven't had thatbefore with confidential client work, but I've had it in the personal setting when I don't
(21:13):
care whether you know something gets leaked out and I'm saying this and laying the tablehere because I think what you're starting to uh drive towards here, which is fascinating
towards me, is this idea of a pervasive kind of larger entity with the subagents thatunderstands all the different projects that I'm working on.
(21:33):
Maybe not thoroughly, but it has knowledge that can be applicable.
So when I'm asking for research, yeah, yeah, yeah, yeah, exactly.
And, I think you're onto something here, Jeff, that I don't think a lot of people arefully appreciating yet about where this game is going.
Um, and so I don't know that I've got a question here necessarily, but what you're sayingreally resonates with me.
(21:55):
And I feel like this is the type of maybe not bleeding edge, uh, usage with these tools,but leading edge usage.
that more folks need to start to wrap their heads around.
I think it has some really interesting human and organizational implications too.
So if you think about it, let's use an analogy, right?
So people working in a job, they have their own professional networks, which are theirown, right?
(22:20):
They don't belong to the company.
And in the same way you, know, connections, whether that's LinkedIn connections or emailaddresses are just people that you know, and trust that you have with other people in the
same way, people are building very sophisticated.
memories in AI systems about who they are, what they know, that, know, the AI systems areunderstanding this tacit knowledge.
(22:44):
Sometimes that's in their own personal AI accounts and they're using that for kind of, youknow, for work things as well.
This kind of shadow AI usage that you, that you've probably heard of a lot.
And then also then they're doing that in their own company systems and the tacit knowledgethat these people have.
(23:05):
And their understanding and their expertise is really valuable.
And so then I think it becomes a question of how do I actually save that for myself?
How do I understand that myself?
Who owns that?
Like, how does the company benefit from that?
There's like an interesting thing to understand there.
And all of that's just talking about an individual.
When you think about the level of teams, right?
(23:27):
So if somebody was to retire for a company or leave a company, how do you think about
carrying over that tacit knowledge and expertise from that person.
That's different.
Like, you know, currently people think about it in terms of like, well, how do I get theirfiles or their contacts from their Outlook calendar?
It's a very different thing to think about how do I actually store and codify and protectsome of the tacit knowledge and workflows and things that they've been doing for years,
(23:58):
which we could benefit from.
And then how do I, an individual think about how I can maintain that to myself?
Yeah, no, this this gets directly to the heart of something that I've been talking aboutin a lot of workshops and in public speaking over the last couple of years, which is when
you've got a general purpose technology, it reshapes everything, uh including law.
(24:18):
And my head as uh a as a classically trained attorney that does not practice, don't callme for advice, uh goes immediately to what is the right of the individual worker like you
laid out with respect to that tacit knowledge?
and where is the line between what I know and what I bring to the organization versus whatthey're now able to capture ah and begin to deploy.
(24:41):
Yeah, absolutely.
It's actually one of the things I've been building.
So based on this study, I've been building a tool that measures your AI collaboration.
So it's called Corex and I can, I can share with you the, the details of the, of the tooland the assessment methodology.
But really what that does is it works with you.
So you can work with it in a live chat assessment, as I described with those sort of 30minute conversations.
(25:04):
You can also actually use it to analyze your past conversations to unpack and
and go through and analyze, well, what's the quality of the collaboration?
How is it changing over time?
Is there evidence that you're improving or learning or taking on new concepts?
And in the design of that tool, it's been really important to think about what does theindividual keep and what's visible to a manager, a training company that might be working
(25:31):
with them on it?
What's the boundaries?
I think from the work I've been doing so far, I think what's really important is that the
The conversations themselves seem to be something that are something that needs to beprotected and private to the individual and the employer too.
But the overall kind of signals and implications of it are something that could be shared,but never at the granular level of like, here's the specific thing that Chad was working
(26:03):
on or doing, but here's the kind of general strengths.
and opportunities that Chad has within this.
And then, you from that, I've been developing a dashboard that teams can look at to see,how well are my people able to collaborate with AI, but that's aggregated and anonymized,
right?
So you can't see how well does Chad do versus how well does Jeff do.
(26:23):
You can see at the aggregate level, how is my team doing.
Well, this, you you gave the example earlier in terms of uh when, I can't remember if youmentioned when someone leaves or when someone starts, but basically talking about how we
think about that traditionally.
And what I'm wondering about here is whether any organizations are starting to grapplewith the idea of what information from our system that the human has been collaborating
(26:50):
with when the human leaves, do we need to provide some type of portability or export orpurge?
uh in a more extreme example from the system.
Like what do I have the right to as uh that individual employee uh to know about myselfand the way I work and is there a way to take some of that with me to my next job, maybe
(27:11):
to see my own agents.
Yeah.
So I think, um, two things I can say.
One is I, I've seen companies start to look at that at the enterprise level, right?
And this is actually something I've been doing research into myself as well is how do youactually understand, um, what's the, what's the kind of decision-making culture of an
(27:32):
organization and how can you make sure that your AI systems are aligned with that and makesure that your
fine tuning and instructing the AI to work within the parameters of that.
But that's more at the enterprise level, at the company level.
What I have also seen, um even actually like two, two and a half years ago at my previouscompany, somebody actually put into their employment contracts a term, like a term in the
(28:05):
employment contract that they couldn't...
The company wasn't allowed to recreate an AI version of that.
So this was like two, two and a half years ago, right?
So I think it's probably coming.
That was a very early example of that.
We were quite surprised by that and we agreed to it.
But I think it's a really interesting emerging area.
(28:26):
I haven't seen companies grapple with it at much scale.
Yeah.
I think the tool that you've talked about establishing in terms of measurement andcollaboration is really interesting.
If I'm a mid-market CEO and maybe I don't have the type of resources of an enterpriselevel organization, and I came to you with a question because I'd heard this episode and I
(28:50):
said something like, I need to know if my investment AI is paying off just beyond licensecounts.
What would you encourage them to track?
Is there a way for them to, without a more sophisticated system, begin to gauge the levelof collaboration that's going on as well?
Yeah, sure.
So the Corex tool is designed exactly for that.
(29:13):
So the goal of it is that there are two ways you can analyze your skills.
The first way, as I mentioned, is actually having a live chat assessment.
And so that's for an individual to have a conversation with.
I think the main applications I'm thinking about for that are you could use it to assesspeople you might be looking to hire.
You could...
(29:35):
use it at the outset of a training program or just even to design a training program andunderstand what are my needs, as you said.
So that gives you sort of a real problem solving challenge and then you work through andget advice on how well you do.
And that's something you can run across your team.
And I've started to see there are similar types of efforts going on in hiring assessmentsacross other companies too.
(30:01):
The second, which is, and so that takes about 30, 40 minutes for an individual to do.
The second form of it, which is a little, little faster and easier too, is that you canbasically log in and you get a really long prompt, which instructs your AI tool of choice,
whether that's, you know, chat, GBT or copilot or Claude to analyze your pastconversations through the lens of a range of different metrics.
(30:29):
I can tell you what they, what they are.
and then give a report back to the tool and then it generates some analysis.
And in doing that, it strips out the actual content of your conversations and anyconfidential information and just gives you an analysis of, when you're doing X, Y, Z
things, here's what you can improve, here's what you can do better at, and here's thetrend over time over the last X weeks or months of you using it.
(30:55):
And so what it looks at is,
Overall three main areas.
the, so the three Rs results, relationship and resilience.
So the first area results is I'm actually getting good results from working with AI.
Like is the output actually really good?
And is it better than a human could actually do by themselves or an AI by itself?
(31:16):
So when you're actually having this live chat conversation, one thing it does in theanalysis is it looks at different branching points in the conversation.
And then it simulates out if the conversation had stopped there, the human had stopped andthe AI had just continued talking with itself, what would the output have gotten to?
(31:37):
And then it compares where you actually got to.
And so it uses that to understand what's the value of the human judgment that you've addedin the conversation.
And it does that at different points in the conversation and figures out not only what'sthe value that you've added as a person, but also when and where did you actually add
particular value?
So then you get an analysis showing, yeah, actually when you, when you provided thatconstraint or when you add the extra content chat, you know, this is when you actually
(32:06):
added a lot of extra value.
And so you can use that to get a real feedback on, your conversation style and, where you,where you sort of did well and didn't do well.
So that's sort of the results.
The second piece is the relationship.
So what's the quality of the dialogue?
So are you pushing back enough?
Are you actually providing.
the right kind of context.
(32:26):
So one of things that was really interesting from the study with about 450 people was thatthere was a general trend.
People actually in their fifties were the highest scoring age group.
People in their twenties were the lowest going age group.
And the reason for that was that people in their twenties were much more likely to justpassively accept the output of the AI and be like, yep, right.
(32:51):
Sounds good.
Okay.
Tell me more.
Or let's go with that versus people who were more in their fifties, they were more likelyto engage, provide constraints, say, I'm not sure that actually works.
Here's a nuance on that.
And so this is like a behavioral thing.
It's very trainable, but had a noticeable impact on the quality of the output.
(33:13):
Can I pause you there before we get to the three Rs?
That distinction is fascinating because one of the things I keep hearing over and overfrom folks is, well, I just need to hire somebody that's younger that can absolutely be
brilliant with our AI tools.
And what I'm wondering here is whether that contrast in the generations is a issue ofexperience and seasoning, or is it a true cultural generational change uh in that?
(33:39):
the 20-somethings are more digitally native uh and are used to a different set of, uh Iguess I'd call it, and usage with uh digital tools than folks in their 50s.
I'm interested in what your thoughts are on that.
Yeah, I I think it's probably a little bit of both Chad, but I think maybe more thelatter.
So I think it's partly about having more experience, being more seasoned and maybe havingmore context.
(34:04):
And I think that's often what people say.
Well, yeah, people, people with more experience are going to be able to provide moredomain expertise.
But I think actually the data showed it was primarily more that, you know, there werepeople in their twenties who were able to do that, but the, it's more of a sort of like
the cultural usage, which is right.
So I think.
And this is, you know, this is a hypothesis.
(34:26):
This is, it's hard to know the exact reason, but my assumption would be that there's justa general experience for people who are younger, more digitally native, who have grown up
more with AI tools, whereby they are more used to the sort of passive response andaccepting of the output.
(34:47):
Uh, whereas people who were generally older, they, they just
work with it a little bit more like you would in a conversation as opposed to working witha tool.
So it was more that kind of culture of, well, let's actually like work with this thing.
Let's collaborate more actively rather than just using it as a thing you can do, you know,have to run a task and say, great, now do the next thing.
(35:13):
Thank you for pausing and interrupting your flow there, because I think that particularexample goes directly to the heart of a lot of conversations we need to be having about
the nature of collaboration in the new work environment and also our personal environmentas well.
mean, you know, there's that old joke on the Simpsons.
uh I, for one, welcome our new robot overlords, right?
We need to figure out how to work with these systems.
(35:36):
And if there are very real generational differences,
in output and the way we're interacting with them, they're showing up.
I just thought that was a very interesting data point.
So thank you for that.
uh Your third R, we talked about results and relationships.
Yeah.
So the third R is resilience.
So this looks at a few different things.
overall it looks at, you building skills?
(35:56):
Are you building?
And so as an example, in the live chat assessment, what it will look for is, is thereevidence that I'm actually taking on an understanding and building upon what the AI has
told me or what solution we've come to together?
Or am I just kind of like, like, or is that like, I'm actually understanding this?
(36:16):
Am I developing new?
skills and understanding as a result.
Also looking at your past conversations over time, it looks at, you building skills?
Are you getting better at your AI collaboration over time?
Are you understanding new concepts over time?
Right?
Is there a sign that in working with AI, you are picking up new things, developing morecomplex framing of issues over time?
(36:42):
This is something that I think is really important because we talked a little bit aboutearlier.
You could be a leader, let's say in finance, right?
Like you have very deep functional expertise.
One of the things that's really important is that AI can actually help you get better atthings outside of that domain of expertise.
Right?
So like R &D was the example I used earlier.
(37:02):
Right?
So what's really interesting is to look at, you know, for a finance leader, are theybetter able to understand and grasp and grapple with concepts outside of their
area of expertise and is that growing over time?
Because that's really one of the most beneficial things is that you become less siloed asan expert and more of a, I don't want to say a generalist, like you just have broader in
(37:26):
your expertise.
Mm-hmm.
I think that out of all of those three Rs, I don't want to say one is more important thanthe next, but that one resonates, the issue of resiliency resonates with me the most
because I think there's a temptation for us to just want to lean into our strengths anduse some of the AI tools to get even better in that zone.
(37:48):
But to your point, knowledge and the value of knowledge and expertise is rapidly gettingcommoditized.
My words, not yours.
I think that if you are able to understand that and to become more of that generalist andto build that resiliency, then perhaps you can begin to work with these tools to help
(38:10):
assess some of these different domains through your particular lens, right?
Again, going back to that idea of the smaller agents coordinated by the super agent andunderstanding who we are, what we do.
I think there are ways to work with these tools in a collaborative environment that domake you more of a powerful generalist.
and add more value to the team as at the same time, the value of traditional uh knowledgesets is perhaps decreasing.
(38:37):
Yeah, there's a huge issue with the sort of the cognitive atrophy.
And also I think just really an understanding the work that you're delivering.
It's so possible now to produce, you know, this would be the same in educational contextsand work contexts to produce a beautiful presentation or a great essay, but not actually
(38:58):
understand this.
And I've seen this myself, actually.
I was at a conference a few months ago, about five months ago, and I saw
a very seasoned executive deliver a presentation and it was clear he did not understandactually what the presentation was.
It was clear that the AI had written the presentation and it was actually great content onthe slides, but it was clear he didn't understand it and didn't have the fluidity to be
(39:23):
able to answer questions about it as a result.
I haven't seen anything that extreme, but I don't have difficulty envisioning that at all.
It was really bad.
It was really embarrassing, actually.
Oh, just because you've got uh the world at your fingertips and the ability to dosomething like that doesn't mean that you should, right?
(39:44):
Yeah.
And it's back to this issue of like, what happens if you don't have the tool, right?
Like if someone asks you a question outside of the meeting, you need to be able to answerthat, right?
You can't just be like, wait, hold on second.
Let me go and ask Chachi if you see the question, the answer to that, right?
You need to actually have the understanding yourself, not just be able to get the thing tospit out the thing.
(40:05):
And so that's the important thing of how you building skills and knowledge through this,not just getting out.
We've been talking an awful lot here about knowledge work and I know because I haveprivate clients in this space, but also we get a lot of email from, from listeners that
are in more tangible uh trades, manufacturing or uh construction, things like that.
(40:26):
So let, let's shift gears a little bit.
know you did some work with Walmart and I believe the system you worked on with them wascalled trend to product that changed.
Yeah.
Okay.
Good.
I got that right.
Changed the way that they design and ship products.
So I'm interested for our folks that work in more tangible areas, not just knowledge work,because there's an intersection here between the knowledge and the tangible.
(40:48):
Can you walk us through what trend to product looked like and what was the starting point?
How did humans and AI divide the work?
What changed in terms of outcomes?
I know I'm throwing a lot up there, but I'm fascinated by this intersection between the AIin knowledge work and then the tangible world that a lot of our listeners live in.
(41:09):
Yeah, sure.
So, sorry.
So, so the work, so the work really started over two, two and a half years ago, looking athow would you actually introduce AI into the product design process?
So Walmart, they produce their own private label brands and products across a range ofdifferent categories.
(41:30):
And the conversation started with where would we apply this and what would be the rightcategory to start with?
So they.
You know, they produce everything from clothes to food to furniture to everything, youknow, across all of the realms of everything you can think about buying in Walmart.
And so we started out with the idea of working in fashion and clothing, partly because itwas a very difficult area.
(41:58):
So actually designing a garment is not easy, right?
There are lots of different parts to it.
You have different suppliers, you have
different components, and just the process of designing something that actually isconstructable is technically feasible is actually quite difficult.
(42:18):
And so we started with that as the area to start with.
And what we looked at first of all was understanding what's the current process for, Iwant to do a blue dress to that blue dress being on a shelf that somebody can actually
pick up or rack in this instance.
And so that process is quite arduous, involves lots of different steps, lots of differentteams, different suppliers, and about a year actually.
(42:47):
So working as a designer, you're having to forecast and use a crystal ball to understand,what could I do that people will want a year from now, or maybe nine months, but basically
a year, depending on the type of garment you're looking at.
And so we mapped out.
what were all the different tasks in the current process?
(43:07):
But instead of then looking at, what are the point solutions to automate all of thosesteps?
We looked at, what were the real outcomes that we're looking to achieve here?
So we're looking at, we want to achieve cohesion across a line.
We want to make sure that we're being efficient.
We want to make sure that we're only producing 20 different garments rather than 30.
(43:29):
Right?
You know, like let's satisfy as many needs as we can with the fewest number of items.
We want to make sure we're being efficient in terms of the use of suppliers and the use ofdifferent fabrics, but also staying on trend.
And so what we identified was a new working system whereby the tool would be picking up onthings going on on TikTok on the runway.
(43:54):
would then generate mood boards and inspiration, you know, know, a
broad range of design inspiration for the teams to analyze.
There would also be a team of agents that would be supporting in that in terms of acreative director agent that would assess, how well was this in line with how a different
(44:14):
brand worked and all sorts of different parts along the process.
But the key thing was actually, how do you design the system that gets to a workable techpack, as it's called, that is
something that can be translated from a computer screen to something that a factory canproduce and do that at scale.
(44:35):
And so that required a lot of work with suppliers, with merchants, with designers workingtogether to redesign what their job would look like.
And so was less about how do we automate these things, it's more about how do we redesignthe workflow.
And now they have managed to successfully launch that, they're shipping clothes every daywith
(44:56):
using this new system called Trend to Product.
And as a result, they have built a capability and an intelligent system to understand andbring back feedback from the market into the design process.
And so that's the real innovation is that they own all of the intellectual property aroundthe garments, but also understand from the system, how to feedback design and demand
(45:20):
information into the next design and the next design and the next design.
That is absolutely fascinating.
mean what you're talking about here is Pattern mining and trend spotting over a large anddiverse set of inputs I mean you mentioned tick-tock and some other things there coming
back in Yeah, that is absolutely wild and I'd love to know Clearly we don't have I guessthe the data or the timeline yet.
(45:48):
I'd love to know how an approach like that contrasts with um
the accuracy of more traditional approaches because my sense of fashion is that somebodykind of goes like this ah and you know senses which way the winds are blowing and then
they make their bets in terms of the production.
Yeah, so I think that, yeah, that is true a lot of the time.
(46:10):
And I think, unfortunately, in the fashion industry, about 20 % of all clothes that areproduced end up in landfill or burned.
So this is not like 20 % ends up in TJ Maxx or being sold after the fact.
It's burned or put in landfill, just because it's hard to predict.
So this was one of the reasons to select that as a category.
(46:32):
um
you're never going to have a hundred percent hit rate.
But the key thing was how do you design enough feedback loops early on in the process inso that you could actually improve the predictability and also shorten the time between
the prediction and actually production, right?
(46:53):
How do you shorten that time so that you can at least have a better view of what's goingto be trending and then also where is it trending, right?
So a certain trend in culture
might manifest itself in different parts of the US, you don't want to produce it in Oregonif it's really only a thing in Virginia, right?
(47:14):
So there's things like that you need to understand too.
Wow.
what role, uh no, that's probably the wrong question.
I'm interested in the role that human designers play in the new system and how their roleschanged and where that, you know, you hear people talk an awful lot about human in the
loop.
I'm interested which parts of that loop or which part of that process do the humandesigners play a role in and how does that compare to what they were doing previously?
(47:41):
Yeah, so I think one thing that was really important to understand was that the humanswere involved in the redesign process.
And I think it was framed as, let's not design a tool, let's redesign what your job lookslike.
And so it's a different framing.
so for them, actually, they were very involved in helping the team understand, what dothey do today?
(48:06):
What are all the things they find a pain, frankly?
what are all of the monotonous tasks and the steps that they wish they could skip andunderstanding all of that and then using them as really testers of the new system, both
the actual tool itself, but also the way of working, right?
And so they were critical in that.
(48:27):
And they were very engaged in that because I think the framing of how do we redesign yourjob to do things better, faster.
was more appealing than how do we just automate your job, right?
That's a different framing and they were actively engaged and I think enjoyed thecollaboration as part of that because they could see how it was actually gonna make their
(48:51):
job more enjoyable too.
And I think there's a factor in that too when you're working in an area where people arefeeling like that risk of automation is there, there's a desire to future-proof yourself.
be part of the leading wave of how things are going to change, both in terms of shapingwhat your job looks like, but also being somebody who's ahead of the curve.
(49:17):
There's a future-proofing element that strengthens your resume as well.
That future proofing piece, I see an awful lot of organizations out there uh throwingaround some version of the idea that we are going AI first, right?
We're an AI leader, we're AI first.
And I know that you are very focused on outcomes first, AI second.
(49:41):
So I'm interested, have you ever gotten some calls?
I know I have, but have you ever gotten calls from organizations that desperately want toadd AI to an existing process?
And when you get in there, you realize that the entire workflow needs to be redesigned.
Yes.
I've also, think I've also myself been guilty of that as well.
(50:04):
Right.
And so I can think of, I can think of projects I've done with clients where that wasreally what we did is we automated an old process rather than fundamentally redesigning
it.
And so that's why I I've taken this view that you need to really think about what's theoutcome, not how do you automate the tasks?
Cause I, there are projects I've done in the last
(50:27):
three years, probably sort of, you know, in the sort of two to three years ago timeframe,when a lot of this technology was newer, where that was really what we did.
We actually automated the existing workflows meticulously and to a high level of quality.
But I think it made me realize how much we were missing out on by not redesigning the waythat the work could be done.
(50:52):
And I also would imagine you've walked into a number of organizations that want to deploysome type of AI and found that more traditional solutions are actually in their better
interest.
Yeah, well, absolutely.
I think, um, yeah, there's another, another project I'm thinking of with a client where webuilt this amazing AI tool, but it was, it was something that people didn't really need,
(51:13):
right.
And it didn't get used and the existing ways of working were clunky, but fine.
And there wasn't the need to change that there wasn't sufficient need to change.
Yeah.
I think that's a good hinge point to pivot into a related area, which is common mistakes.
(51:36):
So organizations that try to roll out AI collaboration at scale, you know, they've donethe pilots, the leadership's bought in, they've done some training, but then something
stalls out.
In your experience, where do you see this break?
Yeah, I think it can break it a few stages.
So think the first thing is just releasing access to the tools without active guidance onhow to get the best out of them.
(52:03):
Right.
So what kind of tasks could I be doing with this?
You know, beyond just getting a good prompt out there, like how do I actually bestcollaborate with this?
Some of the things we talked about, I think the second layer is it's great to have peopleusing and experimenting, but
how do you actually identify, what are some real opportunities to reinvent workflows orreinvent how we work that require some active sponsorship and saying, okay, this team,
(52:33):
this team, this team, we're gonna do something together to reinvent how we work together.
And it's gonna require some real effort, but it's gonna be transformative for us.
And so identifying some of those, it doesn't have to be everywhere in the business all atonce, but some of those that you think,
have particular interest for the business.
And you could say the work with Walmart was like that.
(52:54):
They identified that product design was something that was really important.
And to do that, they had to get a few different functions working together differently,right?
So you couldn't just leave it to the teams to experiment with AI and get to a goodoutcome.
And then the third piece I would say is how do you actually provide kind of theinfrastructure and graduating path to go from
(53:19):
good pilots and experiments to actually, how do you reorganize how the business works?
So this gets into talent, right?
How do you actually hire people?
How do you promote people?
Like, do you promote people because they're good at working with AI or not?
Like you're starting to see some consulting firms, like I think it was Accenture now,they're looking at integrating your usage of AI into whether or not you can get promoted.
(53:46):
Right?
there's all sorts of how do you rewire how the organization works?
Once you've started the usage, how does it redesign the way you look at hiring talent?
As well as how do you make sure that once you have deployed something, you stay fresh andyou use the new tools and the latest infrastructure so that people aren't tempted to go
(54:09):
outside of the company walls and use something that's kind of better.
for their purposes that's outside of the system, right?
And so this is really important because you don't want to get into situation where peopleare doing all of their experimentation outside of the company's both from a
confidentiality standpoint, compliance standpoint, as well as, we talked about earlier,all of that memory and that knowledge is lost to the company's systems in terms of the
(54:39):
tacit expertise that's captured by it.
and the way that the work could be reinvented, like all that kind of seeps out of thecompany's purview.
You just touched on something that I think is critically uh important and verymisunderstood.
So I work with a group of CTOs, a chair group that meets every other month and the groupdoes nothing except talk about AI in terms of the impact on their organizations, what
(55:04):
they've done, where they are, where they're going.
And one of the recurrent themes, because we just had a meeting the other week, was thatthey've got a lot of folks in the organization uh that are doing the whole vibe coding
thing, right?
And they're
They're getting minimum viable product and in many instances it's quite good, but they'recoming uh to the IT department, right?
(55:24):
Because that is where everything gets funneled right now for right or wrong in mostorganizations and they're getting very frustrated then because IT has got three, six, nine
month type uh lead times where they need to secure, harden, right?
So there's a traditional path there that is bumping directly up against what a lot of thefrontline employees want to do.
(55:46):
And so I think there's a challenge there for organizations to figure out how do you havethe security and the stability that you've always needed that centralized IT provides
while accelerating and not chilling the enthusiasm of the folks uh that uh are working onthese different ideas uh so that they then don't go, in your words, outside the system.
(56:06):
Yeah, that is, that is, that is a, that is a huge tension.
And I think, you know, what I have seen is, sometimes larger companies have been furtherahead of this.
Actually Walmart actually is a really good example of this.
So they actually have even, uh, like earlier last year, they were implementing effectivelylike a, internal vibe coding infrastructure so that whoever it was, if they had a really
(56:31):
good idea, they could experiment with the latest stuff, but it would be compliant.
And not just compliant, but it would actually be able to pull from the data that they hadand the systems that they have.
Right.
So it's actually a better thing you can create because you create something that'sactually connected to your systems and has the context of your business rather than just
(56:52):
being developed on the outside.
And they are trying to figure out how do I bring this in and make it of any use to mybusiness.
And so I think that's creating that kind of sandbox for experimentation is important.
but it also has to stay outdated.
And that's the real challenge is that you could do it one time and then some new thingcomes out and no one's using Claude code anymore because it looks outdated compared to
(57:18):
some new thing.
And you need to make sure that you can stay at the cutting edge, at least for some of yourpeople too.
Now I think that's a really uh good set of points, Jeff.
All right, so our time always goes very quickly on these calls.
So as we look to wrap up here in the next six or seven minutes, uh I'm interested in twothings.
(57:42):
Number one, I'm interested in middle market organizations that let's say they've gotanywhere from 10 or 20 million to a billion and a half in revenue.
How does the playbook for them change versus some of the enterprise level organizations?
What can they actually do with fewer resources?
Where's a good starting point there?
(58:02):
Yeah, so I think for a mid-market organization, the questions are the same in terms of howdo we actually figure out what we're trying to achieve?
What's the right mix of types of solutions that we might look at?
(58:22):
But the way you might get there is different.
So I think...
Overall, the vision of understanding, what are we trying to achieve with AI is important.
And I think it's important to pass apart.
Well, we might want to have some efficiency gains, but also what are our hypotheses on howwe could use this thing and work with this thing in a different way in order to do things
(58:46):
that we couldn't have done before?
And what do we think that looks like for us as a business?
So beyond like, okay, we could be more efficient at these tasks.
What are the new types of outcomes, the new types of services that we couldn't haveachieved before?
And what do we want to aim towards?
So I think that that's an exercise that every company should do no matter their size.
And I think that that, yeah, that's the first step.
(59:10):
The second step then is thinking about, well, what's the portfolio of different effortsthat I might need.
And so that can look, that can work at different levels, kind of as I've described.
So firstly, well, how do we kind of deploy some tools?
and give people enough infrastructure and knowledge to be able to use that in theirday-to-day work.
The second thing is how do we actually identify some select areas where we want to reshapeand redesign how we work, redesign some workflows, not just automate them, actually like
(59:40):
rethink what's possible.
And then the third piece is what are new value streams that we might be able to create?
And not all organizations are going to be there yet, but
at least having a hypothesis as to, could we use this to create new types of services?
Could we use this to offer a different level of value to our customers?
(01:00:02):
Is there something that could reshape the type of business that we become based on thedata that we have and based on the problems that our customers or clients have?
The how you get there though is different.
So I think
The first thing to understand is that for a large enterprise, in the five upwards billionin revenue, they will be hounded by large consulting firms all the time.
(01:00:30):
And so they'll have no shortage of people pitching them solutions, and they can get freeadvice from them as well, which is good and bad.
It's always going to come with some kind of attachment to some kind of solution, usually,in mind.
Um, so I think for a mid-market firm, the big challenge is what's the extent to which youshould think about using off the shelf tools and systems and what should you actually
(01:00:59):
build and maintain yourselves?
And I think that's the key thing that really is important to understand and where are theareas you want to develop a focused team or effort to have that capability and that
expertise in house.
to understand what are the new tools that you should be keeping up on so that you preventpeople from using outside.
(01:01:21):
And that needs to be within the business, not just within IT.
And then how do you actually start to cultivate some kind of system for looking atidentifying efforts and funding them and making sure that you're making the most out of
them rather than just...
(01:01:41):
you know, rather than just kind of like letting people experiment.
And so for a lot of mid-market firms that probably won't look like creating a complexcenter of excellence and having teams in different functions and different business units.
But you still need to make sure that you have enough of that expertise in house so thatyou can effectively evaluate vendors and also evaluate consultancies and advisors to make
(01:02:05):
sure that your m
really maintaining the value for your company too.
Because I think the risk is that as a mid-market company, you end up just becomingsubsumed by the roadmaps and the decisions made by SaaS vendors rather than shaping your
own destiny.
(01:02:27):
And many of those vendors may not be here in another few years.
Yes.
All right, this has been thoroughly enjoyable for me, Jeff.
As we conclude here, let's bring this back to where we started and let's just uh hit onelast question here about collaboration.
So if I'm a leader, I've been listening to this episode, again, bringing us back to wherewe started here today.
(01:02:51):
If I want to start shifting my organization from using AI to actually collaborating withit, what are maybe the first two or three things that I should do uh next Monday when I
walk back in?
Yeah, so I think as a leader, would think about three things.
One is how do I shift my mental model?
How do I think about if there was an expert form of intelligence sat next to me every day,how would I do my day differently?
(01:03:19):
What does that mean for me personally?
And how do I look for that as a way to improve who I am as a leader, not just get stuffdone?
The second thing I would look at is how do I find an area of my work that I want to get myhands dirty and do something.
Right.
So a little bit about, you know, people might be vibe coding using Claude code.
(01:03:44):
I think it's really important.
The best way to actually understand how to work with AI is to work with AI to do somethingand build it.
So it's not about like taking a prompting class.
Well, finding a problem that you care about an area of your job that really sucks.
Mm-hmm.
would agree needs to change and start doing something to see how you can change that.
(01:04:04):
Right.
And it might be that the outcome you get to is not worth scaling or not worthimplementing, but you will actually learn what it's like to work with this different form
of intelligence, how you can get good results, how you can get bad results and what youneed to do differently to steer more towards the good results rather than the bad results.
(01:04:26):
And then the third thing is to think about
what that means for how you show up and lead, right?
And so this is about your day-to-day interactions with folks.
It's about sharing some of your learnings with humility and making it comfortable forpeople to say, well, yeah, I'm trying to figure this out too.
As well as actually helping to embody that questioning mindset and that critical thinkingmindset in your day-to-day conversations with your team and with your peers.
(01:04:56):
so that you're cultivating that kind critical intelligence in your day-to-day work.
Wonderful.
think that is a great place for us to conclude today as we continue on our journey with AIfrom just treating it like a tool that we search for things to collaborating then to some
(01:05:16):
hopefully semi-autonomous and autonomous features here in the near future.
uh Jeff, I really appreciate your time and uh everything that you brought to ourconversation today.
for our listeners that want to learn more about you, your work, maybe they want to buyyour book, maybe they want to connect with you.
What are the ways and the paths for them to find you online?
(01:05:37):
Yeah, sure.
So my company website is human-machines.com.
And then I'm also on LinkedIn, and I have a weekly newsletter.
And people can reach out to me by any of those two means.
Lovely.
Jeff Gibbons, thank you very much for your time today.
Wonderful, great, thank you, Jan.
And for our listeners, thank you for listening once again to AI for the C-suite.
(01:06:00):
If this episode was useful for you, subscribe wherever you get your podcasts, follow us onLinkedIn and check out AI for the C-suite.com.
So until next time, keep your algorithms running, your leadership evolving and your AI incheck.
Take care everybody.
Popular Podcasts
Betrayal Weekly
Betrayal Weekly is back for a new season. Every Thursday, Betrayal Weekly shares first-hand accounts of broken trust, shocking deceptions, and the trail of destruction they leave behind. Hosted by Andrea Gunning, this weekly ongoing series digs into real-life stories of betrayal and the aftermath. From stories of double lives to dark discoveries, these are cautionary tales and accounts of resilience against all odds. From the producers of the critically acclaimed Betrayal series, Betrayal Weekly drops new episodes every Thursday. If you would like to share your story, you can reach out to the Betrayal Team by emailing them at betrayalpod@gmail.com and follow us on Instagram at @betrayalpod and @glasspodcasts. Please join our Substack for additional exclusive content, curated book recommendations, and community discussions. Sign up FREE by clicking this link Beyond Betrayal Substack. Join our community dedicated to truth, resilience, and healing. Your voice matters! Be a part of our Betrayal journey on Substack.
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by Audiochuck Media Company.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.