September 30, 2025 • 42 mins
Join the Tool Use Discord: https://discord.gg/PnEGyXpjaX
Unlock the power of AI for your business, even if you're not a tech expert. In this episode, AI strategist Marni Wills reveals how non-technical founders can leverage AI as a new operating system to amplify their intelligence. Learn to build powerful AI workflows, create your own AI operating system, and personalize tools like ChatGPT, Claude, and Perplexity for massive business growth. Marni breaks down her exact process for using AI agents, developing custom instructions to avoid hallucinations, and building a "second brain" to automate tasks from email marketing to securing podcast interviews. This is your guide to making AI work for you, strategically, not technically.
Connect with Marni Wills:
LinkedIn: https://www.linkedin.com/in/marnie-wills-entrepreneur
Website: https://businesswithaistrategist.com/
Instagram: https://instagram.com/businesswithai_strategist/
Digital Store: https://stan.store/business_with_ai_strategist
Connect with us
https://x.com/ToolUsePodcast
https://x.com/MikeBirdTech
00:00:00 - Intro
00:02:19 - First Steps for Non-Technical Founders
00:08:13 - How to Avoid AI Hallucinations
00:11:04 - Building a Custom Prompt Library
00:27:17 - AI Agents vs. Automation (Lindy vs. Make)
00:32:48 - Creative AI Workflows You Haven't Thought Of
Subscribe for more insights on AI tools, productivity, and workflows.
Tool Use is a weekly conversation with the top AI experts, brought to you by ToolHive.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
Do not look at a is artificial intelligence.
Look at it as a completely new operating system and it's to
amplify your intelligence. You as the business leader.
What if we could just change ourjob and do whatever we want to
because of AII? Actually, don't think AI and the
use of generative AI is technical.
I think it's very much more strategical.
(00:20):
If you want to run a successful business, using AI is mandatory.
There's so many tools and systems available that
everything's become hyper competitive, but AI still feels
very technical, very intimidating to a lot of people.
So how does the non-technical founder leverage AI?
Well, in episode 59 of Tool Use,brought to you by Tool Hive,
we're joined by Marni Wills, an AI trainer and strategist.
(00:41):
She's going to teach us how AI can be leveraged by the
non-technical founder. We're going over workflows,
tools, and how a non techie can make an AI operating system.
So I hope you enjoyed this conversation with Marni Wills.
So when chat PT came out, I kindof just, you know, embraced it
massively. And Fast forward a couple and
like another 18 months and I pivoted my franchise to a
(01:03):
business partnership thanks to ChatGPT.
I completely streamlined, streamlined all of our processes
from e-mail marketing campaigns to sales to content creation to
creating a board of directors with a ChatGPT project to yeah,
for, for my business. And we had massive, massive
(01:23):
growth. So that was middle of 2024.
And then I sold and so exited myhealth and fitness business
because I created so many assetsthanks to ChatGPT and I was
definitely using quite a lot of Gemini back then actually.
And then yes, it exited that business.
And then I don't know how it happened, but I ended up
(01:44):
delivering AI workshops and my brand business with AI
Strategist started and my very popular newsletter is called
Master AI without Overwhelm. So here we are today.
Mike and I love to guest some podcasts and talk about AI from
a non techie background because there's absolutely nothing
techie about me. However, I do say to a lot of
(02:08):
people, I actually don't think AI and the use of generative AI
is technical. I think it's very much more
strategical. A lot of it's been built up in
such a way that people get intimidated by it.
So if there is a non-technical founder out there who's trying
to dabble but they don't know where to begin, what's your
general advice? Is it look at the existing
processes they have and try to find AI tool for that?
(02:28):
Or do you have almost like a playbook to someone to dip their
toes into the AI world? Yeah.
So my view on it is do not look at AI as artificial
intelligence. Look at it as an A new,
completely new operating system and it's to amplify your
intelligence. You is the business leader.
(02:48):
It is a complete new operating system.
It's, you know, just like when we first had print to when we
then had the Internet to when wethen had mobile phones like the
way we use our mobiles like and our smartphones.
It's changed like the way us humans communicate, talk, like
everything, it changes everything.
You know, we all, we all know, not all of us, but many of us
(03:11):
now have like a scrolling issue,like, you know, and we're going
to have that issue in, you know,a few years times.
We're all going to have an agentissue where we've all got these
agents working for us. So, you know, it is a complete
operating change. It is not just seen as a tool.
So that mindset shift of this isa new operating system and I
need to use a new operating system, an AI operating system
(03:34):
to so I need to personalize to amplify my intelligence.
I try to not focus too much on tools or automations to start
with. It is if you're going to use
chat sheet, if you're going to use Clawed or maybe you use a
model. My favorite is UNET AI, where I
can access 22 AI models in one platform that links to my
(03:57):
business brain. So I have created an AI
operating system that way that helps amplify my intelligence
and I really, really personalized it.
And I think that has probably put me in really good stead to
be able to develop all my workflows, start to use agentic
workflows, start to vibe code. Like I can't even believe, Mike,
(04:19):
I said this to my 8 year old daughter when I made her a few
apps just recently. She said, what have you done?
I'm like, I'm a vibe coder. She's a what a vibe coder.
So I don't even really know, butI'm a vibe coder.
I just like laughed about it. It was hilarious.
But yeah, that's what AI has done for me.
That's that's amazing. And I try to recommend people
similar, like you said, use it as a thought partner.
(04:40):
You can have conversations with AI to figure out what AI can do
for you and you can just bounce ideas off and just start
dabbling by by playing experimenting.
It's it's a reoccurring theme onthe show in the notes of
experimentation. If someone were to go to
chatgbt.com for the first time, do you have any advice on what
they should talk to it about, what they should give it?
(05:01):
Like there, there is some concern around privacy and
people don't necessarily want toovershare, but the same time to
really reap the benefit, The more information you provide the
better. So what are some advice around
experiments to run or just kind of getting a general
understanding of what these systems are capable?
Of yeah, I love that question. I'll tackle the security bet
because Claude is just also now Claude is probably my favorite
(05:22):
AI to go to, model to go to. I've just brought out their new
privacy and it's toggled naturally toggled on.
So if you go to your settings in, in almost all the AI models,
there is a bit settings and it says you can toggle on or off
whether you want the AI to be. If if the if you would like to
help the AI train on your information so your chats and
(05:44):
help train the AI, you can toggle that on or off.
I personally leave it on becauseI am a massive believer that AI
has got to where it's got because of all of our human
knowledge and, and everything weput out there to the world thus
far. And if we stop doing that, you
know, I just think it's going tobe lazy AI regurgitated content
(06:06):
going in and out. So I love to help the AI
learning, the AI learning journey for the next generation.
But I do understand that some people, you know, have sensitive
information or maybe some IP that they don't want to share
and those type of things. So first of all, 100% you, you
make your decision of what, whatyou prefer to do.
But the way I would view it is if you're very new to it, it's
(06:28):
all about the questions. I was at a digital women's
conference today and one of the advice was where should you
start? And someone said you should
learn prompt engineering. And I was like, no, you
shouldn't because the AI, all the AI models are getting so
much smarter now that they don'tneed the same prompting
frameworks that they used to usewhen they 1st come out
(06:50):
2022-2023. The, the art is asking
questions. The art of it is asking
questions and asking me a, what it can do to help you.
So the first thing that I would say is I would like to
personalize you Chatty. I'm going to call you Chatty or
Derek or whatever. Like, you know, I would like to
personalize you, ask me questions that are going to
(07:11):
allow you to get the best out ofme.
That's where I would start. And then I would just go back
and forth with it. I often use voice mode, but I
also on my laptop have and on myphone have a couple of different
programs. One's called voice type and the
other is called whisper, whisperflow.
(07:32):
So I just voice, obviously I'm abit of a talker rather than a
typer. So I just voice a lot.
And actually it's just way more natural conversation going back
and forth with the AI. So yeah, I would start with
questioning. It's personalized your AI with
questioning and just remember that we are the experts as in we
have the human domain expertise,the knowledge, the experience,
(07:54):
and it's the perspective, right?And the emotional intelligence
we've got. AI doesn't have that, but it's
like a PhD and everything. So it's super, super smart, but
without that emotional IQ, so and without any type of
perspective and enough context. So that's where we still have to
add that domain expertise. 1 justifiable concern that people
(08:16):
have when they're starting to play with AI is they hear all of
these conversations around the hallucinations and AI being
confidently wrong. If someone is in a situation of
yours where they're trying to expand their business or go into
domains that they're not too familiar with, do you have any
advice on how people should approach the risk of
hallucination and verify the output and make sure that
(08:37):
they're able to get accurate results?
Definitely. Well, first thing I would say is
if you can customize your A so AI.
So for example, like everybody knows chat.
So if you go to settings and yougo to customize and one of the
instructions I give, I always say respond to me in bullet
points. Always ask 2 clarifying
questions. Write in UK English.
(08:57):
No offense but ZS come out and all of them are writing and I
need s s. I say no dashes and I say lots
of commas and then that reduces the M dashes.
I say many emojis because we allknow emojis are, you know, AI
content. And then I say you need to cite
(09:17):
any statistics or information that you are giving me.
I want justification about whereit's come from.
And at the end you also tell me,would you like me to show you
where I got this information from?
I've added that in my custom instructions.
So all of that. So that would be the first
thing. And the second thing is if
you're if you're not quite sure,you can also use the deep
(09:38):
research option. So in the plus button often when
I'm trying to gather informationand I'm generally probably
asking quite surface level, if I'm a bit like, I'm not sure,
like that statistic sounds a bitcrazy or you know, where did
they make that up? Because remember, especially
chat 2BT5, the vibe is like justwants to please you, you know?
(09:59):
I don't know if you're liking the new 5.0 voice, but honestly,
I don't think I can handle it anymore.
It's just so like, it's just so I don't know, like best friend
next door trying to be super cool.
I can't deal with it and Claude's not much better.
(10:20):
When I, when I talk to Claude and it comes back, it goes
gotcha or great idea. Like like super enthusiastic.
I'm like, calm down. So I think, hey, I say elucinate
because they're trying to pleaseyou or they're trying to be
creative or you know, they're trying to fit the context you've
given it. So customizing, but also giving
it enough context and enough instructions.
(10:40):
And I often will then if I'm, ifI'm not sure, I'll re ask the
question and I'll press the deepresearch button because the deep
research button makes it go out,find the sources and cite the
sources so you you can see whereit's got its information from.
So they're the three things thatI do.
Perfect. Yeah, being able to to verify
and actually the sources is so important.
You mentioned the custom instructions, which I think is
great. It really allows people to
(11:01):
tailor to their own uses and their own preferences.
Do you do anything else in regards to having a library or a
store of, whether it's prompts or like primer prompts, like
partial prompts you can add to get consistent outputs?
And if So, what are some examples that you've put
together? Okay, so I have a few here.
So most of my projects as they're known as ChatGPT and
(11:26):
Claude actually use perplexity spaces, which is really, really
similar. So I actually think perplexity
is an amazing AI model for a couple of reasons.
One is the inbuilt model for perplexities called Sonar.
But you can actually choose the model that you can access and
it's got access to Claude Gemini.
(11:47):
It definitely doesn't have alarm.
I think it's got Mistral and open AI, so ChatGPT.
So the most amazing thing about perplexes, you can choose the
model. So that means that I have
spaces, so they're like projects.
And in those spaces, I have given them custom instructions
and a lot of context. So a lot of files, a lot of
(12:07):
connections to like to my GoogleDrive as an example.
And if I'm doing a project and it's maybe an e-mail marketing
campaign, I'm probably using clawed because it's quite good
at writing. So I change the model to clawed.
If I'm doing something creative,like maybe it's content ideation
or repurposing, I'm probably using ChatGPT.
If it's something educational, I'm probably using Gemini.
So each of my projects I actually have different AI
(12:30):
models connected to them, which is a little bit more of a step
than chat TBT projects, for example.
So, so number one is I have assistants that I set up via
spaces or they're also called projects, as I mentioned, and
they have a brain and they have custom instructions.
And then the one that I have, and I actually use it mainly on
clawed is I've got a, I've actually called it custom
(12:53):
instruction builder. So whenever I want to write new
instructions to create a projector a space or gems, if you're in
Gemini or I think agents are from Co pilots in Microsoft.
So they're all slightly different, but they're the same
thing. I go to Claude and I my Claude
custom instruction builder and Isay this is the outcome I want.
(13:14):
Write me my instructions now. I got that prompt from Claude
code. So I just went to code and I
said, this is what I need. I need a custom instruction
builder, you know, and, and I also have some instructions
because each model likes to produce different instructions
that likes. I didn't quite know exactly how
it is because again, I'm not notthe programmer, right?
(13:37):
But it's the way that it outputsor the OR the context that you
give it slightly different. So on my custom instruction
builder, I've got the details ofwhat what open AI like, so
ChatGPT, what Gemini likes, whatClaude likes.
And then I'll just go into my custom instruction bid or say I
want to build a project, I need custom instructions on ChatGPT
(13:59):
for XYZ and it comes up for me. So that's how I do it.
I love that we've been talking abit more about that on the show.
Just the the broad category of meta prompting where you use
prompts to get more prompts. Customer instructions, exact
same thing. Do you have any other workflows
where the output from your interaction with an AI is not
the final product, but it's justa step in the process?
(14:20):
Yeah. So I don't know if I was talking
to you about this, you know, prepodcast interview, but I get
these guest podcasting spots viamy GPT that I've created.
So my workflow is there's a few podcasting platforms as you
(14:40):
know, because that's how we connected and the match will
come up or my VA might put in some of the key contents from my
newsletter and see if she can find some links there.
We then go away and we find the information about the podcast
more on perplexity, especially like the last few episodes that
they've had, the number of downloads, the guests and we do
(15:02):
Perplexity does all the background research.
We then take that information, we put that information into a
GPT that I've created, which is my podcast picture.
And I've added the instructions to make sure that all of my, I
call it my thought leadership, right.
So all of the I guess my again, my domain expertise, my
knowledge, my experience and my perspective.
(15:23):
Everything comes from my newsletters that I write, my
podcasts that I guessed in, any articles I write and workshops I
deliver. All that information goes into
my podcast pictures. So it is my second brain as my
thought leader. And then the perplexity
information goes in is has instructions to align any of my
thought leadership concepts withthe podcast, either the topics
(15:47):
they've asked for or resonates with their audience.
And normally, normally it's both, to be fair, but a lot of
the time it's more about the audience because that's what
perplexity is great at, like finding you know, the, the
audience for the for the podcast, especially the episodes
as well. And then it writes me a nice
little pitch and then my VA sends it.
(16:08):
And then I get a invitation to come and join.
And that's where I have amazing conversations like this one.
So that is a part of my workflow.
Still has a human in the loop and it's still me because I'm
training the GPT about my thought leadership.
So yeah. I think it's amazing and the
pitch clearly worked. You had just mentioned human in
the loop, which is one of the most important principles in my
(16:29):
mind right now. Everyone kind of has different
boundaries and thresholds for when human in the loop should
actually occur. How much can they automate?
How much do they stay in charge of things?
What are your rules or heuristics around when it's OK
let an AI do its own thing versus when you should
definitely have a human in the loop?
I think human in the loop shouldbe in all of that right now,
mainly because I'm a big believer in the need for human
(16:53):
connection at some point. And again, still wanting to.
I think a generative AI is stillvery in its infancy and where
we're going to be in like 5-10 years time and we need to be
part of that journey. And if we just let agents or AI
automations go off and do their thing, I believe that we're not
(17:13):
going to get the best out of theGen.
AI use and out of ourselves as humans.
So for me, it's still human in the loop.
I do do some automations, but the automations I normally do
last within the business and I normally do at last because I
always start with the leader. I always start with the leader
in the businesses and the leadership team in the
businesses making sure they're personalizing their AI models.
(17:35):
Then we look at ways in which wecan create AI systems within the
business for whether that's likean OS model or whether that's,
you know, white labeling some platform.
So, you know, not everyone's gottheir own little chat TBT
account like, you know, or we know that Microsoft agents are
OK, but they're not the best in the world yet.
(17:56):
So you know that other people, other like most people in
business are using other tools and then everybody has all this
sporadic and there's nothing together.
And then you go write an AI compliance policy and you're
like not quite working here, right?
So I do leaders in the business personalize their AI models,
make sure they are using it to amplify their intelligence in
every way, shape and form that allow.
(18:18):
And then we look at the businessoperating system, but that
allows because they're the experts in their industry.
I'm the expert in strategy, but they're the expert in their
industry. So if I can, you know, give them
an environment where they realize the potential of AI and
exactly what it can do because they know their operations and
you know, the business mission goals and you know the customers
(18:42):
and the customer journey, they'll come up with the AI
workflows that suit them and thebusiness.
So I don't tell them what to do.We do it together because they
are their domain expertise in there.
And then once we've got those workflows in, we go, OK, what
are these a repetitive that we can then automate.
And if I'm honest, I probably haven't built any automations
(19:05):
from like I used to use make a lot, although Zapier has just
brought out some really good newagents in there, but I've moved
more to agents. So I use Lindy quite a lot and
Replic quite a lot now. So I've moved away from
automations from that bit. And I do honestly think we're
moving to agents anyway. And to come back to your
original question about human inthe loo, even with your agents,
(19:29):
you should have some, I believe you should have some type of
reminder, you know, in your calendar, every fortnight to to
one month, Max, you're checking in on those agents, you know,
because they've got to get smarter and we've got to get
smarter and we've all got to work together.
So we can't let them off the leash just yet.
Yeah. And everything moves and changes
(19:50):
so quickly that maybe there's a new integration or a new
alternative that you want to implement.
Absolutely. One thing that stood out to me
when you're going in and workingwith the people to help develop
their systems as a partnership. One area that people like to
grow in is not just learning about the different tools and
how to implement them, but actually what skills they should
develop. Because as soon as you leave
(20:10):
then they're stuck with these tools, but what about the next
progress? So when you're advising people,
what type of skills or mindsets do you recommend that they work
on developing so they can take these learnings and apply them
more deeply to their businesses?I believe the AI literacy is
around that first principle I talked about around questioning
the ability to question and to use AI to get the answer like
(20:33):
so. So question everything that you
need to be done and see if AI can collaborate with you to do
it. So I think that's like one O 1
of AI literacy. I think the second thing that I
would highlight, as you know, a real strength for people to
adopt and to make sure that everybody in the business is, is
aligned with, is the fact that you need a way of which your AI,
(21:02):
we can understand the business, the context of the business
enough so it never hallucinates because that is the bit that's
often missing is there's not enough.
I, I call it second brain because I would say, let's
create your second brain. Like there's just not enough
business second brain in, in much of the AI tools or AI
(21:25):
models that are being used. It's very generic stuff like
even to the point where, you know, I talked about my custom
instruction builder. I tell my clients to build that
and I give them a, an example ofmy custom instruction builder.
But I make them after we've given the AI the brain.
(21:45):
I say to them, now say to your AII need a custom instruction
builder. Here's an example, but I need it
unique to me. So all of my clients have
different custom instruction builders because they're all
unique to them. So I think that is the next AI
literacy skill of really making sure that it's super customized
(22:07):
and you're giving as much context to the AI as possible.
So it doesn't matter when you leave or if somebody leaves in
that like the the projects, the workflows, they're already there
and they have enough context. It's not so much about relying
on the human. And then I think like the last,
like, I don't, I guess you couldcall it AI literacy is just this
(22:29):
for me, it's a bit of a mindset piece, like you said around
mindset. For me, it's a mindset piece.
Like AI, as I've said already, is there to amplify us.
But actually, what if it's get, what if we can view it as like,
it creates an abundance of opportunities of which our job
descriptions massively change because what we can do with AI
(22:50):
is crazy. But also like, what if we could
just change our job and do whatever we want to because of
AI, you know? So we need that abundance
mindset of like the opportunities are amazing if you
let it and if you have that mindset to allow it to help you
guide on whatever is needed for you to do.
(23:10):
As you start incorporating more AI systems into your business,
you're going to want to give it real data with MCP, and that can
be a little bit scary, which is why I've been using Tool Hive.
Tool Hive makes it simple and secure to use MCP.
It includes a registry of trusted MCP servers and lets me
containerize any other server with a single command.
I can install it in the client seconds, and secret protection
(23:30):
and network isolation are built in.
You can try Tool Hive too. It's free and it's open source
and you can learn more at Tool Live dot dev.
Now back to the conversation with Marty, There's tons of
concerns around people have around job displacement, but
it's very easy to look at that versus job creation.
If you go back to the Internet, to mobile, all of them have had
a similar thing where some jobs are lost, some are created and
it's about staying flexible, agile learning and a lot of the
(23:52):
things that you mentioned. And I say this that a lot.
It's either 64 or 65% of Gen. Alpha which are born the
generation born 2010 to 2013, who if they don't go to higher
education are probably coming into the workforce now or
they'll be in the workforce by twenty. 3064% of their jobs do
(24:12):
not currently exist, which meansthere's massive job creation.
But not only that, the reasons that is 76% of them would love
to be entrepreneurs. So yeah, I think as business
leaders, I think we've got to, Ipersonally think we've got to
view that in two ways. One is, can we be a thought
(24:34):
leader and think about our customer journey and and what
our customers want from us in the future. 2027 is like a
really big AI number at the moment that 2027 report came
out. So like everyone's kind of
focusing on 2027. How are your customers
operating? What's their human operating
system when consuming your products or services or join you
on your customer journey? Work back from there and think
(24:57):
about the different ways that the jobs are going to change
because there's going to be massive transformation.
There'll be jobs created. And then think about if, if Jen
Elfa is coming in and they want to be an entrepreneur and you've
done all the hard work in your industry to get to where you
are, where are they picking up from?
And if they are picking up, you know, not copycatting, but
(25:18):
business is built on other businesses, it's just the way it
is. You solve a problem, right?
Good businesses solve problems for people and that's how it
goes. And you could have three of the
same business on the High Streetif you wanted to.
And it's a competitive race, so it is going to be there.
But what then is your USP? Is it you've been a business
longer? Is it you've transformed?
Is it the type of clientele or is it the brand itself?
(25:41):
And there's so much talk out there at the moment about, you
know, making sure, you know, companies and people have
personal brands. And that's why you see bigger
companies now where the the CEO's, especially all these tech
companies, the CEO's are guessing on so many podcasts,
aren't they? Because they're trying to build
a brand so that when Replit and Relay go against each other,
(26:01):
someone goes, oh, I'm going withReplit because the CEO did this
awesome podcast. Like that is where it's gone.
And I think we need to get aheadof the next kind of generation
coming in who are gonna, I thinkprobably flourish on our fruits
left. What's the saying our labour of
fruits or fruits of labour, I can't remember, you know, but
what we're doing now is businessleaders and, you know, drudging
(26:23):
through this AI transition and then we'll get to the end.
And then we're like all these kids taking out, taking our
businesses. But you know, good on them, I
say. Yeah, I'm, I'm so thrilled at
the idea of a one to five personcompany being able to be
leveraged so much that they compete with the big people.
It's really going to change the landscaper things.
It's so exciting. One thing that I, before I
(26:46):
forget about it, I want to bringup that you mentioned is, is the
whole concept of the second brain.
I've put a lot of effort into building my personal one where I
use an Obsidian vault. I have an Obsidian plug in
generator that automates things.It's very technical and I know
that's not for everybody. And when you mentioned some of
the tools like make and Lindy, I'd love if you could kind of
explain a bit about what their strengths are, when people
should look into them, what typeof benefit they can expect to
(27:07):
drive because everyone will havedifferent needs.
But having a journal idea of where they should even start to
approach for building the systemto help them out, whether it's
their personal or professional life would be helpful.
So could you just explain like what is Lindy?
What is make? When would you choose one over
the other? Yeah.
So I'll come back to Lindy and make only because I think
there's a few other operating system that people can look into
(27:27):
first and probably should look into UNET as one which I
mentioned and the similar one tothatisyou.com.
So they're all and there's quitea few.
Basically the concept is that it's one platform, so you like
open AI, for example. So if you go to your chat, but
on that in your platform you'll have access to 22 different AI
(27:51):
models that you can choose. And then you have access to the
files and the context. So there'll be connectors which
you can do in your chat account as well, but you'll have
connectors that can connect to different files and different, I
guess calendars and you know, different contexts that you want
to give it. And then you can create the AI
(28:12):
assistant or the agent. So slight different.
The AI assistant is a general large language model of which
you will chat back and forth to it.
Whereas an agent, it has the agentic mode on and you will
give it a task and instead of you giving it the next step, it
will go off. It will decide and do the
thinking about how it's going tocomplete that task.
(28:33):
It will go off and it will do the research and it will have
another agent that writes the copy.
Then it might have another agentthat writes the code for a
landing page. So that's the agentic mode where
it does it all. Most of the large language
language models now have them built in chat.
GBT's agent mode perplexities called Labs, Claude is called
Artifacts and Gemini is called Gem Gem apps.
(28:56):
So they all have agent mode in them now.
But these platforms also have itwhere they'll have some inbuilt
templates. So it might be like, I don't
know, a content repurposing agent of which because you've
got your business brain already in your platform of which you
could have so many users sign in.
So you're all signing in to the one kind of AI model that's for
(29:18):
your business and therefore you could have different agents set
up for different departments fordifferent people within the
department, but it all connects into, you know, a live drive
type thing that will always get updated.
So I would probably look into those type of AI models first
for businesses, especially businesses you know, of more
(29:39):
than I say, you know, probably 5people.
So yeah, I'll probably look intothat first.
And then when you're thinking about agents, so Lindy is
probably my favorite because I'ma talker, as I've said many
times, and I can just go, right,this is what I think I need my
agent to do. Just as a really simple example,
(30:03):
You know, I need my agent to look at my emails that come in
daily and create three folders for me.
One is AI information, one is urgent action.
The other, you know, probably follow-ups, you know, people
that I need to follow up with that I haven't followed up with
or that they, you know, followedup with me type thing.
And then and I get, and then I say to the agent, well, I also
(30:26):
want drafts. I want draft replies.
And with the AI information, I then say, here's my 4 areas that
I like to talk about, like my 4 pillars, I like to talk about,
take the information from those emails and put them in those
into those pillars. And then from there, you know,
and I like to to check it at 8:00 AM in the morning and 5:00
PM in the evening. And then I'll get a notification
(30:48):
to say, here you go. And I'll have my drafts there
ready for me to check and my content ready for me to go
through. And another little workflow
which I haven't spoken about is I then take that content that
Lindy has done in my emails and I put it into Notebook LM, my
absolute favorite tool because on Notebook LM you can upload
(31:09):
sources. I think it's like 300 sources.
It could be YouTube videos, websites, pasted texts.
It can even search the net and then you can talk to those
sources. It can create a podcast for you.
It can create interactive like how to videos, it can create
study guides, it can create reports you could talk to the
(31:31):
AI. So I often add it as my
knowledge base and then I'll sayI'm writing my newsletter.
Here's my 4 things that I like to talk about in the newsletter.
And I also have my previous newsletters update.
Is there anything in here I haven't spoken about or have I
spoken about at all? And there's been an update and
it will, you know, look through all my information and be able
to give that back to me. So yeah, notebook LM is is
(31:51):
absolutely amazing. But again, that's all because I
basically, I asked Lindy just said to Lindy, exactly like
pretty much how I've spoken to you just now, this is what I
need. And Lindy goes and builds it.
Whereas different to say make orNAN where you've got to, it's an
automation of which you've got to put a trigger.
Then you have to connect it to alarge language model.
Then you have to say, this is what I want to happen.
(32:12):
If this happens. Like the automations, they are
getting better, don't get me wrong.
And I'm probably can't speak anymore because I don't actually
use them anymore. I probably haven't used them now
for about 3-4 months because again, I just like talking to an
AI agent rather than doing the doing.
That's awesome. That's that's one of the best
referrals for Lindy I've heard. I'm also a huge fan of Notebook
(32:35):
LM. Where I live, it's very rural,
everything's an hour drive away.I can't be reading papers and
trying to keep up with things. So I just throw a few resources
into Notebook LM. I haven't used the interactive
mode, but you've got me excited about that one.
No, this is great. This is super exciting.
Are there any other unexpected workflows that have come up with
your business or your consultingwhere even if it wasn't the
(32:57):
initial ID or thought gradually just evolved out of your
interactions where you thought Icould actually make this
automation or that's an idea fora tool that I never even
considered, but it works in my workflow?
Have been any unexpected surprises through your discovery
process? Yeah, I think 2 come to mind.
So the first one is a tool I came across.
(33:19):
I always pick myself up when I say tool because here I am
preaching you need an AI operating system.
And then I go, I've got this tool and I've got this tool.
But I generally have an AI operating system, as you know,
I've spoken about, but there aredefinitely different tools.
So interestingly, one of my clients does a lot of driving to
(33:39):
and from well within the countryhere in the UK.
He believe it or not, he's actually in cybersecurity.
So and he was really struggling to keep up with a lot of work.
So we found a tool and I use it now.
It's amazing. It's called voice pen.
So voice pen and basically we have set up his voice pan.
(33:59):
So he will just press record andhe'll start talking and
sometimes he'll talk out an e-mail or sometimes he'll, you
know, just talk about a client he's just gone and visit and you
know, brain dump or he'll talk out an article he wants to write
on LinkedIn and he just talks. And then there's a transcribe
button down the bottom and you can transcribe that into
(34:20):
anything you like. So we've set up.
So for his LinkedIn, we've customized it to write like he
writes on LinkedIn. So he pressed transcribe and
then the way that it transcribesit to LinkedIn, he can just copy
that and put that straight on LinkedIn and actually it now
connects straight to LinkedIn. So it will go straight and post
it for him. Emails, he likes to reply in a
certain way. So we've customized it so it
(34:41):
says e-mail and he just, you know, basically sends it that
way. And we've done a WhatsApp one as
well because he, you know, say WhatsApp, you don't want to have
the transcription all there likethe WhatsApp couple of lines,
bit of a space and lots of bullet points.
So that's how he does his Whatsapps.
Yeah. So we've we've customized his
voice pan so that he can literally work while driving.
(35:04):
And it yeah, I think it's been massively game changing for him.
And I really do hate that word because it's a very AI word,
game changing, revolutionized. But honestly, I think he would
and all of his team would 100% agree.
Like they're not waiting on halfthe stuff for him now because
he's doing it as he's driving. So that's one workflow.
And then surprisingly, and this comes back to vibe coding I was
(35:26):
working with, it's called a CIC over here.
So it's a community interest club.
So they have to get funding, they have to apply for grants
and funding, but they're not quite a charity.
So they can't do like lots of fundraising like other charities
can, but they can, they can apply for different grants from
government or from private companies.
And it's a bit of a numbers game, like there's a whole lot
(35:47):
of grants out there, but you've got to, you know, apply for them
and it takes them so long and you know, they're a community
interest club. So they don't, you know, they're
not paying themselves a lot and they're very time poor and they
can't afford a VA and those typeof things.
So what I did is on the most amazing Lovable with my vibe
coding skills is I did a whole lot of deep research and I found
(36:10):
all of the grants that and the funds that were eligible for
this CIC within the year. I then uploaded that into
Superbase which connected to Lovable and I said I need a
almost like a CRM that is able to divide this up into monthly
applications for this company and gave it all the information.
(36:33):
And then it had the database of the of the of the research that
I did. And then every time this company
director goes into this lovable platform that created every 30
days it comes up and it says, sothe last 30 days they had 238 to
apply for. So it comes up, it's got all the
information about the application.
(36:56):
It just gives you a link to apply straight away.
We then created a GPT, which we didn't use it as a traditional
GPT. It was a project within the AI
only because the project updatesin memory and every time they do
an application, the update memory.
And when they have a successful application, they're able to
then upload that as well. So therefore, it's constantly
updating its memory. Whereas the GPT, the memory
(37:17):
doesn't update unless you go in and update it.
So yeah, they then put that information that come off that
lovable platform, put that straight into the GPT, It fills
out the application, they go back, they fill it in.
Done. It was taking them roughly about
3 hours per application. It now takes them anywhere
between 20 and 30 minutes per application, and they're able to
do 30 a month. And it's just a numbers game.
(37:39):
So that's a work, though, that Inever, like thought would ever
happen. It was just like working with
the guy to help him, you know, with his personalized AI
operating system, like I said. And then he said, this is my
biggest bottleneck. And I'm like, we can solve that,
surely, you know. Yeah.
And those are both phenomenal use cases where we're taking
something that would happen anyway over a certain time and
(38:00):
compressing it, giving people their time back, allowing to
have a bigger impact. There's been so many times I've
been driving, I'm just like, I'mstuck in the car for three
hours. I can't do anything.
I'm offline and if you can just enable a little bit more
productivity, it gives you that evening back that you'd have to
like work late. You're clearly on top of a lot
of this stuff. You you've mentioned a lot of
great tools. What are your go to sources for
(38:20):
learning about these things, discovering them?
Do you have any pipelines that help exploit that process?
There's how do you how do you learn about these AI tools?
So I get a lot of newsletters. There's an AI tool for that.
I have the newsletter come through.
I think there's an AI for business.
There's quite a few. So and then my Lindy kind of
splits them up for me, which is good, but I'm a bit I'm very
addicted to YouTube. Another job of mine is I happen
(38:43):
to be mum's taxi. So I have to drive my children
to and from school. Lots of activities.
I do a lot of driving myself. So I do get to do a lot of
learning. I follow Matt Wolf future tools,
AIIO and is his ones hubspots marketing gets the grain is
absolutely phenomenal. Rob, the AI guy, I hope, I hope
(39:04):
he doesn't listen. He's slightly annoying, but
really good content. And then there's a guy, I think
he's based in Lisbon. He's cool.
I think he might be called the AI guy.
He's also really, really good bit slow.
So I listened to him on 1.5 speed.
But all those podcasts that I doabsolutely love.
And again, I, I really utilize notebook LM to help like kind of
(39:29):
help with my learning by either condensing it, especially I'm
trying to really stay up to datewith like SEO and AI search.
A lot of my business, I work with a lot of marketing
agencies, so they really need tokeep up with that.
And you know, there's so many different opinions and thoughts
out there that actually instead of following one person, I try
and get as many as I can. I put it into a notebook and I'm
(39:49):
like, right, let's compare them.Let's think about this from a
perspective of, you know, this type of business, you know,
because I think everybody, you know, without sounding rude or
snobby by any means, but I do think there's a big rise of AI
experts in industries. And that's awesome because, you
know, a year ago I was trying toconvince people to adopt AI.
(40:10):
Now I'm like, every time I open,I'm like, oh, this person's
become an AI expert all of a sudden.
Amazing. But it's very industry specific.
And because I work across a lot of industries, I just have to be
really careful to make sure thatI'm gaining a whole lot of
resources and not relying too much on kind of 1 prospective.
But if you haven't heard already, HubSpot are really
(40:30):
pushing the recent kind of coined term called loop
marketing, which discusses how to market in the age of AI and
talks a lot a little bit about AI search as well, which has
been really, really helpful because there's a lot of
agencies going through massive transformations and really
struggling to get their head around it.
Yeah, absolutely. It's, it's a firehose of
(40:51):
information. But if you can find a way to
distill it and and get the gems out, it's awesome.
Marty, this was a phenomenal chat.
I learned a bunch. This is really enjoyable to get
a perspective and your expertise.
Before we let you go, is there anything you want the audience
to know? I'd love to chat on LinkedIn if
anyone wants to find me on LinkedIn.
I think you said you'll put the details in the show notes.
And I love working with people. I call it a 90 day Sprint, your
(41:12):
AI, your personal AI accelerator.
I think 90 days is a really awesome day.
We awesome timeline to get a lotdone, get a lot of processes
done. So if you think you're behind,
you're not. And a 90 day Sprint with me and
you will be flying. So yeah, I welcome anybody to
come join me on that. Thank you for listening to this
conversation with Marnie Wills. I had a great time talking to
(41:32):
Marnie, full of energy, a lot ofgreat insights and ideas, and I
really hope it inspired you to start exploring and playing with
these AI tools. You don't need to know how to
code to leverage AI to help you run your business.
And if you're thinking about starting a startup or maybe a
little side hustle, there are many options available so that
you can leverage these benefits and get to your end goal faster.
I also hope you consider checking out making a second
(41:55):
brain, being able to start leveraging AI in your personal
life or have different automations happen.
It all begins by just asking questions to AI.
What's even possible? What tools exist?
What can you do? I want to give a quick shout out
to Tool Hive for supporting thischannel so I can have these type
of conversations and I'll see you next week.
Do not look at a is artificial intelligence.
Look at it as a completely new operating system and it's to
amplify your intelligence. You as the business leader.
What if we could just change ourjob and do whatever we want to
because of AII? Actually, don't think AI and the
use of generative AI is technical.
I think it's very much more strategical.
(00:20):
If you want to run a successful business, using AI is mandatory.
There's so many tools and systems available that
everything's become hyper competitive, but AI still feels
very technical, very intimidating to a lot of people.
So how does the non-technical founder leverage AI?
Well, in episode 59 of Tool Use,brought to you by Tool Hive,
we're joined by Marni Wills, an AI trainer and strategist.
(00:41):
She's going to teach us how AI can be leveraged by the
non-technical founder. We're going over workflows,
tools, and how a non techie can make an AI operating system.
So I hope you enjoyed this conversation with Marni Wills.
So when chat PT came out, I kindof just, you know, embraced it
massively. And Fast forward a couple and
like another 18 months and I pivoted my franchise to a
(01:03):
business partnership thanks to ChatGPT.
I completely streamlined, streamlined all of our processes
from e-mail marketing campaigns to sales to content creation to
creating a board of directors with a ChatGPT project to yeah,
for, for my business. And we had massive, massive
(01:23):
growth. So that was middle of 2024.
And then I sold and so exited myhealth and fitness business
because I created so many assetsthanks to ChatGPT and I was
definitely using quite a lot of Gemini back then actually.
And then yes, it exited that business.
And then I don't know how it happened, but I ended up
(01:44):
delivering AI workshops and my brand business with AI
Strategist started and my very popular newsletter is called
Master AI without Overwhelm. So here we are today.
Mike and I love to guest some podcasts and talk about AI from
a non techie background because there's absolutely nothing
techie about me. However, I do say to a lot of
(02:08):
people, I actually don't think AI and the use of generative AI
is technical. I think it's very much more
strategical. A lot of it's been built up in
such a way that people get intimidated by it.
So if there is a non-technical founder out there who's trying
to dabble but they don't know where to begin, what's your
general advice? Is it look at the existing
processes they have and try to find AI tool for that?
(02:28):
Or do you have almost like a playbook to someone to dip their
toes into the AI world? Yeah.
So my view on it is do not look at AI as artificial
intelligence. Look at it as an A new,
completely new operating system and it's to amplify your
intelligence. You is the business leader.
(02:48):
It is a complete new operating system.
It's, you know, just like when we first had print to when we
then had the Internet to when wethen had mobile phones like the
way we use our mobiles like and our smartphones.
It's changed like the way us humans communicate, talk, like
everything, it changes everything.
You know, we all, we all know, not all of us, but many of us
(03:11):
now have like a scrolling issue,like, you know, and we're going
to have that issue in, you know,a few years times.
We're all going to have an agentissue where we've all got these
agents working for us. So, you know, it is a complete
operating change. It is not just seen as a tool.
So that mindset shift of this isa new operating system and I
need to use a new operating system, an AI operating system
(03:34):
to so I need to personalize to amplify my intelligence.
I try to not focus too much on tools or automations to start
with. It is if you're going to use
chat sheet, if you're going to use Clawed or maybe you use a
model. My favorite is UNET AI, where I
can access 22 AI models in one platform that links to my
(03:57):
business brain. So I have created an AI
operating system that way that helps amplify my intelligence
and I really, really personalized it.
And I think that has probably put me in really good stead to
be able to develop all my workflows, start to use agentic
workflows, start to vibe code. Like I can't even believe, Mike,
(04:19):
I said this to my 8 year old daughter when I made her a few
apps just recently. She said, what have you done?
I'm like, I'm a vibe coder. She's a what a vibe coder.
So I don't even really know, butI'm a vibe coder.
I just like laughed about it. It was hilarious.
But yeah, that's what AI has done for me.
That's that's amazing. And I try to recommend people
similar, like you said, use it as a thought partner.
(04:40):
You can have conversations with AI to figure out what AI can do
for you and you can just bounce ideas off and just start
dabbling by by playing experimenting.
It's it's a reoccurring theme onthe show in the notes of
experimentation. If someone were to go to
chatgbt.com for the first time, do you have any advice on what
they should talk to it about, what they should give it?
(05:01):
Like there, there is some concern around privacy and
people don't necessarily want toovershare, but the same time to
really reap the benefit, The more information you provide the
better. So what are some advice around
experiments to run or just kind of getting a general
understanding of what these systems are capable?
Of yeah, I love that question. I'll tackle the security bet
because Claude is just also now Claude is probably my favorite
(05:22):
AI to go to, model to go to. I've just brought out their new
privacy and it's toggled naturally toggled on.
So if you go to your settings in, in almost all the AI models,
there is a bit settings and it says you can toggle on or off
whether you want the AI to be. If if the if you would like to
help the AI train on your information so your chats and
(05:44):
help train the AI, you can toggle that on or off.
I personally leave it on becauseI am a massive believer that AI
has got to where it's got because of all of our human
knowledge and, and everything weput out there to the world thus
far. And if we stop doing that, you
know, I just think it's going tobe lazy AI regurgitated content
(06:06):
going in and out. So I love to help the AI
learning, the AI learning journey for the next generation.
But I do understand that some people, you know, have sensitive
information or maybe some IP that they don't want to share
and those type of things. So first of all, 100% you, you
make your decision of what, whatyou prefer to do.
But the way I would view it is if you're very new to it, it's
(06:28):
all about the questions. I was at a digital women's
conference today and one of the advice was where should you
start? And someone said you should
learn prompt engineering. And I was like, no, you
shouldn't because the AI, all the AI models are getting so
much smarter now that they don'tneed the same prompting
frameworks that they used to usewhen they 1st come out
(06:50):
2022-2023. The, the art is asking
questions. The art of it is asking
questions and asking me a, what it can do to help you.
So the first thing that I would say is I would like to
personalize you Chatty. I'm going to call you Chatty or
Derek or whatever. Like, you know, I would like to
personalize you, ask me questions that are going to
(07:11):
allow you to get the best out ofme.
That's where I would start. And then I would just go back
and forth with it. I often use voice mode, but I
also on my laptop have and on myphone have a couple of different
programs. One's called voice type and the
other is called whisper, whisperflow.
(07:32):
So I just voice, obviously I'm abit of a talker rather than a
typer. So I just voice a lot.
And actually it's just way more natural conversation going back
and forth with the AI. So yeah, I would start with
questioning. It's personalized your AI with
questioning and just remember that we are the experts as in we
have the human domain expertise,the knowledge, the experience,
(07:54):
and it's the perspective, right?And the emotional intelligence
we've got. AI doesn't have that, but it's
like a PhD and everything. So it's super, super smart, but
without that emotional IQ, so and without any type of
perspective and enough context. So that's where we still have to
add that domain expertise. 1 justifiable concern that people
(08:16):
have when they're starting to play with AI is they hear all of
these conversations around the hallucinations and AI being
confidently wrong. If someone is in a situation of
yours where they're trying to expand their business or go into
domains that they're not too familiar with, do you have any
advice on how people should approach the risk of
hallucination and verify the output and make sure that
(08:37):
they're able to get accurate results?
Definitely. Well, first thing I would say is
if you can customize your A so AI.
So for example, like everybody knows chat.
So if you go to settings and yougo to customize and one of the
instructions I give, I always say respond to me in bullet
points. Always ask 2 clarifying
questions. Write in UK English.
(08:57):
No offense but ZS come out and all of them are writing and I
need s s. I say no dashes and I say lots
of commas and then that reduces the M dashes.
I say many emojis because we allknow emojis are, you know, AI
content. And then I say you need to cite
(09:17):
any statistics or information that you are giving me.
I want justification about whereit's come from.
And at the end you also tell me,would you like me to show you
where I got this information from?
I've added that in my custom instructions.
So all of that. So that would be the first
thing. And the second thing is if
you're if you're not quite sure,you can also use the deep
(09:38):
research option. So in the plus button often when
I'm trying to gather informationand I'm generally probably
asking quite surface level, if I'm a bit like, I'm not sure,
like that statistic sounds a bitcrazy or you know, where did
they make that up? Because remember, especially
chat 2BT5, the vibe is like justwants to please you, you know?
(09:59):
I don't know if you're liking the new 5.0 voice, but honestly,
I don't think I can handle it anymore.
It's just so like, it's just so I don't know, like best friend
next door trying to be super cool.
I can't deal with it and Claude's not much better.
(10:20):
When I, when I talk to Claude and it comes back, it goes
gotcha or great idea. Like like super enthusiastic.
I'm like, calm down. So I think, hey, I say elucinate
because they're trying to pleaseyou or they're trying to be
creative or you know, they're trying to fit the context you've
given it. So customizing, but also giving
it enough context and enough instructions.
(10:40):
And I often will then if I'm, ifI'm not sure, I'll re ask the
question and I'll press the deepresearch button because the deep
research button makes it go out,find the sources and cite the
sources so you you can see whereit's got its information from.
So they're the three things thatI do.
Perfect. Yeah, being able to to verify
and actually the sources is so important.
You mentioned the custom instructions, which I think is
great. It really allows people to
(11:01):
tailor to their own uses and their own preferences.
Do you do anything else in regards to having a library or a
store of, whether it's prompts or like primer prompts, like
partial prompts you can add to get consistent outputs?
And if So, what are some examples that you've put
together? Okay, so I have a few here.
So most of my projects as they're known as ChatGPT and
(11:26):
Claude actually use perplexity spaces, which is really, really
similar. So I actually think perplexity
is an amazing AI model for a couple of reasons.
One is the inbuilt model for perplexities called Sonar.
But you can actually choose the model that you can access and
it's got access to Claude Gemini.
(11:47):
It definitely doesn't have alarm.
I think it's got Mistral and open AI, so ChatGPT.
So the most amazing thing about perplexes, you can choose the
model. So that means that I have
spaces, so they're like projects.
And in those spaces, I have given them custom instructions
and a lot of context. So a lot of files, a lot of
(12:07):
connections to like to my GoogleDrive as an example.
And if I'm doing a project and it's maybe an e-mail marketing
campaign, I'm probably using clawed because it's quite good
at writing. So I change the model to clawed.
If I'm doing something creative,like maybe it's content ideation
or repurposing, I'm probably using ChatGPT.
If it's something educational, I'm probably using Gemini.
So each of my projects I actually have different AI
(12:30):
models connected to them, which is a little bit more of a step
than chat TBT projects, for example.
So, so number one is I have assistants that I set up via
spaces or they're also called projects, as I mentioned, and
they have a brain and they have custom instructions.
And then the one that I have, and I actually use it mainly on
clawed is I've got a, I've actually called it custom
(12:53):
instruction builder. So whenever I want to write new
instructions to create a projector a space or gems, if you're in
Gemini or I think agents are from Co pilots in Microsoft.
So they're all slightly different, but they're the same
thing. I go to Claude and I my Claude
custom instruction builder and Isay this is the outcome I want.
(13:14):
Write me my instructions now. I got that prompt from Claude
code. So I just went to code and I
said, this is what I need. I need a custom instruction
builder, you know, and, and I also have some instructions
because each model likes to produce different instructions
that likes. I didn't quite know exactly how
it is because again, I'm not notthe programmer, right?
(13:37):
But it's the way that it outputsor the OR the context that you
give it slightly different. So on my custom instruction
builder, I've got the details ofwhat what open AI like, so
ChatGPT, what Gemini likes, whatClaude likes.
And then I'll just go into my custom instruction bid or say I
want to build a project, I need custom instructions on ChatGPT
(13:59):
for XYZ and it comes up for me. So that's how I do it.
I love that we've been talking abit more about that on the show.
Just the the broad category of meta prompting where you use
prompts to get more prompts. Customer instructions, exact
same thing. Do you have any other workflows
where the output from your interaction with an AI is not
the final product, but it's justa step in the process?
(14:20):
Yeah. So I don't know if I was talking
to you about this, you know, prepodcast interview, but I get
these guest podcasting spots viamy GPT that I've created.
So my workflow is there's a few podcasting platforms as you
(14:40):
know, because that's how we connected and the match will
come up or my VA might put in some of the key contents from my
newsletter and see if she can find some links there.
We then go away and we find the information about the podcast
more on perplexity, especially like the last few episodes that
they've had, the number of downloads, the guests and we do
(15:02):
Perplexity does all the background research.
We then take that information, we put that information into a
GPT that I've created, which is my podcast picture.
And I've added the instructions to make sure that all of my, I
call it my thought leadership, right.
So all of the I guess my again, my domain expertise, my
knowledge, my experience and my perspective.
(15:23):
Everything comes from my newsletters that I write, my
podcasts that I guessed in, any articles I write and workshops I
deliver. All that information goes into
my podcast pictures. So it is my second brain as my
thought leader. And then the perplexity
information goes in is has instructions to align any of my
thought leadership concepts withthe podcast, either the topics
(15:47):
they've asked for or resonates with their audience.
And normally, normally it's both, to be fair, but a lot of
the time it's more about the audience because that's what
perplexity is great at, like finding you know, the, the
audience for the for the podcast, especially the episodes
as well. And then it writes me a nice
little pitch and then my VA sends it.
(16:08):
And then I get a invitation to come and join.
And that's where I have amazing conversations like this one.
So that is a part of my workflow.
Still has a human in the loop and it's still me because I'm
training the GPT about my thought leadership.
So yeah. I think it's amazing and the
pitch clearly worked. You had just mentioned human in
the loop, which is one of the most important principles in my
(16:29):
mind right now. Everyone kind of has different
boundaries and thresholds for when human in the loop should
actually occur. How much can they automate?
How much do they stay in charge of things?
What are your rules or heuristics around when it's OK
let an AI do its own thing versus when you should
definitely have a human in the loop?
I think human in the loop shouldbe in all of that right now,
mainly because I'm a big believer in the need for human
(16:53):
connection at some point. And again, still wanting to.
I think a generative AI is stillvery in its infancy and where
we're going to be in like 5-10 years time and we need to be
part of that journey. And if we just let agents or AI
automations go off and do their thing, I believe that we're not
(17:13):
going to get the best out of theGen.
AI use and out of ourselves as humans.
So for me, it's still human in the loop.
I do do some automations, but the automations I normally do
last within the business and I normally do at last because I
always start with the leader. I always start with the leader
in the businesses and the leadership team in the
businesses making sure they're personalizing their AI models.
(17:35):
Then we look at ways in which wecan create AI systems within the
business for whether that's likean OS model or whether that's,
you know, white labeling some platform.
So, you know, not everyone's gottheir own little chat TBT
account like, you know, or we know that Microsoft agents are
OK, but they're not the best in the world yet.
(17:56):
So you know that other people, other like most people in
business are using other tools and then everybody has all this
sporadic and there's nothing together.
And then you go write an AI compliance policy and you're
like not quite working here, right?
So I do leaders in the business personalize their AI models,
make sure they are using it to amplify their intelligence in
every way, shape and form that allow.
(18:18):
And then we look at the businessoperating system, but that
allows because they're the experts in their industry.
I'm the expert in strategy, but they're the expert in their
industry. So if I can, you know, give them
an environment where they realize the potential of AI and
exactly what it can do because they know their operations and
you know, the business mission goals and you know the customers
(18:42):
and the customer journey, they'll come up with the AI
workflows that suit them and thebusiness.
So I don't tell them what to do.We do it together because they
are their domain expertise in there.
And then once we've got those workflows in, we go, OK, what
are these a repetitive that we can then automate.
And if I'm honest, I probably haven't built any automations
(19:05):
from like I used to use make a lot, although Zapier has just
brought out some really good newagents in there, but I've moved
more to agents. So I use Lindy quite a lot and
Replic quite a lot now. So I've moved away from
automations from that bit. And I do honestly think we're
moving to agents anyway. And to come back to your
original question about human inthe loo, even with your agents,
(19:29):
you should have some, I believe you should have some type of
reminder, you know, in your calendar, every fortnight to to
one month, Max, you're checking in on those agents, you know,
because they've got to get smarter and we've got to get
smarter and we've all got to work together.
So we can't let them off the leash just yet.
Yeah. And everything moves and changes
(19:50):
so quickly that maybe there's a new integration or a new
alternative that you want to implement.
Absolutely. One thing that stood out to me
when you're going in and workingwith the people to help develop
their systems as a partnership. One area that people like to
grow in is not just learning about the different tools and
how to implement them, but actually what skills they should
develop. Because as soon as you leave
(20:10):
then they're stuck with these tools, but what about the next
progress? So when you're advising people,
what type of skills or mindsets do you recommend that they work
on developing so they can take these learnings and apply them
more deeply to their businesses?I believe the AI literacy is
around that first principle I talked about around questioning
the ability to question and to use AI to get the answer like
(20:33):
so. So question everything that you
need to be done and see if AI can collaborate with you to do
it. So I think that's like one O 1
of AI literacy. I think the second thing that I
would highlight, as you know, a real strength for people to
adopt and to make sure that everybody in the business is, is
aligned with, is the fact that you need a way of which your AI,
(21:02):
we can understand the business, the context of the business
enough so it never hallucinates because that is the bit that's
often missing is there's not enough.
I, I call it second brain because I would say, let's
create your second brain. Like there's just not enough
business second brain in, in much of the AI tools or AI
(21:25):
models that are being used. It's very generic stuff like
even to the point where, you know, I talked about my custom
instruction builder. I tell my clients to build that
and I give them a, an example ofmy custom instruction builder.
But I make them after we've given the AI the brain.
(21:45):
I say to them, now say to your AII need a custom instruction
builder. Here's an example, but I need it
unique to me. So all of my clients have
different custom instruction builders because they're all
unique to them. So I think that is the next AI
literacy skill of really making sure that it's super customized
(22:07):
and you're giving as much context to the AI as possible.
So it doesn't matter when you leave or if somebody leaves in
that like the the projects, the workflows, they're already there
and they have enough context. It's not so much about relying
on the human. And then I think like the last,
like, I don't, I guess you couldcall it AI literacy is just this
(22:29):
for me, it's a bit of a mindset piece, like you said around
mindset. For me, it's a mindset piece.
Like AI, as I've said already, is there to amplify us.
But actually, what if it's get, what if we can view it as like,
it creates an abundance of opportunities of which our job
descriptions massively change because what we can do with AI
(22:50):
is crazy. But also like, what if we could
just change our job and do whatever we want to because of
AI, you know? So we need that abundance
mindset of like the opportunities are amazing if you
let it and if you have that mindset to allow it to help you
guide on whatever is needed for you to do.
(23:10):
As you start incorporating more AI systems into your business,
you're going to want to give it real data with MCP, and that can
be a little bit scary, which is why I've been using Tool Hive.
Tool Hive makes it simple and secure to use MCP.
It includes a registry of trusted MCP servers and lets me
containerize any other server with a single command.
I can install it in the client seconds, and secret protection
(23:30):
and network isolation are built in.
You can try Tool Hive too. It's free and it's open source
and you can learn more at Tool Live dot dev.
Now back to the conversation with Marty, There's tons of
concerns around people have around job displacement, but
it's very easy to look at that versus job creation.
If you go back to the Internet, to mobile, all of them have had
a similar thing where some jobs are lost, some are created and
it's about staying flexible, agile learning and a lot of the
(23:52):
things that you mentioned. And I say this that a lot.
It's either 64 or 65% of Gen. Alpha which are born the
generation born 2010 to 2013, who if they don't go to higher
education are probably coming into the workforce now or
they'll be in the workforce by twenty. 3064% of their jobs do
(24:12):
not currently exist, which meansthere's massive job creation.
But not only that, the reasons that is 76% of them would love
to be entrepreneurs. So yeah, I think as business
leaders, I think we've got to, Ipersonally think we've got to
view that in two ways. One is, can we be a thought
(24:34):
leader and think about our customer journey and and what
our customers want from us in the future. 2027 is like a
really big AI number at the moment that 2027 report came
out. So like everyone's kind of
focusing on 2027. How are your customers
operating? What's their human operating
system when consuming your products or services or join you
on your customer journey? Work back from there and think
(24:57):
about the different ways that the jobs are going to change
because there's going to be massive transformation.
There'll be jobs created. And then think about if, if Jen
Elfa is coming in and they want to be an entrepreneur and you've
done all the hard work in your industry to get to where you
are, where are they picking up from?
And if they are picking up, you know, not copycatting, but
(25:18):
business is built on other businesses, it's just the way it
is. You solve a problem, right?
Good businesses solve problems for people and that's how it
goes. And you could have three of the
same business on the High Streetif you wanted to.
And it's a competitive race, so it is going to be there.
But what then is your USP? Is it you've been a business
longer? Is it you've transformed?
Is it the type of clientele or is it the brand itself?
(25:41):
And there's so much talk out there at the moment about, you
know, making sure, you know, companies and people have
personal brands. And that's why you see bigger
companies now where the the CEO's, especially all these tech
companies, the CEO's are guessing on so many podcasts,
aren't they? Because they're trying to build
a brand so that when Replit and Relay go against each other,
(26:01):
someone goes, oh, I'm going withReplit because the CEO did this
awesome podcast. Like that is where it's gone.
And I think we need to get aheadof the next kind of generation
coming in who are gonna, I thinkprobably flourish on our fruits
left. What's the saying our labour of
fruits or fruits of labour, I can't remember, you know, but
what we're doing now is businessleaders and, you know, drudging
(26:23):
through this AI transition and then we'll get to the end.
And then we're like all these kids taking out, taking our
businesses. But you know, good on them, I
say. Yeah, I'm, I'm so thrilled at
the idea of a one to five personcompany being able to be
leveraged so much that they compete with the big people.
It's really going to change the landscaper things.
It's so exciting. One thing that I, before I
(26:46):
forget about it, I want to bringup that you mentioned is, is the
whole concept of the second brain.
I've put a lot of effort into building my personal one where I
use an Obsidian vault. I have an Obsidian plug in
generator that automates things.It's very technical and I know
that's not for everybody. And when you mentioned some of
the tools like make and Lindy, I'd love if you could kind of
explain a bit about what their strengths are, when people
should look into them, what typeof benefit they can expect to
(27:07):
drive because everyone will havedifferent needs.
But having a journal idea of where they should even start to
approach for building the systemto help them out, whether it's
their personal or professional life would be helpful.
So could you just explain like what is Lindy?
What is make? When would you choose one over
the other? Yeah.
So I'll come back to Lindy and make only because I think
there's a few other operating system that people can look into
(27:27):
first and probably should look into UNET as one which I
mentioned and the similar one tothatisyou.com.
So they're all and there's quitea few.
Basically the concept is that it's one platform, so you like
open AI, for example. So if you go to your chat, but
on that in your platform you'll have access to 22 different AI
(27:51):
models that you can choose. And then you have access to the
files and the context. So there'll be connectors which
you can do in your chat account as well, but you'll have
connectors that can connect to different files and different, I
guess calendars and you know, different contexts that you want
to give it. And then you can create the AI
(28:12):
assistant or the agent. So slight different.
The AI assistant is a general large language model of which
you will chat back and forth to it.
Whereas an agent, it has the agentic mode on and you will
give it a task and instead of you giving it the next step, it
will go off. It will decide and do the
thinking about how it's going tocomplete that task.
(28:33):
It will go off and it will do the research and it will have
another agent that writes the copy.
Then it might have another agentthat writes the code for a
landing page. So that's the agentic mode where
it does it all. Most of the large language
language models now have them built in chat.
GBT's agent mode perplexities called Labs, Claude is called
Artifacts and Gemini is called Gem Gem apps.
(28:56):
So they all have agent mode in them now.
But these platforms also have itwhere they'll have some inbuilt
templates. So it might be like, I don't
know, a content repurposing agent of which because you've
got your business brain already in your platform of which you
could have so many users sign in.
So you're all signing in to the one kind of AI model that's for
(29:18):
your business and therefore you could have different agents set
up for different departments fordifferent people within the
department, but it all connects into, you know, a live drive
type thing that will always get updated.
So I would probably look into those type of AI models first
for businesses, especially businesses you know, of more
(29:39):
than I say, you know, probably 5people.
So yeah, I'll probably look intothat first.
And then when you're thinking about agents, so Lindy is
probably my favorite because I'ma talker, as I've said many
times, and I can just go, right,this is what I think I need my
agent to do. Just as a really simple example,
(30:03):
You know, I need my agent to look at my emails that come in
daily and create three folders for me.
One is AI information, one is urgent action.
The other, you know, probably follow-ups, you know, people
that I need to follow up with that I haven't followed up with
or that they, you know, followedup with me type thing.
And then and I get, and then I say to the agent, well, I also
(30:26):
want drafts. I want draft replies.
And with the AI information, I then say, here's my 4 areas that
I like to talk about, like my 4 pillars, I like to talk about,
take the information from those emails and put them in those
into those pillars. And then from there, you know,
and I like to to check it at 8:00 AM in the morning and 5:00
PM in the evening. And then I'll get a notification
(30:48):
to say, here you go. And I'll have my drafts there
ready for me to check and my content ready for me to go
through. And another little workflow
which I haven't spoken about is I then take that content that
Lindy has done in my emails and I put it into Notebook LM, my
absolute favorite tool because on Notebook LM you can upload
(31:09):
sources. I think it's like 300 sources.
It could be YouTube videos, websites, pasted texts.
It can even search the net and then you can talk to those
sources. It can create a podcast for you.
It can create interactive like how to videos, it can create
study guides, it can create reports you could talk to the
(31:31):
AI. So I often add it as my
knowledge base and then I'll sayI'm writing my newsletter.
Here's my 4 things that I like to talk about in the newsletter.
And I also have my previous newsletters update.
Is there anything in here I haven't spoken about or have I
spoken about at all? And there's been an update and
it will, you know, look through all my information and be able
to give that back to me. So yeah, notebook LM is is
(31:51):
absolutely amazing. But again, that's all because I
basically, I asked Lindy just said to Lindy, exactly like
pretty much how I've spoken to you just now, this is what I
need. And Lindy goes and builds it.
Whereas different to say make orNAN where you've got to, it's an
automation of which you've got to put a trigger.
Then you have to connect it to alarge language model.
Then you have to say, this is what I want to happen.
(32:12):
If this happens. Like the automations, they are
getting better, don't get me wrong.
And I'm probably can't speak anymore because I don't actually
use them anymore. I probably haven't used them now
for about 3-4 months because again, I just like talking to an
AI agent rather than doing the doing.
That's awesome. That's that's one of the best
referrals for Lindy I've heard. I'm also a huge fan of Notebook
(32:35):
LM. Where I live, it's very rural,
everything's an hour drive away.I can't be reading papers and
trying to keep up with things. So I just throw a few resources
into Notebook LM. I haven't used the interactive
mode, but you've got me excited about that one.
No, this is great. This is super exciting.
Are there any other unexpected workflows that have come up with
your business or your consultingwhere even if it wasn't the
(32:57):
initial ID or thought gradually just evolved out of your
interactions where you thought Icould actually make this
automation or that's an idea fora tool that I never even
considered, but it works in my workflow?
Have been any unexpected surprises through your discovery
process? Yeah, I think 2 come to mind.
So the first one is a tool I came across.
(33:19):
I always pick myself up when I say tool because here I am
preaching you need an AI operating system.
And then I go, I've got this tool and I've got this tool.
But I generally have an AI operating system, as you know,
I've spoken about, but there aredefinitely different tools.
So interestingly, one of my clients does a lot of driving to
(33:39):
and from well within the countryhere in the UK.
He believe it or not, he's actually in cybersecurity.
So and he was really struggling to keep up with a lot of work.
So we found a tool and I use it now.
It's amazing. It's called voice pen.
So voice pen and basically we have set up his voice pan.
(33:59):
So he will just press record andhe'll start talking and
sometimes he'll talk out an e-mail or sometimes he'll, you
know, just talk about a client he's just gone and visit and you
know, brain dump or he'll talk out an article he wants to write
on LinkedIn and he just talks. And then there's a transcribe
button down the bottom and you can transcribe that into
(34:20):
anything you like. So we've set up.
So for his LinkedIn, we've customized it to write like he
writes on LinkedIn. So he pressed transcribe and
then the way that it transcribesit to LinkedIn, he can just copy
that and put that straight on LinkedIn and actually it now
connects straight to LinkedIn. So it will go straight and post
it for him. Emails, he likes to reply in a
certain way. So we've customized it so it
(34:41):
says e-mail and he just, you know, basically sends it that
way. And we've done a WhatsApp one as
well because he, you know, say WhatsApp, you don't want to have
the transcription all there likethe WhatsApp couple of lines,
bit of a space and lots of bullet points.
So that's how he does his Whatsapps.
Yeah. So we've we've customized his
voice pan so that he can literally work while driving.
(35:04):
And it yeah, I think it's been massively game changing for him.
And I really do hate that word because it's a very AI word,
game changing, revolutionized. But honestly, I think he would
and all of his team would 100% agree.
Like they're not waiting on halfthe stuff for him now because
he's doing it as he's driving. So that's one workflow.
And then surprisingly, and this comes back to vibe coding I was
(35:26):
working with, it's called a CIC over here.
So it's a community interest club.
So they have to get funding, they have to apply for grants
and funding, but they're not quite a charity.
So they can't do like lots of fundraising like other charities
can, but they can, they can apply for different grants from
government or from private companies.
And it's a bit of a numbers game, like there's a whole lot
(35:47):
of grants out there, but you've got to, you know, apply for them
and it takes them so long and you know, they're a community
interest club. So they don't, you know, they're
not paying themselves a lot and they're very time poor and they
can't afford a VA and those typeof things.
So what I did is on the most amazing Lovable with my vibe
coding skills is I did a whole lot of deep research and I found
(36:10):
all of the grants that and the funds that were eligible for
this CIC within the year. I then uploaded that into
Superbase which connected to Lovable and I said I need a
almost like a CRM that is able to divide this up into monthly
applications for this company and gave it all the information.
(36:33):
And then it had the database of the of the of the research that
I did. And then every time this company
director goes into this lovable platform that created every 30
days it comes up and it says, sothe last 30 days they had 238 to
apply for. So it comes up, it's got all the
information about the application.
(36:56):
It just gives you a link to apply straight away.
We then created a GPT, which we didn't use it as a traditional
GPT. It was a project within the AI
only because the project updatesin memory and every time they do
an application, the update memory.
And when they have a successful application, they're able to
then upload that as well. So therefore, it's constantly
updating its memory. Whereas the GPT, the memory
(37:17):
doesn't update unless you go in and update it.
So yeah, they then put that information that come off that
lovable platform, put that straight into the GPT, It fills
out the application, they go back, they fill it in.
Done. It was taking them roughly about
3 hours per application. It now takes them anywhere
between 20 and 30 minutes per application, and they're able to
do 30 a month. And it's just a numbers game.
(37:39):
So that's a work, though, that Inever, like thought would ever
happen. It was just like working with
the guy to help him, you know, with his personalized AI
operating system, like I said. And then he said, this is my
biggest bottleneck. And I'm like, we can solve that,
surely, you know. Yeah.
And those are both phenomenal use cases where we're taking
something that would happen anyway over a certain time and
(38:00):
compressing it, giving people their time back, allowing to
have a bigger impact. There's been so many times I've
been driving, I'm just like, I'mstuck in the car for three
hours. I can't do anything.
I'm offline and if you can just enable a little bit more
productivity, it gives you that evening back that you'd have to
like work late. You're clearly on top of a lot
of this stuff. You you've mentioned a lot of
great tools. What are your go to sources for
(38:20):
learning about these things, discovering them?
Do you have any pipelines that help exploit that process?
There's how do you how do you learn about these AI tools?
So I get a lot of newsletters. There's an AI tool for that.
I have the newsletter come through.
I think there's an AI for business.
There's quite a few. So and then my Lindy kind of
splits them up for me, which is good, but I'm a bit I'm very
addicted to YouTube. Another job of mine is I happen
(38:43):
to be mum's taxi. So I have to drive my children
to and from school. Lots of activities.
I do a lot of driving myself. So I do get to do a lot of
learning. I follow Matt Wolf future tools,
AIIO and is his ones hubspots marketing gets the grain is
absolutely phenomenal. Rob, the AI guy, I hope, I hope
(39:04):
he doesn't listen. He's slightly annoying, but
really good content. And then there's a guy, I think
he's based in Lisbon. He's cool.
I think he might be called the AI guy.
He's also really, really good bit slow.
So I listened to him on 1.5 speed.
But all those podcasts that I doabsolutely love.
And again, I, I really utilize notebook LM to help like kind of
(39:29):
help with my learning by either condensing it, especially I'm
trying to really stay up to datewith like SEO and AI search.
A lot of my business, I work with a lot of marketing
agencies, so they really need tokeep up with that.
And you know, there's so many different opinions and thoughts
out there that actually instead of following one person, I try
and get as many as I can. I put it into a notebook and I'm
(39:49):
like, right, let's compare them.Let's think about this from a
perspective of, you know, this type of business, you know,
because I think everybody, you know, without sounding rude or
snobby by any means, but I do think there's a big rise of AI
experts in industries. And that's awesome because, you
know, a year ago I was trying toconvince people to adopt AI.
(40:10):
Now I'm like, every time I open,I'm like, oh, this person's
become an AI expert all of a sudden.
Amazing. But it's very industry specific.
And because I work across a lot of industries, I just have to be
really careful to make sure thatI'm gaining a whole lot of
resources and not relying too much on kind of 1 prospective.
But if you haven't heard already, HubSpot are really
(40:30):
pushing the recent kind of coined term called loop
marketing, which discusses how to market in the age of AI and
talks a lot a little bit about AI search as well, which has
been really, really helpful because there's a lot of
agencies going through massive transformations and really
struggling to get their head around it.
Yeah, absolutely. It's, it's a firehose of
(40:51):
information. But if you can find a way to
distill it and and get the gems out, it's awesome.
Marty, this was a phenomenal chat.
I learned a bunch. This is really enjoyable to get
a perspective and your expertise.
Before we let you go, is there anything you want the audience
to know? I'd love to chat on LinkedIn if
anyone wants to find me on LinkedIn.
I think you said you'll put the details in the show notes.
And I love working with people. I call it a 90 day Sprint, your
(41:12):
AI, your personal AI accelerator.
I think 90 days is a really awesome day.
We awesome timeline to get a lotdone, get a lot of processes
done. So if you think you're behind,
you're not. And a 90 day Sprint with me and
you will be flying. So yeah, I welcome anybody to
come join me on that. Thank you for listening to this
conversation with Marnie Wills. I had a great time talking to
(41:32):
Marnie, full of energy, a lot ofgreat insights and ideas, and I
really hope it inspired you to start exploring and playing with
these AI tools. You don't need to know how to
code to leverage AI to help you run your business.
And if you're thinking about starting a startup or maybe a
little side hustle, there are many options available so that
you can leverage these benefits and get to your end goal faster.
I also hope you consider checking out making a second
(41:55):
brain, being able to start leveraging AI in your personal
life or have different automations happen.
It all begins by just asking questions to AI.
What's even possible? What tools exist?
What can you do? I want to give a quick shout out
to Tool Hive for supporting thischannel so I can have these type
of conversations and I'll see you next week.
Popular Podcasts
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.
Paper Ghosts: The Texas Teen Murders
Paper Ghosts: The Texas Teen Murders takes you back to 1983, when two teenagers were found murdered, execution-style, on a quiet Texas hill. What followed was decades of rumors, false leads, and a case that law enforcement could never seem to close. Now, veteran investigative journalist M. William Phelps reopens the file — uncovering new witnesses, hidden evidence, and a shocking web of deaths that may all be connected. Over nine gripping episodes, Paper Ghosts: The Texas Teen Murders unravels a story 42 years in the making… and asks the question: who’s really been hiding the truth?
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com