April 18, 2023 • 21 mins
In this episode, Maya Mikhailov (Founder & AI Leader) and Andreas Welsch discuss how AI leaders can set expectations with their leadership. Maya shares examples of stakeholder management and provides valuable advice for listeners looking to balance excitement for AI with the due diligence needed for successful AI projects.
Key topics:
- Manage your senior leadership’s expectations
- Make AI tangible for their leaders
- Focus on generative AI
Listen to the full episode to hear how you can:
- Be realistic about what can be accomplished
- Educate about different types of AI
- View AI projects as a journey
Watch this episode on YouTube:
https://youtu.be/_9XrCpC1zFE
Questions or suggestions? Send me a Text Message.
***********
Disclaimer: Views are the participants’ own and do not represent those of any participant’s past, present, or future employers. Participation in this event is independent of any potential business relationship (past, present, or future) between the participants or between their employers.
Level up your AI Leadership game with the AI Leadership Handbook:
https://www.aileadershiphandbook.com
More details:
https://www.intelligence-briefing.com
All episodes:
https://www.intelligence-briefing.com/podcast
Get a weekly thought-provoking post in your inbox:
https://www.intelligence-briefing.com/newsletter
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Andreas Welsch (00:00):
Today we'll talk about managing your leader's
expectations when running AIprojects.
And I figured who better to talkto about it than someone who's
got a ton of experience instartups and corporates doing
just that.
Maya Mikhailov.
Hey Maya.
Maya Mikhailov (00:13):
Thank you for having me.
Andreas Welsch (00:15):
Awesome.
Hey why don't you tell ourguests a little bit about who
you are, what you do and how yougot here.
Maya Mikhailov (00:24):
Absolutely.
My name is Maya, I'm Co-Founderand CEO of Savvy.Ai, where we
help bring machine learningtools to every team.
Before that, my previouscompany, GPShopper, was bought
by Synchrony Financial, where Iwent on to start a new division
at Synchrony and be the generalmanager of a leading in FinTech
(00:44):
product innovation using AI andmachine learning technologies.
So we built a lot of productsusing AI and machine learning
for the bank.
And here I am.
Andreas Welsch (00:54):
Fantastic.
Hey again, thanks for joining.
Really sounds like you've seenquite a lot especially both in
startup in and corporate.
Maya, should we play a littlegame to kick things off?
What do you say?
Maya Mikhailov (01:06):
I guess we are.
Andreas Welsch (01:09):
Awesome.
Hey, this one is called In YourOwn Words.
When I hit the when I hit thebuzzer, the wheels will start
spinning and when they stop,you'll see a sentence.
And I'd like you to answer withthe first thing that comes to
mind and why.
In your own words.
And to make it a little moreinteresting, you'll only have 60
seconds for your answer.
Maya Mikhailov (01:31):
All I'm ready.
I've been practicing my gameshow.
Andreas Welsch (01:34):
Perfect.
And so folks, for those of youwatching this live, also drop
your answer in the chat and why.
Alright.
So now Maya, are you ready forWhat's the BUZZ?
Maya Mikhailov (01:46):
I am ready.
No whammies.
Andreas Welsch (01:48):
Fantastic.
Then let's roll.
So if AI were a bird, what wouldit be?
60 seconds.
Maya Mikhailov (02:02):
I think recently AI would be a phoenix.
It has risen from the ashes ofan AI winter where everyone said
AI was overhyped.
It was too much.
It's not gonna be useful.
Companies can't execute againstit.
And when ChatGPT dropped on anunsuspecting world at the end of
October, all of a sudden we sawinterest in the AI just rise
(02:22):
from the ashes.
And all of a sudden everyonewants to get in and know how it
can transform their companies,transform their teams and the
way they work.
Andreas Welsch (02:31):
Awesome.
Hey within time and an awesomeanswer.
Thank you so much.
I agree, right?
At the end of last year, we sawa lot of the media and experts
talking up the next AI winter.
And then boom.
Maya Mikhailov (02:45):
Boom.
I think there was something veryvisceral about ChatGPT.
It's not so much yes transformertechnology, large language
models.
Very interesting.
But I think what made it moreinteresting is the visceral
connection that people had.
They got it.
Before AI was always behind thescenes.
Our planes would arrive at thegate on time.
Our packages would come in.
(03:07):
Our Netflix suggestions werepersonalized.
They didn't see the AI and nowthey could see it get their
hands on it.
All of a sudden, in everyboardroom, leaders are like,
what are we doing?
What's our AI strategy?
How are we gonna make this work?
Because they themselves can seehow.
Andreas Welsch (03:23):
Excellent.
Yes.
I think that's exactly thepoint.
And that's why I'm so glad we'rehaving this conversation today
about expectations.
Because I also see them risingright in the conversations that
I have.
To your point, it's always sowhat are we doing with AI and,
to your point right now, you canactually feel it.
So look if you look at anyleadership survey, pretty much
(03:43):
AI is somewhere at the top andtop of mind for business
leaders.
But I also feel when youactually do talk to leaders,
there's still many that arestruggling to make sense of this
whole AI thing and I get it,right?
It's not necessarily their corecompetency.
And most of the time, it doesn'teven have to be.
But I also feel that thislimited literacy on AI creates
(04:05):
some challenges and sometimes amismatch of expectations.
What do you typically see?
Maya Mikhailov (04:11):
First of all, I think that I would challenge how
much literacy they actuallyneed, because they do need a bit
of AI and data literacy.
But look, leaders don'tnecessarily know how to program
yet.
They use computers every singleday.
What most business leaders wannaknow is, how will this be a
workforce multiplier?
How will this drive outcomes?
(04:32):
What they do need to know is arealistic version of what AI can
do for their particular company,given their particular data and
their particular circumstances.
They need to know the risksassociated with AI, whether
that's reputational risk,ethical risks, et cetera.
And they need to know what theycan do often with their current
team.
Many leaders are still hesitantin this economy to necessarily
(04:55):
make a huge investment or a hugepivot into a new technology.
They need to know what they cando today to achieve results in
the short and medium term sothat they can invest in the long
term.
So I think that's the premise ofwhat leaders need.
To be honest with you, I've beentalking to a lot of CEOs and
CTOs recently, and in a weirdway, they feel overwhelmed.
(05:16):
It's becoming everythingeverywhere, all at once.
AI is everywhere.
It's affecting everything.
It's transforming every singlething you do, every job, every
this, every that.
And yet it needs to be boileddown to effectively communicate
with them in a practic timeline,in a practical way.
They need to know, Hey, do youknow this problem we've been
(05:37):
having with how to efficientlyroll trucks to get product from
A to B?
This is actually a machinelearning problem.
You can solve this with AIinstead of trying to solve this
with spreadsheets and withguesses.
So they need to know how it willpractically affect their
business and generate.
Andreas Welsch (05:56):
Excellent.
Yeah.
I think that practicality isreally key.
And making sure it's connectedto the business and actually
starting with a business problemand not just the usual that
we've seen all too often (06:07):
A technology looking for a
problem.
Maya Mikhailov (06:10):
Yeah.
And, I think to a certainamount, they need to know how
they could do it with the toolsthey have.
Because if you start talking toyour leadership about AI and
machine learning, and you startwith a conversation that says:
first, we need to hire thefollowing 10 people that
everybody else is trying to hirein this economy.
(06:32):
We need to invest X, Y, Z moneythat we might not even have
because we're doing some belltightening.
We need to undergo a 24 monthdata re architecture project.
And maybe, after all of thosethings are done, we can start on
our AI journey.
And that's just, that's not whatthey wanna hear.
And that's not the reality ofthe situation.
Andreas Welsch (06:53):
Now, I'm wondering, you said you founded
two startups, you've sold one ofthem to a large financial
institution, and you've joinedthem as an AI leader and
especially in FinTech being sucha hot space and a leading space.
I was wondering how have youactually managed those
expectations towards yourleaders in a corporate
environment?
What's helped you and how haveyou gone about it?
Maya Mikhailov (07:14):
I was really lucky at Synchrony Financial.
I had an incredibly supportiveleadership team.
I had an incredibly supportivebo board, and they saw the
potential of what we were doing.
They saw AI, they saw machinelearning, and they saw that it
had potential not just in thesebig use cases that every bank is
going after, like anti-moneylaundering and fraud.
(07:34):
But it had that potential asthat workforce multiplier, as
that intelligence level up oftheir software and their
products.
So in that sense, I feel like Iwas very fortunate.
But the reality is also a lot ofthe support that we are able to
get for our division was becausewe were able to tell a story and
we were able to tell a storyaround what we are building
(07:55):
around how the goals of what weare building.
Aligned with what the company'sgoals were.
It seems pretty simplistic, butyou'd be surprised at how many
data leaders, AI leaders getstuck in the mechanics of what
they're doing and they forget.
They get lost in the forest andthey forget to bring it back to
the overall corporate goals andobjectives.
(08:16):
We came armed with numbers.
Here's how it's gonna help.
Here are the results, what we'reexpecting.
And we didn't necessarilydiscuss with the board and
leadership things like rootmeans, square error or
heteroscedasticity.
Very few boards wanna discussheteroscedasticity.
Maybe at OpenAI they do.
But more boards are like, whatare the outcomes?
What are the risks?
(08:36):
What can we accomplish by nextquarter, by two quarters from
now, et cetera.
So when we framed this as astory and a narrative that they
can digest and understand andpull back to their own corporate
goals and objectives, we weremuch more successful.
Andreas Welsch (08:50):
I think that's a really good recommendation.
Frame and phrasing it in simpleterms, in terms that relate to
these goals and to theseobjectives.
Look I'm wondering also now,because AI is so much more
accessible and it's accessibleto anyone.
And I feel if you've beenfollowing the news even at the
(09:11):
beginning of the year with.
Maya Mikhailov (09:13):
I've been following the news so much that
I'm inundated with news, likethere's like a new AI
announcement while we're talkinglike.
Four new AI startups havelaunched while we're having this
conversation.
Andreas Welsch (09:24):
Fully agree, right?
It's so hard to stay on top ofthe news and some days it feels
like your head is spinning justtrying to keep up.
Now, I feel with World EconomicForum Davos at the beginning of
the year, that's when I feel itwas really propelled as a topic
again.
And, it was so much top of mindnow with AI and generative AI
(09:45):
being so accessible to anyoneincluding business leaders where
it's no longer just technologyconversation.
How do you see these ormanagement of these
expectations?
Because, hey, look, ChatGPT issuper easy.
It's super simple to use.
I can just pop in a question andI get a response.
Why can't we do that?
How do you manage that gapbetween what you mentioned,
(10:08):
right?
There are some foundationalthings that you need to have in
place.
Some are there, some are on theway of getting there, before you
can use generative AI.
Does it always have to be builtfrom scratch, for example, or
most of all, how do you managethat expectation?
Maya Mikhailov (10:23):
I think the first thing is and I really have
a lot of sympathy right now forAI leaders because they have to
walk such a fine line.
They have to walk a fine linebetween being hype master,
which, and the, hyper of AI iseventually gonna lead into
problems when it doesn't do allthe magical things that they
read about in the media.
(10:44):
And they have to walk a fineline between being like a wet
blanket.
Sandbagging all the results sothat the overall company gets
impatient or the business getsimpatient with waiting for these
deliverables to happen.
So they're right now juggling alot of plates.
But the reality is, that firstof all, a business with
generative AI, they need toestablish a framework.
They need to establish aframework of behavior of what's
(11:05):
okay.
They need to establish a datasecurity framework.
The folks at Samsung, thosedevelopers may not have known
that they can't put proprietarysuper secret chip data into
ChatGPT, because it helps themdocument or QA a process that
they were doing.
We're human beings.
Human beings naturally gravitateto convenience.
(11:26):
So the first thing the companyhas to do is figure out a
framework of what makes themcomfortable with these
generative AI tools, and thenestablish a knowledge base help.
Let that empower the SMEs toestablish a knowledge base,
because they know how thesetools can help their businesses
and help their businessessucceed.
And so they can establish theseare the top prompts I use.
(11:48):
But the second thing is and,this is very, important, AI is
not just ChatGPT.
And even though ChatGPT is athing we can wrap our brains
around, there isn't one model torule them all.
This isn't Westworld quite yet.
So there isn't this one magicmodel that's gonna solve your
(12:09):
problems.
ChatGPT will not tell you whichtruck to roll in your warehouse.
ChatGPT will not make acontinuous decision for you
about what piece of content toput in front of which users at
any given time on your website.
It won't do certain things.
So the first thing you have todo is reframe your management
(12:29):
into, there are certain thingsthat ChatGPT is good at.
There are certain things that AIvis-a-vis machine learning or
optical recognition is good at.
And then look at your problemsets.
So the first thing you have todo is, go look at the problems
across your business andestablish a roadmap of what can
be reasonably accomplished.
What are these small andmid-size wins that can be
(12:51):
accomplished with small tomid-size risks so that you can
start building credibility forlarger AI programs?
If your business is a little bithesitant about it, if your
business is just saying, get onthe gas, there certainly is more
opportunity there, but on theother hand, you also have to
think again about data privacy,about data security, and about
(13:12):
the ethics of what you're doingand where it can be used and
where it cannot be used.
Not just because you're in aregulated industry, but because
you're not gonna fire hose allyour company's private data into
somebody else's open model.
Andreas Welsch (13:26):
There's a lot of a lot of insight and a lot of
truth that, that you're sharing.
Because things that you've builtbefore the advent of generative
AI, they're still valid.
Maya Mikhailov (13:37):
People ask me what I do and I was like, I'm
just in boring machine learning.
You don't wanna talk about that.
But the reality is, it's thatboring machine learning that's
actually gonna be that hiddenworkforce multiplier, while
everyone else is too busyputting into ChatGPT, how do I
write a memo to management aboutthe wonderful things we're doing
with AI?
Andreas Welsch (13:58):
I'm taking a look at the chat and I see
Michael has the insightfulquestion here, and he says, Will
domain specific tools likeBloombergGPT for finance or
Harvey AI for legal researchbecome the new normal in the
next 24 months?
What do you see?
What do you believe?
Maya Mikhailov (14:13):
When you're talking about large language
models and some of thesegenerative AI models.
I really think that you're justgonna see some specialization as
companies discover that theydon't need the whole history of
language.
They need language that'sspecific to their business.
They need knowledge that'sspecific to their business.
So do I see a rise inpurpose-built tools that are
(14:35):
specific to certain industrieslike healthcare or like
BloombergGPT?
Absolutely.
But I still see a wide range ofproblems.
That are very common across manybusinesses that AI and machine
learning can address.
But yeah, I definitely see withlarge language models there's
gonna be a lot of kind of hypertuning, if you will, against
(14:56):
certain industries and certaindata sets as companies will also
demand, like their own LLM.
Because they don't wannanecessarily feed back that data
into a generalized model,especially if it's company if
it's chip design data, that'sthe most proprietary data they
own.
Andreas Welsch (15:13):
That makes sense.
I think it'll be reallyinteresting seeing where this
goes and how many of thesemodels or what types of models
will be available publicly orwill be available commercially
to the point of BloombergGPT andsimilar ones.
I think definitely an excitingspace we get to be a part of and
get to shape over the nextcouple of years.
Maya Mikhailov (15:34):
Yeah.
It's a very exciting space and Ithink it's evolving.
Not just because there's newnews coming out about
transformer models and LLMsevery single day.
I think it's an exciting space,because companies are still
trying to figure out their waythrough it and figure out we've
been told for the last decadethat data is oil.
(15:54):
That our data is the mostprecious resource we have as a
business.
We have to protect it at allcosts.
And now all of a sudden peopleare saying now just put all of
that in my black box model.
So I think a lot of companiesare trying to figure out, is
that the right strategy for us?
What is the right strategy forus?
What types of AI or machinelearning work for us versus what
types don't really generate ahuge benefit to the business?
(16:17):
And again, what can we executetoday in the sh short to medium
term?
To start getting some of theseresults under our belts and some
of this learning and experie.
Andreas Welsch (16:28):
I do have a follow up question to that one.
And what we've also talked aboutnot throwing the old AI stuff
out the window and that is inall of this excitement, hype,
but also complexity, technicalcomplexity new security, ethical
questions coming up, what roledo you see AI leaders play?
(16:49):
What role can they play in theirorganizations right now?
What do they need to do to helptheir organization be successful
and grow and thrive in thegenerative AI era?
Maya Mikhailov (17:00):
I think they play a very important role right
now, because right now they'rethe ones educating the
organization.
They're the ones creating thisframework, and that's what's
really important to create aframework where your company can
succeed to create a securityframework, a data usage
framework a knowledge base.
These data leaders, these AIleaders are at the forefront of
(17:20):
helping their company design howthis AI transformation will
affect the business in the nextyear, 5 years, 10 years.
They have a very important roleto play and they have an
important role to play intoeducating teams into what AI can
do for them and can't do forthem.
Because again, the more there isa sensationalism about what AI
(17:43):
may become and that it may takeover humanity and welcome our
new AI overlords, they have arole to play in saying, look,
this is what this means for youtoday.
This is what we can accomplishwith our team today.
So I, think that they are boththe educator and they are the
realizer of some of thosecorporate goals and objectives.
Andreas Welsch (18:04):
I think that's a fantastic message for those of
you in the audience, if you'rein an AI leadership role or
aspiring to become an AI leader.
There's a real opportunity andespecially also a real need here
to help your company, to helpyour peers, to help your
leadership better understandwhat all of this AI stuff is
about and what's tangible andwhat you can do with it and what
(18:24):
you can't or shouldn't be doingwith it.
Maya Mikhailov (18:28):
Absolutely.
And to lay the foundation rightnow of a successful AI
transformation, because that iswhat's next.
We've we've spent the moneycreating data streams, data
rivers, data lake houses.
I don't know what else we'rebuilding on that data, but we
data whitewater rafting, ifyou're in California this spring
(18:50):
we've spent the resources to dothat, and now it's time to put
that data to action.
And that's what the promise ofAI is.
That automation thatproductivity and efficiency gain
of putting that data to work forus rather than just staring at
it on a dashboard.
Andreas Welsch (19:06):
Fantastic.
Hey, maybe can you summarize thethree key takeaways for our
audience today, because we'regetting close to the end of the
show.
Maya Mikhailov (19:13):
Absolutely.
I think first of all, whentalking to your leadership about
data and AI projects,specifically around AI projects,
you have to ground them in whatcan be accomplished.
You have to ground them in whatcan be accomplished today with
the resources you have and whatyour roadmap is in the future.
I think you also have to helpeducate them as to the different
types of AI that it's not.
(19:35):
ChatGPT is the one model to rulethem all, that there's still
machine learning out there.
There's still other types of AIout there that can help their
business more practically in theshort term and accomplish the
results they need.
And finally, I think you have toremember that this is a longer
term effort.
All of this transformation thatwe talk about, all of these AI
(19:56):
gains that we're talking about,they don't necessarily happen
tomorrow.
So I think there is a line towalk between getting your
leadership excited, bringing itback to their goals and
objectives, and realizing thatyou're at the beginning of a
journey.
Show some wins during thejourney.
Tell the story.
Tell how it relates back to yourbusiness objectives.
But don't forget it is in fact ajourney.
Andreas Welsch (20:18):
That's awesome.
I think that's a very practicaland a very realistic assessment
as, as well that it is a journeyand then you're, in it for, the
long run and, not just not justfor a sprint.
Thank you so much for joiningus, Maya, and for sharing your
expertise with us.
It was great having you on theshow.
Maya Mikhailov (20:34):
Thank you so much for having me, and thank
you everybody who joined us.
Today we'll talk about managing your leader's
expectations when running AIprojects.
And I figured who better to talkto about it than someone who's
got a ton of experience instartups and corporates doing
just that.
Maya Mikhailov.
Hey Maya.
Maya Mikhailov (00:13):
Thank you for having me.
Andreas Welsch (00:15):
Awesome.
Hey why don't you tell ourguests a little bit about who
you are, what you do and how yougot here.
Maya Mikhailov (00:24):
Absolutely.
My name is Maya, I'm Co-Founderand CEO of Savvy.Ai, where we
help bring machine learningtools to every team.
Before that, my previouscompany, GPShopper, was bought
by Synchrony Financial, where Iwent on to start a new division
at Synchrony and be the generalmanager of a leading in FinTech
(00:44):
product innovation using AI andmachine learning technologies.
So we built a lot of productsusing AI and machine learning
for the bank.
And here I am.
Andreas Welsch (00:54):
Fantastic.
Hey again, thanks for joining.
Really sounds like you've seenquite a lot especially both in
startup in and corporate.
Maya, should we play a littlegame to kick things off?
What do you say?
Maya Mikhailov (01:06):
I guess we are.
Andreas Welsch (01:09):
Awesome.
Hey, this one is called In YourOwn Words.
When I hit the when I hit thebuzzer, the wheels will start
spinning and when they stop,you'll see a sentence.
And I'd like you to answer withthe first thing that comes to
mind and why.
In your own words.
And to make it a little moreinteresting, you'll only have 60
seconds for your answer.
Maya Mikhailov (01:31):
All I'm ready.
I've been practicing my gameshow.
Andreas Welsch (01:34):
Perfect.
And so folks, for those of youwatching this live, also drop
your answer in the chat and why.
Alright.
So now Maya, are you ready forWhat's the BUZZ?
Maya Mikhailov (01:46):
I am ready.
No whammies.
Andreas Welsch (01:48):
Fantastic.
Then let's roll.
So if AI were a bird, what wouldit be?
60 seconds.
Maya Mikhailov (02:02):
I think recently AI would be a phoenix.
It has risen from the ashes ofan AI winter where everyone said
AI was overhyped.
It was too much.
It's not gonna be useful.
Companies can't execute againstit.
And when ChatGPT dropped on anunsuspecting world at the end of
October, all of a sudden we sawinterest in the AI just rise
(02:22):
from the ashes.
And all of a sudden everyonewants to get in and know how it
can transform their companies,transform their teams and the
way they work.
Andreas Welsch (02:31):
Awesome.
Hey within time and an awesomeanswer.
Thank you so much.
I agree, right?
At the end of last year, we sawa lot of the media and experts
talking up the next AI winter.
And then boom.
Maya Mikhailov (02:45):
Boom.
I think there was something veryvisceral about ChatGPT.
It's not so much yes transformertechnology, large language
models.
Very interesting.
But I think what made it moreinteresting is the visceral
connection that people had.
They got it.
Before AI was always behind thescenes.
Our planes would arrive at thegate on time.
Our packages would come in.
(03:07):
Our Netflix suggestions werepersonalized.
They didn't see the AI and nowthey could see it get their
hands on it.
All of a sudden, in everyboardroom, leaders are like,
what are we doing?
What's our AI strategy?
How are we gonna make this work?
Because they themselves can seehow.
Andreas Welsch (03:23):
Excellent.
Yes.
I think that's exactly thepoint.
And that's why I'm so glad we'rehaving this conversation today
about expectations.
Because I also see them risingright in the conversations that
I have.
To your point, it's always sowhat are we doing with AI and,
to your point right now, you canactually feel it.
So look if you look at anyleadership survey, pretty much
(03:43):
AI is somewhere at the top andtop of mind for business
leaders.
But I also feel when youactually do talk to leaders,
there's still many that arestruggling to make sense of this
whole AI thing and I get it,right?
It's not necessarily their corecompetency.
And most of the time, it doesn'teven have to be.
But I also feel that thislimited literacy on AI creates
(04:05):
some challenges and sometimes amismatch of expectations.
What do you typically see?
Maya Mikhailov (04:11):
First of all, I think that I would challenge how
much literacy they actuallyneed, because they do need a bit
of AI and data literacy.
But look, leaders don'tnecessarily know how to program
yet.
They use computers every singleday.
What most business leaders wannaknow is, how will this be a
workforce multiplier?
How will this drive outcomes?
(04:32):
What they do need to know is arealistic version of what AI can
do for their particular company,given their particular data and
their particular circumstances.
They need to know the risksassociated with AI, whether
that's reputational risk,ethical risks, et cetera.
And they need to know what theycan do often with their current
team.
Many leaders are still hesitantin this economy to necessarily
(04:55):
make a huge investment or a hugepivot into a new technology.
They need to know what they cando today to achieve results in
the short and medium term sothat they can invest in the long
term.
So I think that's the premise ofwhat leaders need.
To be honest with you, I've beentalking to a lot of CEOs and
CTOs recently, and in a weirdway, they feel overwhelmed.
(05:16):
It's becoming everythingeverywhere, all at once.
AI is everywhere.
It's affecting everything.
It's transforming every singlething you do, every job, every
this, every that.
And yet it needs to be boileddown to effectively communicate
with them in a practic timeline,in a practical way.
They need to know, Hey, do youknow this problem we've been
(05:37):
having with how to efficientlyroll trucks to get product from
A to B?
This is actually a machinelearning problem.
You can solve this with AIinstead of trying to solve this
with spreadsheets and withguesses.
So they need to know how it willpractically affect their
business and generate.
Andreas Welsch (05:56):
Excellent.
Yeah.
I think that practicality isreally key.
And making sure it's connectedto the business and actually
starting with a business problemand not just the usual that
we've seen all too often (06:07):
A technology looking for a
problem.
Maya Mikhailov (06:10):
Yeah.
And, I think to a certainamount, they need to know how
they could do it with the toolsthey have.
Because if you start talking toyour leadership about AI and
machine learning, and you startwith a conversation that says:
first, we need to hire thefollowing 10 people that
everybody else is trying to hirein this economy.
(06:32):
We need to invest X, Y, Z moneythat we might not even have
because we're doing some belltightening.
We need to undergo a 24 monthdata re architecture project.
And maybe, after all of thosethings are done, we can start on
our AI journey.
And that's just, that's not whatthey wanna hear.
And that's not the reality ofthe situation.
Andreas Welsch (06:53):
Now, I'm wondering, you said you founded
two startups, you've sold one ofthem to a large financial
institution, and you've joinedthem as an AI leader and
especially in FinTech being sucha hot space and a leading space.
I was wondering how have youactually managed those
expectations towards yourleaders in a corporate
environment?
What's helped you and how haveyou gone about it?
Maya Mikhailov (07:14):
I was really lucky at Synchrony Financial.
I had an incredibly supportiveleadership team.
I had an incredibly supportivebo board, and they saw the
potential of what we were doing.
They saw AI, they saw machinelearning, and they saw that it
had potential not just in thesebig use cases that every bank is
going after, like anti-moneylaundering and fraud.
(07:34):
But it had that potential asthat workforce multiplier, as
that intelligence level up oftheir software and their
products.
So in that sense, I feel like Iwas very fortunate.
But the reality is also a lot ofthe support that we are able to
get for our division was becausewe were able to tell a story and
we were able to tell a storyaround what we are building
(07:55):
around how the goals of what weare building.
Aligned with what the company'sgoals were.
It seems pretty simplistic, butyou'd be surprised at how many
data leaders, AI leaders getstuck in the mechanics of what
they're doing and they forget.
They get lost in the forest andthey forget to bring it back to
the overall corporate goals andobjectives.
(08:16):
We came armed with numbers.
Here's how it's gonna help.
Here are the results, what we'reexpecting.
And we didn't necessarilydiscuss with the board and
leadership things like rootmeans, square error or
heteroscedasticity.
Very few boards wanna discussheteroscedasticity.
Maybe at OpenAI they do.
But more boards are like, whatare the outcomes?
What are the risks?
(08:36):
What can we accomplish by nextquarter, by two quarters from
now, et cetera.
So when we framed this as astory and a narrative that they
can digest and understand andpull back to their own corporate
goals and objectives, we weremuch more successful.
Andreas Welsch (08:50):
I think that's a really good recommendation.
Frame and phrasing it in simpleterms, in terms that relate to
these goals and to theseobjectives.
Look I'm wondering also now,because AI is so much more
accessible and it's accessibleto anyone.
And I feel if you've beenfollowing the news even at the
(09:11):
beginning of the year with.
Maya Mikhailov (09:13):
I've been following the news so much that
I'm inundated with news, likethere's like a new AI
announcement while we're talkinglike.
Four new AI startups havelaunched while we're having this
conversation.
Andreas Welsch (09:24):
Fully agree, right?
It's so hard to stay on top ofthe news and some days it feels
like your head is spinning justtrying to keep up.
Now, I feel with World EconomicForum Davos at the beginning of
the year, that's when I feel itwas really propelled as a topic
again.
And, it was so much top of mindnow with AI and generative AI
(09:45):
being so accessible to anyoneincluding business leaders where
it's no longer just technologyconversation.
How do you see these ormanagement of these
expectations?
Because, hey, look, ChatGPT issuper easy.
It's super simple to use.
I can just pop in a question andI get a response.
Why can't we do that?
How do you manage that gapbetween what you mentioned,
(10:08):
right?
There are some foundationalthings that you need to have in
place.
Some are there, some are on theway of getting there, before you
can use generative AI.
Does it always have to be builtfrom scratch, for example, or
most of all, how do you managethat expectation?
Maya Mikhailov (10:23):
I think the first thing is and I really have
a lot of sympathy right now forAI leaders because they have to
walk such a fine line.
They have to walk a fine linebetween being hype master,
which, and the, hyper of AI iseventually gonna lead into
problems when it doesn't do allthe magical things that they
read about in the media.
(10:44):
And they have to walk a fineline between being like a wet
blanket.
Sandbagging all the results sothat the overall company gets
impatient or the business getsimpatient with waiting for these
deliverables to happen.
So they're right now juggling alot of plates.
But the reality is, that firstof all, a business with
generative AI, they need toestablish a framework.
They need to establish aframework of behavior of what's
(11:05):
okay.
They need to establish a datasecurity framework.
The folks at Samsung, thosedevelopers may not have known
that they can't put proprietarysuper secret chip data into
ChatGPT, because it helps themdocument or QA a process that
they were doing.
We're human beings.
Human beings naturally gravitateto convenience.
(11:26):
So the first thing the companyhas to do is figure out a
framework of what makes themcomfortable with these
generative AI tools, and thenestablish a knowledge base help.
Let that empower the SMEs toestablish a knowledge base,
because they know how thesetools can help their businesses
and help their businessessucceed.
And so they can establish theseare the top prompts I use.
(11:48):
But the second thing is and,this is very, important, AI is
not just ChatGPT.
And even though ChatGPT is athing we can wrap our brains
around, there isn't one model torule them all.
This isn't Westworld quite yet.
So there isn't this one magicmodel that's gonna solve your
(12:09):
problems.
ChatGPT will not tell you whichtruck to roll in your warehouse.
ChatGPT will not make acontinuous decision for you
about what piece of content toput in front of which users at
any given time on your website.
It won't do certain things.
So the first thing you have todo is reframe your management
(12:29):
into, there are certain thingsthat ChatGPT is good at.
There are certain things that AIvis-a-vis machine learning or
optical recognition is good at.
And then look at your problemsets.
So the first thing you have todo is, go look at the problems
across your business andestablish a roadmap of what can
be reasonably accomplished.
What are these small andmid-size wins that can be
(12:51):
accomplished with small tomid-size risks so that you can
start building credibility forlarger AI programs?
If your business is a little bithesitant about it, if your
business is just saying, get onthe gas, there certainly is more
opportunity there, but on theother hand, you also have to
think again about data privacy,about data security, and about
(13:12):
the ethics of what you're doingand where it can be used and
where it cannot be used.
Not just because you're in aregulated industry, but because
you're not gonna fire hose allyour company's private data into
somebody else's open model.
Andreas Welsch (13:26):
There's a lot of a lot of insight and a lot of
truth that, that you're sharing.
Because things that you've builtbefore the advent of generative
AI, they're still valid.
Maya Mikhailov (13:37):
People ask me what I do and I was like, I'm
just in boring machine learning.
You don't wanna talk about that.
But the reality is, it's thatboring machine learning that's
actually gonna be that hiddenworkforce multiplier, while
everyone else is too busyputting into ChatGPT, how do I
write a memo to management aboutthe wonderful things we're doing
with AI?
Andreas Welsch (13:58):
I'm taking a look at the chat and I see
Michael has the insightfulquestion here, and he says, Will
domain specific tools likeBloombergGPT for finance or
Harvey AI for legal researchbecome the new normal in the
next 24 months?
What do you see?
What do you believe?
Maya Mikhailov (14:13):
When you're talking about large language
models and some of thesegenerative AI models.
I really think that you're justgonna see some specialization as
companies discover that theydon't need the whole history of
language.
They need language that'sspecific to their business.
They need knowledge that'sspecific to their business.
So do I see a rise inpurpose-built tools that are
(14:35):
specific to certain industrieslike healthcare or like
BloombergGPT?
Absolutely.
But I still see a wide range ofproblems.
That are very common across manybusinesses that AI and machine
learning can address.
But yeah, I definitely see withlarge language models there's
gonna be a lot of kind of hypertuning, if you will, against
(14:56):
certain industries and certaindata sets as companies will also
demand, like their own LLM.
Because they don't wannanecessarily feed back that data
into a generalized model,especially if it's company if
it's chip design data, that'sthe most proprietary data they
own.
Andreas Welsch (15:13):
That makes sense.
I think it'll be reallyinteresting seeing where this
goes and how many of thesemodels or what types of models
will be available publicly orwill be available commercially
to the point of BloombergGPT andsimilar ones.
I think definitely an excitingspace we get to be a part of and
get to shape over the nextcouple of years.
Maya Mikhailov (15:34):
Yeah.
It's a very exciting space and Ithink it's evolving.
Not just because there's newnews coming out about
transformer models and LLMsevery single day.
I think it's an exciting space,because companies are still
trying to figure out their waythrough it and figure out we've
been told for the last decadethat data is oil.
(15:54):
That our data is the mostprecious resource we have as a
business.
We have to protect it at allcosts.
And now all of a sudden peopleare saying now just put all of
that in my black box model.
So I think a lot of companiesare trying to figure out, is
that the right strategy for us?
What is the right strategy forus?
What types of AI or machinelearning work for us versus what
types don't really generate ahuge benefit to the business?
(16:17):
And again, what can we executetoday in the sh short to medium
term?
To start getting some of theseresults under our belts and some
of this learning and experie.
Andreas Welsch (16:28):
I do have a follow up question to that one.
And what we've also talked aboutnot throwing the old AI stuff
out the window and that is inall of this excitement, hype,
but also complexity, technicalcomplexity new security, ethical
questions coming up, what roledo you see AI leaders play?
(16:49):
What role can they play in theirorganizations right now?
What do they need to do to helptheir organization be successful
and grow and thrive in thegenerative AI era?
Maya Mikhailov (17:00):
I think they play a very important role right
now, because right now they'rethe ones educating the
organization.
They're the ones creating thisframework, and that's what's
really important to create aframework where your company can
succeed to create a securityframework, a data usage
framework a knowledge base.
These data leaders, these AIleaders are at the forefront of
(17:20):
helping their company design howthis AI transformation will
affect the business in the nextyear, 5 years, 10 years.
They have a very important roleto play and they have an
important role to play intoeducating teams into what AI can
do for them and can't do forthem.
Because again, the more there isa sensationalism about what AI
(17:43):
may become and that it may takeover humanity and welcome our
new AI overlords, they have arole to play in saying, look,
this is what this means for youtoday.
This is what we can accomplishwith our team today.
So I, think that they are boththe educator and they are the
realizer of some of thosecorporate goals and objectives.
Andreas Welsch (18:04):
I think that's a fantastic message for those of
you in the audience, if you'rein an AI leadership role or
aspiring to become an AI leader.
There's a real opportunity andespecially also a real need here
to help your company, to helpyour peers, to help your
leadership better understandwhat all of this AI stuff is
about and what's tangible andwhat you can do with it and what
(18:24):
you can't or shouldn't be doingwith it.
Maya Mikhailov (18:28):
Absolutely.
And to lay the foundation rightnow of a successful AI
transformation, because that iswhat's next.
We've we've spent the moneycreating data streams, data
rivers, data lake houses.
I don't know what else we'rebuilding on that data, but we
data whitewater rafting, ifyou're in California this spring
(18:50):
we've spent the resources to dothat, and now it's time to put
that data to action.
And that's what the promise ofAI is.
That automation thatproductivity and efficiency gain
of putting that data to work forus rather than just staring at
it on a dashboard.
Andreas Welsch (19:06):
Fantastic.
Hey, maybe can you summarize thethree key takeaways for our
audience today, because we'regetting close to the end of the
show.
Maya Mikhailov (19:13):
Absolutely.
I think first of all, whentalking to your leadership about
data and AI projects,specifically around AI projects,
you have to ground them in whatcan be accomplished.
You have to ground them in whatcan be accomplished today with
the resources you have and whatyour roadmap is in the future.
I think you also have to helpeducate them as to the different
types of AI that it's not.
(19:35):
ChatGPT is the one model to rulethem all, that there's still
machine learning out there.
There's still other types of AIout there that can help their
business more practically in theshort term and accomplish the
results they need.
And finally, I think you have toremember that this is a longer
term effort.
All of this transformation thatwe talk about, all of these AI
(19:56):
gains that we're talking about,they don't necessarily happen
tomorrow.
So I think there is a line towalk between getting your
leadership excited, bringing itback to their goals and
objectives, and realizing thatyou're at the beginning of a
journey.
Show some wins during thejourney.
Tell the story.
Tell how it relates back to yourbusiness objectives.
But don't forget it is in fact ajourney.
Andreas Welsch (20:18):
That's awesome.
I think that's a very practicaland a very realistic assessment
as, as well that it is a journeyand then you're, in it for, the
long run and, not just not justfor a sprint.
Thank you so much for joiningus, Maya, and for sharing your
expertise with us.
It was great having you on theshow.
Maya Mikhailov (20:34):
Thank you so much for having me, and thank
you everybody who joined us.
Popular Podcasts
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
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Therapy Gecko
An unlicensed lizard psychologist travels the universe talking to strangers about absolutely nothing. TO CALL THE GECKO: follow me on https://www.twitch.tv/lyleforever to get a notification for when I am taking calls. I am usually live Mondays, Wednesdays, and Fridays but lately a lot of other times too. I am a gecko.