May 14, 2025 • 26 mins
In this episode of the Better Boards Podcast, Professor Katja Langenbucher explores how boards can embrace AI to future-proof their decision-making.
Dr. Sabine Dembkowski speaks with Katja, a law professor at Goethe-University in Frankfurt and affiliated with SciencesPo, Paris. She serves on the supervisory boards of BaFin and IEP and brings extensive boardroom and academic experience.
Making Better Judgements: Why Boards Must Embrace AI
AI is rapidly reshaping industries—from pharmaceuticals to finance—and boards can no longer afford to stand still. Katja outlines why boards must move past hesitation and actively integrate AI into their processes.
She explains how leading organisations embed AI into strategy, what this means under the business judgment rule, and why AI should challenge—not replace—human insight.
AI Isn’t a Trend—It’s Becoming a Legal Expectation
AI may still seem opaque to some directors—but that view is increasingly out of step with governance expectations. In jurisdictions applying the business judgment rule, directors must demonstrate informed, reasonable decision-making. AI is becoming part of that expectation.
“Very soon, you cannot claim to be well-informed without consulting an AI.”
Boards have long leaned on expert input for board evaluations and strategic oversight. Going forward, AI must be part of that toolkit—or boards risk falling short of legal standards.
From Coffee Chains to Capital Markets: The Real-World Power of AI
Katja cites practical use cases—like how Starbucks applies AI to optimise store locations using behavioural, geographic, and competitor data.
“You can use AI to identify an M&A target, spot a hostile takeover risk, or even test how markets might respond to your messaging.”
Yet, she observes that AI is still rarely referenced in board evaluations or agendas, despite its ability to surface risks, run scenario models, and sharpen decision-making.
The New Role of Company Secretaries
Company secretaries are ideally placed to help boards adopt AI meaningfully. Katja is clear: directors don’t need to code—they need to ask better questions.
“Nobody is asking directors to code—but boards must ask the right questions.”
Understanding a company’s proprietary data and strategic priorities is a governance task. AI experts deliver the tools, but boards must frame the questions.
Challenging Groupthink and Elevating Debate
Groupthink continues to undermine board effectiveness. Katja shares a compelling example of using AI to simulate press responses—ranging from neutral to harsh—on a sensitive issue.
“Seeing a mock ‘nasty article’ on the big screen challenged the entire board’s thinking.”
Used this way, AI becomes a catalyst for challenge and debate, broadening the board’s perspective.
AI as Induction, Humans as Interpretation
AI and human judgment are not competing forces—they are complementary. AI finds patterns. Humans interpret them.
“A good strategic decision is always a combination of AI and human thinking.”
Board evaluation frameworks must reflect this dual approach. AI accelerates insight; humans weigh impact.
Three Key Take
Remember to subscribe and never miss an episode of the Better Boards Podcast Series. It’s available on Apple, Spotify, or Google.
To find out how you can participate in the Better Boards Podcast Series or for more information on Better Boards’ solutions, please email us at info@better-boards.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Sabine (00:02):
Will AI make the better business judgment?
Welcome to the BetterBotspodcast series.
The podcast for chairs, CEOs,non-executive directors, company
secretaries and their advisors.
Every episode is filled withpractical insights and learnings
from those who are sittingaround the table.
or those that are very close tothis table.
(00:24):
We discuss what really mattersand highlight actionable steps
you can take to enhance theperformance of your board.
Join us this time where weexplore how artificial
intelligence is transformingdecision-making in the
boardroom.
As technology rapidly advances,AI is becoming an essential
tool for board members to copewith risk and uncertainty.
(00:48):
But what I'm seeing actuallywhen I analyze board papers is I
can't see very much that AI isexplicitly on the agenda.
But hey, it might change afterpeople listen to this episode.
When you ask AI for guidance,the process can look very
different from consulting ahuman expert.
(01:09):
I will analyze, can analyzevast amounts of data sets,
surface hidden patterns, andgenerate recommendations that
might never occur to even themost seasoned board
professional.
Still many board members arehesitant, and I see a lot of
this hesitation, especially asto rely on so-called black
(01:30):
boxing.
What is it actually?
What does it do with theinformation I feed?
How is my usage tracked, etcetera?
Discover why this lack ofexplainability can be daunting.
But why it is important torecognize that boards routinely
make high-stake decisions withincompatible, incomplete
(01:51):
information?
Will future-proofing boarddecision-making mean embrace AI
while remaining vigilant aboutits limitations?
Join me in a conversation withProfessor Katja Langbucher.
Katja is law professor at theGoethe University in Frankfurt
and at Science Po in Paris.
Each year she teaches also acourse at U.S.
(02:13):
law schools such as Columbia,Penn, Fordham, New York
University.
Katja is also sitting onvarious boards.
For example, she is a member ofthe German BaFin supervisory
board and was for a long time aYou are in for a really
(02:35):
interesting conversation.
And because we record thisintroduction actually after the
main part, what I really likeabout this podcast is Katja
makes it really relatable.
And there are some really greatexamples in here.
So I hope you enjoy listening.
Katja, fantastic to have you onthe Better Boards podcast
(02:55):
series.
Thank you so, so much forcontributing.
Guest Speaker - Katja L (02:58):
Sabine, thank you for the invitation.
I'm really happy to be here.
Sabine (03:02):
Fantastic.
Let's jump straight in.
I hear some noises that boardsare starting to use AI for their
decision-making.
Can you give us some examplesof what you have seen, how an AI
tool might be valuable toboards?
Guest Speaker - Katja Lan (03:18):
Yeah, and I've seen a couple of
dashboard-like things.
The one I like because it'ssuch an easy example to explain
is also available online ifpeople want to double check it,
is on Starbucks.
And they use AI to figure outthe best place for store
location, right?
(03:40):
So where should I open?
Yeah, it's like, where should Iopen a new one?
Where would I close an existingone?
And I mean, what's thechallenge?
Of course, you need to predictthe success of potential
locations.
And also you should be able tomore or less assess performance
of the existing ones in realtime.
(04:00):
And what traditionallyStarbucks and other coffee shops
have been doing is, you know,sort of a human mix of a couple
of demographic studies, foodtraffic analysis, expert
judgment, stuff like that.
And then they've started todevelop a machine learning
model.
And this analyzes, of course,an enormous data set, right?
(04:22):
So as to demographics, it'sage, it's income level, it's
lifestyle, it's what you prefer.
Then traffic, you look atpedestrian movement, you look at
vehicle movement You includecompetition.
Where are the coffee shopslocated?
Can you figure out where theycome from, where they're going
to open new ones?
You look at the localecosystem.
Is there a university?
(04:44):
Is there an office?
What about the tourist spots?
And what's also reallyinteresting, you include
behavior.
Customer behavior.
And we're already, from aEuropean point of view, getting
into tricky waters here.
But you figure out theirpurchasing patterns, previous
sales data, stuff like that.
(05:04):
Their heat maps that show youwhere are the potential areas
which would really make a lot ofmoney.
So it's really, really asignificant improvement.
And it's just one sort of easy,small-scale example.
And if you think about whereyou could use that, you know,
think about geopolitical risk.
That's something, for instance,a Spanish bank does.
(05:26):
Yeah.
But you can also identify anM&A target or a hostile takeover
threat or an IPO price, youknow, all kinds of things.
So it's interesting that we seesmall steps right now, but
there is Incredible potential atthe horizon.
I
Sabine (05:44):
mean, there is this enormous potential, but I can
tell you when I analyze theboard agenda as part of a board
evaluation, and I can say in thelast 12, 24 months, I have
seen, I can't actually evenrecall an example where AI was
explicitly on the agenda anddiscussed how the board could
(06:08):
make use of it.
Guest Speaker - Katja Lan (06:09):
Yeah, and that is interesting.
So, I mean, there are a coupleof studies out, and in my
experience, putting it on theagenda is something that at
least I have seen all over theplace.
But of course, there's more tobe done than putting it on the
agenda, right?
The really big gap we're seeingis between kind of the normal
(06:31):
middle of the road corporationin the EU or in the UK and the
big tech players is so bigbecause the big tech players
have early on understood howimportant it is to put all the
data that this corporationproduces, collects samples in
(06:54):
sort of one big data pool,metaphorically speaking, in the
middle of the corporation.
And every part of thecorporation has access to that
data pool and can use it forwhatever its task is sort of in
this world of the corporation.
And I mean, a lot of it mightbe culture.
A lot of it might be due toboard members having a certain
(07:19):
age.
We don't know.
But what is clear is that Thefuture is we will need it.
We should use it both forbusiness reasons and maybe also
for legal reasons, because atsome point the law is going to
expect you to use it.
Sabine (07:35):
That's interesting when you say this, because here where
I'm sitting in London, a lot ofpeople are seeing also the
dangers.
What does AI do and how do weneed to contain it?
So you say now almost theopposite, that the law expects
us to use it.
Guest Speaker - Katja Lan (07:52):
Yeah, I do say that.
And I mean, of course, lawyers,you know, you're all often the
killjoy.
But if you think about what thelaw says about board decision
making, in many areas, we havean explicit kind of safe harbor
from board member liability,that's called the business
(08:15):
judgment rule.
You know, codified in manyjurisdictions, not in the UK,
but the UK courts really do thesame thing without having kind
of it written in the text of alaw.
And so the idea behind that is,if there is a business judgment
on the table, and that's moreor less everything, it's geo
(08:35):
strategy, M&A, takeover, IPOpricing, whatever it is, the law
really does two things.
It says, look, I'm notinterested in a judge second
guessing a management decision.
I would like you board memberto kind of act with due care and
good faith and take decisionsyou reasonably believe in,
(08:57):
right?
But what does that mean for theboard?
If we're saying you need to actwith due care, What boards have
been doing in the past is, ofcourse, pick experts, human
experts, to inform them,challenge them, etc.
And in the same way, in myopinion, boards will need to
(09:20):
understand that Very soon, youcannot claim to be well enough
informed in that do care part ofthe law without an AI.
And so an AI, in my view, isgoing to become very soon market
standard.
And then the question a courtmight, after all, ask you is,
(09:41):
why didn't you ask the AI thatall your competitors have
started using?
So
Sabine (09:47):
what do you think?
What's the timeframe on this?
You say it soon becomes astandard.
What do you think?
How long are we talking aboutit until it is a standard in
boardrooms?
Guest Speaker - Katja Langbuc (09:57):
I mean, it depends, of course, on
what the corporation does,right?
If we're talking about techcorporations, I think we're
already there.
If we're talking about areas ofwhat the corporation does,
which has to do with huge datapools that need to be gathered,
(10:17):
evaluated, and used towards somecorporate purpose, I also think
we're already there.
So the question is more, whyand where are we not there?
And it might be reallytraditional industries, but you
look even there.
A couple of months ago, I wasvisiting an automotive factory.
(10:38):
And of course, they have robotsall over the place.
And this is AI too, right?
And then the question, whatkind of AI are we going to use
to, for instance, predictmaintenance?
It's just better.
than humans, especially ifyou're able to meaningfully
combine the two.
Sabine (10:59):
I mean, what I'm saying, of course, companies,
secretaries are using AI toolsor teams are starting to use AI
tools.
I think also directors use itas do their searches.
But what I'm not seeing enoughon the agenda is really how do
we do it as a group?
How do we deal with it as aboard?
(11:20):
And I feel really it should beon the agenda.
Guest Speaker - Katja Lan (11:24):
Yeah, I can totally agree.
And The fun thing is you say,as a group.
And part of my research outsideAI has been to look at group
decision-making on corporateboards and think about corporate
scandals such as Enron yearsago in the US or Wirecard, which
(11:46):
was a German corporate scandal,accounting scandal.
And you know what?
Groupthink is something thatmany have identified as one of
the problems that plaguecorporate boards.
And now you draw the link to AIand you can say, look, if I
(12:06):
have a meaningful, well-trainedAI, this is one way in which I
can use an AI.
I can have the AI challenge me.
A friend of mine, for instance,who is sitting on a bank board
lately told me, look, what wehave been doing is we've been
discussing a bit of a problem wehave in the bank.
And we have, while we were allsitting around the table, asking
(12:29):
an AI to draft a a pressstatement and a couple of media,
potential media articlestalking about the scandal.
And you can do that on a range,right?
Sabine (12:43):
Absolutely.
Guest Speaker - Katja Lang (12:44):
have it draft a nice kind of
article, have it draft a reallynasty article.
And then suddenly you sit therearound the board table and you
see this truly nasty article onthe big screen.
And this is going to challengewhat you've done before maybe.
And then you assess it, reviewit, and you say, well, could
that be?
What could we do if this reallyhappens?
(13:06):
So it's not in any way saying,oh, we're all going to go back
outside the boardroom and onlythe AI is taking over.
This is not what I'm talkingabout.
I'm talking about meaningfullyusing an AI to challenge,
review, critique, and integrateit in board decision making.
Sabine (13:25):
And where do you really see the difference to these
traditional support tools?
Guest Speaker - Katja Lan (13:30):
Yeah, and that's a good question.
In my view, it's reallydifferent in two ways.
So The first one is, well, ofcourse, data, right?
So depending on whatever it is,the Starbucks example, but also
just reviewing documents, whichif you're a board member, of
course, you get enormous amountsof documents you have to review
(13:53):
before you go into the boardmeeting.
Mostly, you're going to askyour staff to do that.
increasingly, you might be ableto ask an AI to do that, which
can very quickly answer concretequestions you might have
because it's so much easier forthe AI to run through these
thousands and thousands ofpages.
I mean, the second way is maybeeven more interesting and might
(14:18):
require a short explanation.
So, What an AI does is calledpattern recognition, right?
So it's really a learningmachine.
If you like metaphors, it'ssort of like taking a walk
through data.
We feed it with.
And the more sophisticated onesdo it without our help.
(14:39):
You just simply program them torespond to one initial
question.
It's called a loss function.
And then they take this lossfunction and walk through the
data and try and findcorrelations between all the
different data points.
And if you compare it to ahuman expert, I mean, we start
(15:03):
usually with a hypothesis.
Should I open the Starbucks inthis area or somewhere else?
Or should I price my IPO hereor there, right?
It's sort of a conceptual,theory-driven type of thinking.
And the AI doesn't do that atall.
And many have said, oh, an AI,and I think it's a nice word, is
(15:25):
an induction machine.
You know, it scours data anddata and data.
And what it does is identify acorrelation.
And, you know, I mean, if youthink about the IPO example, of
course, the human expert isgoing to talk about numbers,
right?
And about business plans andstuff.
But who knows, the AI might addstuff you would have never
(15:47):
thought about.
Maybe, and that maybe you wouldhave thought about a coverage
in social media, but maybe alsodata.
very specific words that wererepeatedly used during
roadshows.
There's a couple of studies onthat, for instance.
Or maybe the good looks or thebad looks, interestingly, there
are studies on that also, thebad looks of a CEO.
(16:09):
Right.
And so you're suddenly like,what?
What does that mean?
Right.
And then you need to processit.
And this is where kind of theclassic board comes back in and
the classic type of thinking,usually human thinking, causal
thinking.
Right.
And the really interestingthing I focus on in my own work
(16:31):
is thinking.
to come up with ideas on how tobring together these really
different ways of thinking, soto say, even though, of course,
the AI in itself doesn't think.
If it does that, it's sort oflike a black box maybe, right?
It spits out its prediction.
It doesn't explain it in thehuman causal way.
(16:54):
And if you think about it, itmakes perfect sense.
The AI identifies correlationsand correlations only.
And we all know correlation isnot causation.
So it's just two different waysof approaching a problem.
Sabine (17:07):
So if people are listening now, to this podcast?
And they say, gosh, yes, we useit a little bit, but we are on
the fringes.
We are still on the fringes ofit and haven't really integrated
it into our board meetings andour decision-making processes.
What have you seen working?
How can boards really getstarted?
(17:28):
I
Guest Speaker - Katja Lan (17:28):
mean, as I say, it depends very much
on what you would like to dowith the AI and what is the
context, you're asking it toprovide meaningful input, right?
So one question is, or oneexample I like to use because
it's so easy to get it, there isAlphaFold, which is a really
(17:49):
well-known foundation model AIthat has been programmed to
identify, so that's verymedical, protein folding
structures.
And why is that important?
Why could that be important fora pharmaceutical corporation?
Because if you use a medicaldrug, you need to understand how
(18:10):
it works well enough to kind ofclick the drug to the body,
right?
To the cells in the body.
So if I'm a board, forinstance, and I'm interested in
identifying new drug developmentoptions, that's a mini input an
AI gives me.
It only triggers an idea.
(18:32):
And then The classic boardstuff starts, right?
Understanding it, understandingI'm evaluating a known unknown.
Maybe it's wrong.
Maybe it's not going to work.
Let's assess it.
Let's evaluate the probabilitythat this might not work.
What would be the error costs?
(18:52):
How much do we want to investin that and stuff?
So that's one very specificone.
And AI could be used as kind ofgenerating a hint, right?
But they're very differentexamples.
You could also use it, as Iexplained before, as sort of a
discussion partner asking it,for example, for the press
(19:13):
statement or asking it tochallenge what the board came up
with in a certain situation orask it, how would a US, a German
or a Chinese market respond towhat we just said?
So, you know, there's a millionways in which you could use an
AI.
So it's hard to say, look, thisis the one way to do it.
(19:35):
The question is more Wherecould you see using it?
I think this is the firstquestion the board needs to ask
itself.
Where is an opportunity that anAI might help us?
And from there, it would moveon and say, okay, if this is
what we would like the AI to do,what should the AI look like?
Sabine (19:56):
Now people are listening and go, oh my God, yet another
thing for us to do.
We don't want yet anotherthing.
You know, the board agenda isalready full as it is and God
knows what problems we all have,you know, also on the
operational side at the moment.
Still the question, sorry torepeat.
Where to start?
(20:16):
How to make sure that we arenot bloating up the agenda even
further, rather use it to savetime?
What's the best way to do it?
Guest Speaker - Katja Lan (20:26):
Well, I mean, you know, AI itself is
going to help you to reduce theagenda.
Because another example, thereis a type of AI that is going to
analyze your board decision andthe discussions you had.
And then it's going to dowhat's called argument mapping.
So it's going to show you veryquickly, okay, these were the
(20:49):
one, two, three arguments, andthis is how I would rank them,
and now you decide.
There's also another companyI've seen.
They're going to front load.
A lot of board discussions.
So how do they do that?
Sort of like you and me, thereare pre-discussions you record,
and then you feed those in theEAI, and then the EAI does just
(21:13):
what I explained.
It's going to draw out therelevant three or four
arguments, rank and contrastthem, and then it's going to be
shorter rather than longer, theboard decision.
Because as we all know, oftenquite a bit of board discussions
Board member meetings are aboutwho gets to speak first, who
(21:34):
gets to speak second.
And, you know, everybody wantsto make space for him or
herself.
And so AI can really beefficiency enhancing rather than
bloating it up.
Sabine (21:47):
So is it actually a task for the company secretaries to
suggest to the board, here iswhere you could use AI on this
board meeting?
Guest Speaker - Kat (21:57):
Absolutely.
Absolutely.
That would be a very good wayforward to identify things that
specifically for thiscorporation might be relevant
and where AI could be valueenhancing.
And it could be about boardmeetings only.
That could be about criticallychallenging what directors are
(22:18):
proposing, right?
That could be even a compliancequestion.
It's super specific, right?
and requires a deepunderstanding of what that
corporation does and where youcould actually enhance what it
does and do it better.
And part of that is really thedata, you know?
(22:39):
And the AI, I think what I'veseen in practice mostly is that
people are afraid and they'relike, oh, AI and it's technique
and I don't understand it,right?
But this is not, yeah, but Imean, this in many ways is kind
of the wrong fear because nobodyasks the board to sit down and
code machines.
What the board is asked to dois to say, look, We have this
(23:05):
specific data pool and this iswhat we have.
This is our proprietary data.
And this is where we can use itand make it interesting and get
a competitive edge.
And the coder is not going toknow about that, right?
So that's the unique expertisethat the board is going to bring
to the table.
I mean, Amazon has been a, youknow, often quoted example,
(23:28):
started out as a book sellingbusiness, right?
And what was their unique datapool?
Well, they had access toaddresses of people and to their
preferences as to books.
And that's what I was talkingabout earlier.
If you put this data pool kindof in the middle of the
corporation, and then all kindsof departments of your
(23:49):
corporation can kind of ask thatdata pool questions.
And so if you know that a lotof people are interested in
cooking books, you might as wellsend them an email, would you
be interested in buying a set ofpot and pans, right?
And so that's why I'm sayingit's corporation specific.
The question you need to comeup with, you're interested in,
(24:13):
you wanna have the AI give youan answer to, That's the stuff
the board must think about.
And that's nothing an AI expertcan tell you.
The AI expert takes the nextstep.
You tell the AI expert, look,this is what we would like to
know.
Is there any way to get an AIto do that?
Sabine (24:31):
Great.
We could talk a lot longer, butI'm eyeing the timing here on
our recording.
And still to come, our mainquestion at the end.
What are really the three keytakeaways our listeners, what
should they take away from Yeah,I
Guest Speaker - Katja Lan (24:48):
mean, you won't be surprised that my
first takeaway would be don't belate to the party.
Using an AI one way or anotherwill become market standard and
you want to be ahead of thecurve.
And then the second one maybeis a good strategic decision.
is always a combination of thesort of thinking style of an AI
and the thinking style of ahuman.
(25:10):
And then, you know, takingthese together, the third one
might be pick the right AI andunderstand your data suited to
your corporation's specific usecase.
And the law is going toencourage, even urge you to do
Sabine (25:27):
it.
Fantastic.
Katja, thank you so, so muchfor contributing to the Better
Bots podcast series.
Guest Speaker - Katja Lan (25:33):
Thank you so much, Sabine.
It was a real pleasure to behere.
Sabine (25:36):
If you want to know where you stand in terms of the
readiness, your readiness forAI, reach out to us.
Together with the CantellusGroup, we developed an AI
readiness questionnaire.
So if you would like to haveaccess for this or like to have
information, reach out to us.
We are very happy to share itand to work with you.
(25:57):
As always, we love yourfeedback on this episode.
If you have any ideas fortopics you would like to see
covered, please do get in touch.
If you would like to hear moreabout our work, which is our
day-to-day job to do boardeffectiveness evaluations, reach
out.
Will AI make the better business judgment?
Welcome to the BetterBotspodcast series.
The podcast for chairs, CEOs,non-executive directors, company
secretaries and their advisors.
Every episode is filled withpractical insights and learnings
from those who are sittingaround the table.
or those that are very close tothis table.
(00:24):
We discuss what really mattersand highlight actionable steps
you can take to enhance theperformance of your board.
Join us this time where weexplore how artificial
intelligence is transformingdecision-making in the
boardroom.
As technology rapidly advances,AI is becoming an essential
tool for board members to copewith risk and uncertainty.
(00:48):
But what I'm seeing actuallywhen I analyze board papers is I
can't see very much that AI isexplicitly on the agenda.
But hey, it might change afterpeople listen to this episode.
When you ask AI for guidance,the process can look very
different from consulting ahuman expert.
(01:09):
I will analyze, can analyzevast amounts of data sets,
surface hidden patterns, andgenerate recommendations that
might never occur to even themost seasoned board
professional.
Still many board members arehesitant, and I see a lot of
this hesitation, especially asto rely on so-called black
(01:30):
boxing.
What is it actually?
What does it do with theinformation I feed?
How is my usage tracked, etcetera?
Discover why this lack ofexplainability can be daunting.
But why it is important torecognize that boards routinely
make high-stake decisions withincompatible, incomplete
(01:51):
information?
Will future-proofing boarddecision-making mean embrace AI
while remaining vigilant aboutits limitations?
Join me in a conversation withProfessor Katja Langbucher.
Katja is law professor at theGoethe University in Frankfurt
and at Science Po in Paris.
Each year she teaches also acourse at U.S.
(02:13):
law schools such as Columbia,Penn, Fordham, New York
University.
Katja is also sitting onvarious boards.
For example, she is a member ofthe German BaFin supervisory
board and was for a long time aYou are in for a really
(02:35):
interesting conversation.
And because we record thisintroduction actually after the
main part, what I really likeabout this podcast is Katja
makes it really relatable.
And there are some really greatexamples in here.
So I hope you enjoy listening.
Katja, fantastic to have you onthe Better Boards podcast
(02:55):
series.
Thank you so, so much forcontributing.
Guest Speaker - Katja L (02:58):
Sabine, thank you for the invitation.
I'm really happy to be here.
Sabine (03:02):
Fantastic.
Let's jump straight in.
I hear some noises that boardsare starting to use AI for their
decision-making.
Can you give us some examplesof what you have seen, how an AI
tool might be valuable toboards?
Guest Speaker - Katja Lan (03:18):
Yeah, and I've seen a couple of
dashboard-like things.
The one I like because it'ssuch an easy example to explain
is also available online ifpeople want to double check it,
is on Starbucks.
And they use AI to figure outthe best place for store
location, right?
(03:40):
So where should I open?
Yeah, it's like, where should Iopen a new one?
Where would I close an existingone?
And I mean, what's thechallenge?
Of course, you need to predictthe success of potential
locations.
And also you should be able tomore or less assess performance
of the existing ones in realtime.
(04:00):
And what traditionallyStarbucks and other coffee shops
have been doing is, you know,sort of a human mix of a couple
of demographic studies, foodtraffic analysis, expert
judgment, stuff like that.
And then they've started todevelop a machine learning
model.
And this analyzes, of course,an enormous data set, right?
(04:22):
So as to demographics, it'sage, it's income level, it's
lifestyle, it's what you prefer.
Then traffic, you look atpedestrian movement, you look at
vehicle movement You includecompetition.
Where are the coffee shopslocated?
Can you figure out where theycome from, where they're going
to open new ones?
You look at the localecosystem.
Is there a university?
(04:44):
Is there an office?
What about the tourist spots?
And what's also reallyinteresting, you include
behavior.
Customer behavior.
And we're already, from aEuropean point of view, getting
into tricky waters here.
But you figure out theirpurchasing patterns, previous
sales data, stuff like that.
(05:04):
Their heat maps that show youwhere are the potential areas
which would really make a lot ofmoney.
So it's really, really asignificant improvement.
And it's just one sort of easy,small-scale example.
And if you think about whereyou could use that, you know,
think about geopolitical risk.
That's something, for instance,a Spanish bank does.
(05:26):
Yeah.
But you can also identify anM&A target or a hostile takeover
threat or an IPO price, youknow, all kinds of things.
So it's interesting that we seesmall steps right now, but
there is Incredible potential atthe horizon.
I
Sabine (05:44):
mean, there is this enormous potential, but I can
tell you when I analyze theboard agenda as part of a board
evaluation, and I can say in thelast 12, 24 months, I have
seen, I can't actually evenrecall an example where AI was
explicitly on the agenda anddiscussed how the board could
(06:08):
make use of it.
Guest Speaker - Katja Lan (06:09):
Yeah, and that is interesting.
So, I mean, there are a coupleof studies out, and in my
experience, putting it on theagenda is something that at
least I have seen all over theplace.
But of course, there's more tobe done than putting it on the
agenda, right?
The really big gap we're seeingis between kind of the normal
(06:31):
middle of the road corporationin the EU or in the UK and the
big tech players is so bigbecause the big tech players
have early on understood howimportant it is to put all the
data that this corporationproduces, collects samples in
(06:54):
sort of one big data pool,metaphorically speaking, in the
middle of the corporation.
And every part of thecorporation has access to that
data pool and can use it forwhatever its task is sort of in
this world of the corporation.
And I mean, a lot of it mightbe culture.
A lot of it might be due toboard members having a certain
(07:19):
age.
We don't know.
But what is clear is that Thefuture is we will need it.
We should use it both forbusiness reasons and maybe also
for legal reasons, because atsome point the law is going to
expect you to use it.
Sabine (07:35):
That's interesting when you say this, because here where
I'm sitting in London, a lot ofpeople are seeing also the
dangers.
What does AI do and how do weneed to contain it?
So you say now almost theopposite, that the law expects
us to use it.
Guest Speaker - Katja Lan (07:52):
Yeah, I do say that.
And I mean, of course, lawyers,you know, you're all often the
killjoy.
But if you think about what thelaw says about board decision
making, in many areas, we havean explicit kind of safe harbor
from board member liability,that's called the business
(08:15):
judgment rule.
You know, codified in manyjurisdictions, not in the UK,
but the UK courts really do thesame thing without having kind
of it written in the text of alaw.
And so the idea behind that is,if there is a business judgment
on the table, and that's moreor less everything, it's geo
(08:35):
strategy, M&A, takeover, IPOpricing, whatever it is, the law
really does two things.
It says, look, I'm notinterested in a judge second
guessing a management decision.
I would like you board memberto kind of act with due care and
good faith and take decisionsyou reasonably believe in,
(08:57):
right?
But what does that mean for theboard?
If we're saying you need to actwith due care, What boards have
been doing in the past is, ofcourse, pick experts, human
experts, to inform them,challenge them, etc.
And in the same way, in myopinion, boards will need to
(09:20):
understand that Very soon, youcannot claim to be well enough
informed in that do care part ofthe law without an AI.
And so an AI, in my view, isgoing to become very soon market
standard.
And then the question a courtmight, after all, ask you is,
(09:41):
why didn't you ask the AI thatall your competitors have
started using?
So
Sabine (09:47):
what do you think?
What's the timeframe on this?
You say it soon becomes astandard.
What do you think?
How long are we talking aboutit until it is a standard in
boardrooms?
Guest Speaker - Katja Langbuc (09:57):
I mean, it depends, of course, on
what the corporation does,right?
If we're talking about techcorporations, I think we're
already there.
If we're talking about areas ofwhat the corporation does,
which has to do with huge datapools that need to be gathered,
(10:17):
evaluated, and used towards somecorporate purpose, I also think
we're already there.
So the question is more, whyand where are we not there?
And it might be reallytraditional industries, but you
look even there.
A couple of months ago, I wasvisiting an automotive factory.
(10:38):
And of course, they have robotsall over the place.
And this is AI too, right?
And then the question, whatkind of AI are we going to use
to, for instance, predictmaintenance?
It's just better.
than humans, especially ifyou're able to meaningfully
combine the two.
Sabine (10:59):
I mean, what I'm saying, of course, companies,
secretaries are using AI toolsor teams are starting to use AI
tools.
I think also directors use itas do their searches.
But what I'm not seeing enoughon the agenda is really how do
we do it as a group?
How do we deal with it as aboard?
(11:20):
And I feel really it should beon the agenda.
Guest Speaker - Katja Lan (11:24):
Yeah, I can totally agree.
And The fun thing is you say,as a group.
And part of my research outsideAI has been to look at group
decision-making on corporateboards and think about corporate
scandals such as Enron yearsago in the US or Wirecard, which
(11:46):
was a German corporate scandal,accounting scandal.
And you know what?
Groupthink is something thatmany have identified as one of
the problems that plaguecorporate boards.
And now you draw the link to AIand you can say, look, if I
(12:06):
have a meaningful, well-trainedAI, this is one way in which I
can use an AI.
I can have the AI challenge me.
A friend of mine, for instance,who is sitting on a bank board
lately told me, look, what wehave been doing is we've been
discussing a bit of a problem wehave in the bank.
And we have, while we were allsitting around the table, asking
(12:29):
an AI to draft a a pressstatement and a couple of media,
potential media articlestalking about the scandal.
And you can do that on a range,right?
Sabine (12:43):
Absolutely.
Guest Speaker - Katja Lang (12:44):
have it draft a nice kind of
article, have it draft a reallynasty article.
And then suddenly you sit therearound the board table and you
see this truly nasty article onthe big screen.
And this is going to challengewhat you've done before maybe.
And then you assess it, reviewit, and you say, well, could
that be?
What could we do if this reallyhappens?
(13:06):
So it's not in any way saying,oh, we're all going to go back
outside the boardroom and onlythe AI is taking over.
This is not what I'm talkingabout.
I'm talking about meaningfullyusing an AI to challenge,
review, critique, and integrateit in board decision making.
Sabine (13:25):
And where do you really see the difference to these
traditional support tools?
Guest Speaker - Katja Lan (13:30):
Yeah, and that's a good question.
In my view, it's reallydifferent in two ways.
So The first one is, well, ofcourse, data, right?
So depending on whatever it is,the Starbucks example, but also
just reviewing documents, whichif you're a board member, of
course, you get enormous amountsof documents you have to review
(13:53):
before you go into the boardmeeting.
Mostly, you're going to askyour staff to do that.
increasingly, you might be ableto ask an AI to do that, which
can very quickly answer concretequestions you might have
because it's so much easier forthe AI to run through these
thousands and thousands ofpages.
I mean, the second way is maybeeven more interesting and might
(14:18):
require a short explanation.
So, What an AI does is calledpattern recognition, right?
So it's really a learningmachine.
If you like metaphors, it'ssort of like taking a walk
through data.
We feed it with.
And the more sophisticated onesdo it without our help.
(14:39):
You just simply program them torespond to one initial
question.
It's called a loss function.
And then they take this lossfunction and walk through the
data and try and findcorrelations between all the
different data points.
And if you compare it to ahuman expert, I mean, we start
(15:03):
usually with a hypothesis.
Should I open the Starbucks inthis area or somewhere else?
Or should I price my IPO hereor there, right?
It's sort of a conceptual,theory-driven type of thinking.
And the AI doesn't do that atall.
And many have said, oh, an AI,and I think it's a nice word, is
(15:25):
an induction machine.
You know, it scours data anddata and data.
And what it does is identify acorrelation.
And, you know, I mean, if youthink about the IPO example, of
course, the human expert isgoing to talk about numbers,
right?
And about business plans andstuff.
But who knows, the AI might addstuff you would have never
(15:47):
thought about.
Maybe, and that maybe you wouldhave thought about a coverage
in social media, but maybe alsodata.
very specific words that wererepeatedly used during
roadshows.
There's a couple of studies onthat, for instance.
Or maybe the good looks or thebad looks, interestingly, there
are studies on that also, thebad looks of a CEO.
(16:09):
Right.
And so you're suddenly like,what?
What does that mean?
Right.
And then you need to processit.
And this is where kind of theclassic board comes back in and
the classic type of thinking,usually human thinking, causal
thinking.
Right.
And the really interestingthing I focus on in my own work
(16:31):
is thinking.
to come up with ideas on how tobring together these really
different ways of thinking, soto say, even though, of course,
the AI in itself doesn't think.
If it does that, it's sort oflike a black box maybe, right?
It spits out its prediction.
It doesn't explain it in thehuman causal way.
(16:54):
And if you think about it, itmakes perfect sense.
The AI identifies correlationsand correlations only.
And we all know correlation isnot causation.
So it's just two different waysof approaching a problem.
Sabine (17:07):
So if people are listening now, to this podcast?
And they say, gosh, yes, we useit a little bit, but we are on
the fringes.
We are still on the fringes ofit and haven't really integrated
it into our board meetings andour decision-making processes.
What have you seen working?
How can boards really getstarted?
(17:28):
I
Guest Speaker - Katja Lan (17:28):
mean, as I say, it depends very much
on what you would like to dowith the AI and what is the
context, you're asking it toprovide meaningful input, right?
So one question is, or oneexample I like to use because
it's so easy to get it, there isAlphaFold, which is a really
(17:49):
well-known foundation model AIthat has been programmed to
identify, so that's verymedical, protein folding
structures.
And why is that important?
Why could that be important fora pharmaceutical corporation?
Because if you use a medicaldrug, you need to understand how
(18:10):
it works well enough to kind ofclick the drug to the body,
right?
To the cells in the body.
So if I'm a board, forinstance, and I'm interested in
identifying new drug developmentoptions, that's a mini input an
AI gives me.
It only triggers an idea.
(18:32):
And then The classic boardstuff starts, right?
Understanding it, understandingI'm evaluating a known unknown.
Maybe it's wrong.
Maybe it's not going to work.
Let's assess it.
Let's evaluate the probabilitythat this might not work.
What would be the error costs?
(18:52):
How much do we want to investin that and stuff?
So that's one very specificone.
And AI could be used as kind ofgenerating a hint, right?
But they're very differentexamples.
You could also use it, as Iexplained before, as sort of a
discussion partner asking it,for example, for the press
(19:13):
statement or asking it tochallenge what the board came up
with in a certain situation orask it, how would a US, a German
or a Chinese market respond towhat we just said?
So, you know, there's a millionways in which you could use an
AI.
So it's hard to say, look, thisis the one way to do it.
(19:35):
The question is more Wherecould you see using it?
I think this is the firstquestion the board needs to ask
itself.
Where is an opportunity that anAI might help us?
And from there, it would moveon and say, okay, if this is
what we would like the AI to do,what should the AI look like?
Sabine (19:56):
Now people are listening and go, oh my God, yet another
thing for us to do.
We don't want yet anotherthing.
You know, the board agenda isalready full as it is and God
knows what problems we all have,you know, also on the
operational side at the moment.
Still the question, sorry torepeat.
Where to start?
(20:16):
How to make sure that we arenot bloating up the agenda even
further, rather use it to savetime?
What's the best way to do it?
Guest Speaker - Katja Lan (20:26):
Well, I mean, you know, AI itself is
going to help you to reduce theagenda.
Because another example, thereis a type of AI that is going to
analyze your board decision andthe discussions you had.
And then it's going to dowhat's called argument mapping.
So it's going to show you veryquickly, okay, these were the
(20:49):
one, two, three arguments, andthis is how I would rank them,
and now you decide.
There's also another companyI've seen.
They're going to front load.
A lot of board discussions.
So how do they do that?
Sort of like you and me, thereare pre-discussions you record,
and then you feed those in theEAI, and then the EAI does just
(21:13):
what I explained.
It's going to draw out therelevant three or four
arguments, rank and contrastthem, and then it's going to be
shorter rather than longer, theboard decision.
Because as we all know, oftenquite a bit of board discussions
Board member meetings are aboutwho gets to speak first, who
(21:34):
gets to speak second.
And, you know, everybody wantsto make space for him or
herself.
And so AI can really beefficiency enhancing rather than
bloating it up.
Sabine (21:47):
So is it actually a task for the company secretaries to
suggest to the board, here iswhere you could use AI on this
board meeting?
Guest Speaker - Kat (21:57):
Absolutely.
Absolutely.
That would be a very good wayforward to identify things that
specifically for thiscorporation might be relevant
and where AI could be valueenhancing.
And it could be about boardmeetings only.
That could be about criticallychallenging what directors are
(22:18):
proposing, right?
That could be even a compliancequestion.
It's super specific, right?
and requires a deepunderstanding of what that
corporation does and where youcould actually enhance what it
does and do it better.
And part of that is really thedata, you know?
(22:39):
And the AI, I think what I'veseen in practice mostly is that
people are afraid and they'relike, oh, AI and it's technique
and I don't understand it,right?
But this is not, yeah, but Imean, this in many ways is kind
of the wrong fear because nobodyasks the board to sit down and
code machines.
What the board is asked to dois to say, look, We have this
(23:05):
specific data pool and this iswhat we have.
This is our proprietary data.
And this is where we can use itand make it interesting and get
a competitive edge.
And the coder is not going toknow about that, right?
So that's the unique expertisethat the board is going to bring
to the table.
I mean, Amazon has been a, youknow, often quoted example,
(23:28):
started out as a book sellingbusiness, right?
And what was their unique datapool?
Well, they had access toaddresses of people and to their
preferences as to books.
And that's what I was talkingabout earlier.
If you put this data pool kindof in the middle of the
corporation, and then all kindsof departments of your
(23:49):
corporation can kind of ask thatdata pool questions.
And so if you know that a lotof people are interested in
cooking books, you might as wellsend them an email, would you
be interested in buying a set ofpot and pans, right?
And so that's why I'm sayingit's corporation specific.
The question you need to comeup with, you're interested in,
(24:13):
you wanna have the AI give youan answer to, That's the stuff
the board must think about.
And that's nothing an AI expertcan tell you.
The AI expert takes the nextstep.
You tell the AI expert, look,this is what we would like to
know.
Is there any way to get an AIto do that?
Sabine (24:31):
Great.
We could talk a lot longer, butI'm eyeing the timing here on
our recording.
And still to come, our mainquestion at the end.
What are really the three keytakeaways our listeners, what
should they take away from Yeah,I
Guest Speaker - Katja Lan (24:48):
mean, you won't be surprised that my
first takeaway would be don't belate to the party.
Using an AI one way or anotherwill become market standard and
you want to be ahead of thecurve.
And then the second one maybeis a good strategic decision.
is always a combination of thesort of thinking style of an AI
and the thinking style of ahuman.
(25:10):
And then, you know, takingthese together, the third one
might be pick the right AI andunderstand your data suited to
your corporation's specific usecase.
And the law is going toencourage, even urge you to do
Sabine (25:27):
it.
Fantastic.
Katja, thank you so, so muchfor contributing to the Better
Bots podcast series.
Guest Speaker - Katja Lan (25:33):
Thank you so much, Sabine.
It was a real pleasure to behere.
Sabine (25:36):
If you want to know where you stand in terms of the
readiness, your readiness forAI, reach out to us.
Together with the CantellusGroup, we developed an AI
readiness questionnaire.
So if you would like to haveaccess for this or like to have
information, reach out to us.
We are very happy to share itand to work with you.
(25:57):
As always, we love yourfeedback on this episode.
If you have any ideas fortopics you would like to see
covered, please do get in touch.
If you would like to hear moreabout our work, which is our
day-to-day job to do boardeffectiveness evaluations, reach
out.
Popular Podcasts
United States of Kennedy
United States of Kennedy is a podcast about our cultural fascination with the Kennedy dynasty. Every week, hosts Lyra Smith and George Civeris go into one aspect of the Kennedy story.
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.
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