Latest / The In-Between Tech and Trust Podcast / The Widening AI Value Gap: What Scaling AI Really Demands (EP 16)

Transcript

The Widening AI Value Gap: What Scaling AI Really Demands (EP 16) · The In-Between Tech and Trust Podcast

0:14
Hello, this is Eva and you're listening to the In Between Tech
0:17
and Trust podcast. And this week's episode gets
0:21
quite practical because I talked to Doctor Mark Roman Frankel and
0:25
he's partner and Associate Director at BCG, responsible for
0:29
the AI and digital transformation practice.
0:32
And together with him, we dive into one of the latest studies
0:36
that was published in September 2025, and it's called The
0:40
Widening Value Gap. And if you're like me and you
0:43
read a lot about AI and tech transformations, you've heard
0:46
about the stat that only 5% of businesses actually gain value
0:52
from their AI projects. And Roman was one of the experts
0:57
that interviewed good approximately over 1200
1:02
organizations and found out that this gap actually exists and
1:06
that the statistic is representative for how AI use
1:11
cases are being incorporated or how they also fail.
1:15
So we dove a lot into what it means when organizations talk
1:19
about AI transformation, how they get AI ready, what it means
1:24
to put an organization into an AI first status, and how
1:30
processes and people are the means to a successful AI
1:35
transformation. Welcome Roman, I'm so glad to
1:41
have you. Hello, if I thanks for inviting
1:44
me, I'm very happy to spend the next couple of minutes or hour
1:48
with you. And before we dive into like
1:51
your new findings, you've just launched a huge report on AI and
1:56
the use cases and also the findings that you have and
2:00
guards of the implementation and so on.
2:02
I do want to start with you with a quite overarching question,
2:06
because what does trust mean to you in the context of AI
2:11
transformation? And is there a way or has it
2:15
also evolved over time on how you viewed it?
2:18
I think that's a pretty profound question for the first as a
2:22
beginning and prior some some reflection.
2:25
I personally experience trust when priorities are pretty
2:29
clear, so and companies or people see that they work
2:34
towards value, which is tangible and when individuals are
2:38
recognized and rewarded. And when in these transformation
2:42
types of work, everything is very much solution and not
2:44
probably oriented. So it has a positive spin.
2:47
Sometimes I see it as a risk for trust if things have to be too
2:52
fast and too bold. Sometimes people like also
2:55
reliability and steady progress and experience.
2:59
Yeah, resilience in activities and it's it's quite fun.
3:03
I mean we both know each other from the House of beautiful
3:05
business and we have attended great sessions by the famous
3:09
coach Michael Bunyas and he and he did a great research and when
3:14
situations, relationships are trustworthy and love lasting.
3:18
And I think resilience was a key password that took take away.
3:21
So when everything doesn't go right, people in an environment
3:24
of trust, they know that things progress and that resilience is
3:30
key in this this setting. So when, when, when I look into
3:33
what what has changed over the last years, I mean AI
3:36
transformation or digital transformation or IT
3:38
transformation before has always been organised large programs
3:44
and major investments. I think taking lots of risks,
3:48
but also promising big change. I see that nowadays things need
3:52
to be more Swifty like budgets are tighter, time is more
3:56
scarce. So if you want to build trust
3:58
with your seniors, you need to prove value very fast.
4:02
So you need to demonstrate impact.
4:04
So you can't just expect that that people wait for years until
4:08
you get an ROI of any transformation.
4:10
So I think trust in AI transformations nowadays is
4:14
really getting things done as talking more doing and proving
4:18
value very fast. And you've just described what
4:22
we could call a widening value gap, right?
4:26
Because there's on the one hand side, like the speed, on the
4:29
other side, like the steadiness that we want to have.
4:31
But it's both also in an economic and an ethical term.
4:36
And when you like also from your experience, look into today's AI
4:40
landscape, what is the single biggest driver of like this
4:44
specific gap, particularly in terms of technology leadership
4:49
or the mindset that is required? I mean, as you know, I mean I'm
4:52
work for BCGI, am driving AI and digital strategy topics and we
4:59
have, we are following a core principle which is called the
5:03
10:20 seventies principle. We always say that 10% of the
5:07
success comes from great algorithms.
5:09
Fine, 20% comes from outstanding technology, but 70% is truly
5:14
about people, progress and how you organize.
5:18
And in the center of the 70s is clearly the sea level alignment
5:22
and leadership fluency in AI and digital.
5:25
And you need to implement the culture that embraces change.
5:28
So success and short start at the very top.
5:32
I mean today technology matters more than than ever.
5:35
We have seen great companies creating great ideas, thriving
5:40
concepts, do rapid innovation, experiment, have great faith,
5:44
fast culture and they they build impressive prototypes and
5:48
customers love it and everything gets adopted.
5:51
But suddenly they don't scale and the question is why?
5:54
And often the tech stack was neglected.
5:57
So the data is siloed, end to end processes breakdown.
6:01
There's lots of menu interventions.
6:03
So as I say, everything needs to start from the top and from the
6:06
business. You need to protest by value,
6:09
but you also need to make sure that the foundation is, is, is
6:12
right. So the tech and data.
6:14
So if you have both, I think then then you make it or at
6:18
least the companies who don't have both, they see their scale
6:22
inhibited after a while. And you've just touched upon
6:26
some of the findings that you've like collaborated on with like
6:30
the report or also your clients and the role that you have at
6:33
BCG. But also there is like quite a
6:36
specific number that most of like us have heard about the 5%
6:41
of companies that generates a stem substantial value from AI
6:46
and like the rest actually is doomed or is set to stagnate.
6:52
And can you share with give us a bit more background of what
6:56
differentiates those 5% of organizations also in terms of
7:02
looking beyond just the resources, I was wondering if
7:05
it's like more about structure, courage or potentially also the
7:10
clarity of ambition towards their road map ahead?
7:14
Yeah, great question. And then to maybe take one step
7:18
back and give the audience some context, for a decade around
7:22
now, we are serving every year senior executive, senior leaders
7:26
covering all industries, all regions and serving them where
7:30
they stand in digital. And of course, nowadays with AI,
7:33
we check how they invest, we check how they succeed, we check
7:37
how they measure value, but also how they leverage tech and
7:39
technology and data. And in this year's study, which
7:43
is actually called dividing AI value Gap, we surveyed 1250
7:48
companies and you mentioned only 5% of them are really realizing
7:52
meaningful business value from AI.
7:54
So this is like I think it was 68 companies out of the 1250.
8:00
Of course, they invest more into AI, but the real differentiator
8:04
is that they strategize and act differently and they don't read
8:07
AI just as a technology project, but as a longer full business
8:12
transformation. They did operating models where
8:15
the data, algorithms and people work together seamlessly
8:18
integrated and they set the right ambitions.
8:21
They focus on those areas where I creates the most significant
8:25
advantage. And they also don't just focus
8:28
on small incremental productivity gains, like making
8:31
a big office function more efficient, but they really focus
8:34
on reshaping the core of the business.
8:37
That's a bigger task to crack, but much more valuably at the
8:40
end. And as I heard your question
8:43
about structure, courage and clarity, but I actually observe
8:46
is the clarity of ambition is by far the most specific factor.
8:50
I mean, everything follows from it.
8:52
I would even say that courage is you need to be gracious to also
8:56
have clarity because you, you can't just do minor small
8:59
investments or have some chase, some single digit efficiency
9:02
gains. You really need to be clear
9:05
where you put your, your place, what are your board
9:07
transformative moves and, and where you want to have double
9:11
digit growth and faster innovation.
9:13
So I think courage, clarity go goes, goes hand in hand.
9:17
And of course, if you have that, you need to build the right
9:20
structures to execute. And that's why companies
9:23
eventually need to build an AI first organization.
9:26
AI first is like a term that we've have heard many times in
9:30
the past, but there are also those ones who claim to be AI
9:34
ready. And I do think that putting AI
9:36
first doesn't mean that you're AI ready, right?
9:39
So what's the key shift required to move somewhat from, let's say
9:44
an experiment or mentioned kind of approach to an enterprise
9:48
wide transformation? Good point.
9:51
You are right. AI ready first means, I mean,
9:54
it's good when you're AI ready. So you already start to run
9:58
pilots. You, you, you upscale talent,
10:01
you maybe start to build a data platform.
10:03
But the pivot to shift to be AI first means that eventually your
10:07
entire business model depends on AI to operate.
10:11
So AI is not something you used to optimize, but it's really
10:14
core. And everything you do has an AI
10:17
first paradigm when you design it.
10:19
It's not about asking all where we can apply AI and our business
10:22
model to become a bit better, have incremental benefits, but
10:26
you should really think about how should our workflows,
10:29
decisions, the customer experience, our entire company
10:32
like be redesigned to succeed in an high power word that requires
10:36
fundamentally shifting how you work.
10:38
And it means also embedding AI into every process than just
10:43
laying it on top of it. And that's why you also
10:45
establish eventually trust in AI because grounding it in very
10:49
robust data, you implement these reliable algorithms and the
10:53
humans and AI seamlessly collaborate.
10:55
And then if when they are people eventually and employees and
10:59
also customers and everyone sees AI as a colleague and not just a
11:02
tool, then adoption accelerates. And what I learned from our
11:06
recent study is if we think about what really makes AI
11:10
organizations first, AI first is they have much more governance
11:14
in place to measure AI value. So, but because they can prove
11:18
its value, they dedicate significant more time in
11:21
learning and, and, and, and protecting also time for
11:24
upscaling employees. They have, I think the number
11:27
was three times more centrally defined data policies.
11:31
So they have data offices in place and eventually that helps
11:35
them to to scale five times more workflows and deploy them.
11:39
And the biggest, the by far biggest number was the
11:42
engagement of the C-Suite. And it brings us back to our
11:44
first question that everything starts from the top.
11:46
I think among the share of future build companies which we
11:50
use these or name these companies which are AI first,
11:53
over 10 times more of the sea levels in these organizations
11:56
are engaged versus companies who are lagging behind.
11:59
So it's again comes back to the point it needs to start from
12:02
from the very top. But at the same time, your study
12:05
also points out one of the buzzwords of like, let's say,
12:09
2025, and that is agentic or agentic services.
12:13
And it's claimed to be one of the biggest accelerators of also
12:18
the value gap that organizations can foster.
12:20
And in your opinion, what makes agentic AI different and how can
12:26
companies use it without falling into let's say complexity traps?
12:30
Also like the paralyzes that governance structure might bring
12:34
with it and potentially also translate it into more of an
12:37
intelligence choice architecture rather than make a process
12:42
driven approach. I mean, first, I mean, what is a
12:44
genetic AII mean? It is the systems that can now
12:48
act autonomously and have still clear goals defined.
12:52
And it, it completely reshapes how I functions because it's,
12:56
it's not just longer about automation or having the new
13:00
creative at it's about not AI that can really act and decide
13:04
and execute everything across different systems.
13:07
We have this analogy with the brain, the human brain, where
13:10
predictive AI is really the left brain.
13:12
It's logical rule based. It's very structured and
13:15
generative AI is, is the crazy one, creative, intuitive,
13:19
holistic thinking and agents now become some more kind of a
13:23
frontal cortex because it organizes, it prioritizes, it
13:26
acts. This is extremely powerful, but
13:29
of course, you can imagine it introduces quite a big new
13:31
complexity. So without what you meant a
13:34
strong choice architecture where it's very clear who is leading
13:38
when. So when do humans lead?
13:40
When is AI leading? And how do they intact?
13:42
When you don't follow such an architecture or think about it,
13:45
then organizations risk extreme chaos or human paralysis.
13:49
So the best companies they defined is upfront.
13:52
So they test agents and controlled workflows.
13:55
They ensure that decisions are transparent and they also know
13:58
that humans need to keep a supervisory role.
14:01
And also, yeah, make clear that ethics and and judgement are
14:05
essential. So for every case along your,
14:07
your life cycle and work workflow or use case, you need
14:11
to clearly define when it's fully AI autonomy of guard
14:14
rates, when human is in the loop or when it's just humans only.
14:18
And if you do this well, it becomes such a multiplier.
14:21
You have human creativity, you have great decision quality.
14:25
The potential is already visible in our research.
14:28
We also question like first, how many companies are experimenting
14:31
with agents already? And it's, it's almost 50%, but
14:34
just a fraction of those see value.
14:36
I think that's also in line with our overall study that few
14:39
companies see value, but the share of agentic value from the
14:43
total value business value of AI is supposed to double in the
14:47
next three years. So it's a, it's a big thing
14:50
coming up now you need to act with this responsibly, also
14:53
create trust. But then I think the impact will
14:56
be will be major and we will all experience this in the next
15:00
couple of years. And what I find so intriguing
15:02
about including genetic system services or like also like
15:08
particular structures into workflows is that it has and
15:13
requires A strategic view on the operating model because it's not
15:17
a tool. It's like how future built
15:20
organizations are defined. And in your expertise, how does
15:24
that what really definition of like the AI change and the
15:28
interference within organizations links to how
15:32
organizations think about trust when it comes to systems, teams
15:38
or in the end also the decisions that are taken or influenced and
15:44
for employees and like their specific roles.
15:47
Yeah, I mean, clearly as AI evolves from currently being
15:51
often set a tool to becoming your your business, I think
15:55
trust itself needs to be redefined.
15:58
I mean, previously and when you have traditional systems or you
16:01
had to implement compliance, you had to implement control its But
16:05
now in AI driven systems, I think what gives you trust is
16:08
also transparency, reliability, but also shared accountability.
16:12
True leaders would never ask you, oh, can we trust this
16:14
model? They would say, oh, have we
16:16
built a system that earns trust from employees, stakeholders,
16:21
regulators, everyone, because it controls AI itself and then it
16:25
becomes kind of systematic. So I mean decision making
16:28
authority. So how you make decision must
16:31
align with with risk exposure. I mean, when when there's a
16:34
likelihood that you have a massive irreversible catastrophe
16:38
of impact from, from a false decision on AI, then you can't
16:42
really fully replace human judgement with AI.
16:45
Nowadays, whatever you do with AI, the tech stack must be
16:48
highly secure against cyber threats.
16:50
Otherwise you have yeah, all these problems from infringement
16:53
and provency risk, etcetera. And now the more you scale these
16:57
systems and the more you build ecosystem, the ETA connectivity
17:02
increases exponentially. So everything becomes much more
17:04
vulnerable. So you need to even have more
17:07
secure operations. So training is important.
17:10
And I have also a very interesting example because I
17:13
mean, OK, trust, if you think about trust for AI, I mean, we
17:16
all use these tools and we get outputs and you initially try or
17:20
you initially learn to trust it. To which degree do you challenge
17:24
what you what you receive and do your challenges every time or
17:27
just sometimes? So so when do you think you need
17:28
to challenge an output? And when we did our study at one
17:31
point in time, I just put the entire data into our models.
17:35
And yeah, of course we have protected environment.
17:38
It was not public, but I simply wanted AI to give me hypothesis
17:42
to give me great insights. And I got 5 very stunning
17:47
out-of-the-box, non obvious ideas from all the data we had.
17:51
And I was really impressed. I mean, one of the fives turned
17:54
out to be like, like fully, fully off, like wrong
17:58
constraints and. But the other four were
18:00
brilliant. And then I wanted to say, OK,
18:01
let's be mindful, let's challenge it.
18:04
Let's ask the algorithm to give me the Python scripts.
18:07
Which data was used? Which assumptions did you take?
18:09
And it turned out after a while, the two of these four very
18:13
impressive insights were completely wrong.
18:15
Like, like the algorithm was cycling me.
18:18
I use variable epsilon, but they use variable X that there was a
18:22
mathematical flaw in a statistical model.
18:24
So so that that was very eye opening.
18:27
So you still need to do your own homework and create your own
18:31
judgement on what you see. I think that's core.
18:33
Also to implement it in the operating model to make it
18:36
trustworthy at the end. Your situation just shows how we
18:41
collaborate with agents or systems or also the
18:45
infrastructures that we've built.
18:46
And now we don't act like a loan and we want to integrate it into
18:52
the ways that we work and want to make sure that the infos that
18:55
we share are accurate. I do think that it comes a lot
18:58
to the kind of orchestration that we choose between humans,
19:02
data and like agents that we incorporate.
19:05
And if you now translate it into your like expertise and the
19:11
experience that you've made, what does good orchestration
19:14
look like? And what happens when
19:17
orchestrations fail in terms of connecting humans, data and
19:21
agents overall? I mean, I mean AI, as I said,
19:26
they genetic systems are super powerful and they but they, as I
19:30
described, can never really work in full isolation.
19:32
It's the same with people. People also don't work in
19:34
isolation. Yeah.
19:35
So the re value emerges from the entire orchestration of humans
19:39
data working as one team. And I think in in the well run
19:44
organizations, it's all about data at the end.
19:46
So, so agents use the data to act, to plan and humans set the
19:50
contracts at the context and boundaries and, and what they
19:54
want to achieve. Yeah, when this orchestration
19:56
fails, you experience the opposite like I, I experienced
19:59
with my example. You get either, I mean, with
20:01
agents, you probably lose accountability or you get
20:04
conflicting outputs, you get wrong decisions.
20:08
Maybe the work is being duplicated, which is even a big
20:10
thing. Like you ask agents to do
20:12
something and you never really when they go in cycles and then
20:15
suddenly you burn a lot of tokens and it becomes very
20:18
costly. So you need to also make sure
20:20
that this is under control. And that's why.
20:23
The MIT published once a great article about intelligence
20:28
choice architectures and here they made pretty clear that you
20:32
need to design to provide decision makers with meaningful
20:36
choices. So the agents or the algorithms,
20:38
they give you choices that they also give you or predict the
20:41
outcome of every of these choices.
20:43
What does it mean? And then you interact with them.
20:46
If you have humans in the loop and then you improve these
20:48
choices. And then you also see how these
20:50
choices interconnect with maybe other connections you don't even
20:54
know about. Because very often you have
20:56
these models and you say, hey, let's do XY that, but you don't
20:59
know how this interferes with other parts of the systems.
21:02
And eventually you have these great choices, which enables
21:04
humans to make much better decisions.
21:07
So these agents, in short, can never work alone.
21:11
So we we all become one system to a certain degree and need to
21:16
make sure that designing these architectures becomes kind of a
21:19
new management discipline. It's not just the CEO or the CFO
21:23
in the past where design is, we really need to design
21:26
intelligent governance architectures around agents.
21:29
Now and like digging into the report that you've published and
21:34
looking at the stats there, it's shown that only are like over 70
21:41
percent of companies actually report unmanaged AI security
21:46
risks exactly in those systems or potentially also in the way
21:49
that they integrate AI into their daily work whatsoever.
21:53
And what do you think? Like is this a trust issue or is
21:57
it a leadership issue? Is it like potentially also
22:00
responsibility issue that we have?
22:02
Or is it just a governance failure because people think
22:06
governance is boring or like costly complex to include?
22:09
It's interesting if if a leadership issue is always a
22:12
trust issue or vice versa. I think leadership is also
22:15
challenged at times. So sometimes if you lose trust
22:18
in something, it's not even a leadership to blame.
22:20
It's maybe the complexity that's that's having much faster
22:24
psychic times at the moment. I mean, yes, you're right.
22:26
More than 70% of companies have reported that their AI security
22:30
risks are not managed. It's very interesting that
22:32
almost 80% of companies, they use a third party AI tools, of
22:36
course, and not everyone buys or builds their LLMS themselves.
22:40
They, we all use third parties and many even know how many
22:44
tools they are using. That's also a problem with
22:46
shadow AI nowadays that, you know, in the past with shadow
22:49
IT, people buy their own tools. And now with AI business units
22:53
buy their own AI tools and, and automatively.
22:56
And there was a research and an experiment where we realized
23:00
that almost 50% / 50% of AI failures originate from these
23:05
third party tools. That's not surprising, but many
23:07
companies still don't, don't mitigate or manage these risks.
23:10
So some of your own risk management ends at your own
23:13
company, but you're not really considering third parties,
23:16
especially I and nowadays agents and multi agent architectures.
23:19
So you really open Pandora's box and then people lose confidence
23:23
in the system and then suddenly you have leaders who say we need
23:26
to be super compliant here and otherwise you, I don't know, get
23:30
a big issue. Yeah, you're personally reliable
23:33
if something goes wrong and then people don't use AI anymore, and
23:37
then what started with a technical issue or wrong risk
23:39
management becomes a completely trust issue.
23:41
People don't adopt. So it's why the more you scale,
23:46
the more you need to make sure that you design.
23:50
And that's also core in the book we wrote together a couple of
23:53
years or two, 2-3 years, Yeah, two years ago, two years ago.
23:56
That exactly time flies that you can't just design a process and
24:02
then at the end when it's done, think about, oh, what do we do
24:05
with security, privacy, ethics? That needs to be core in the
24:09
design when you start designing these things that at the end, if
24:13
you run them, this topic does not pop up.
24:15
So that's that's very important. So governance is not about
24:19
boring bureaucracy. I think it's the foundation when
24:22
well done and to create confidence and only then people
24:25
use it and you adopt and you innovate and at the end still
24:29
stays safe. And one of the key pillars that
24:31
you've also worked on was the Future Build Playbook.
24:35
And within it, there were like 5 highlights that I found like
24:40
particularly worthwhile to dive into because they are also
24:44
positioned as 5 strategic imperatives that we need to look
24:47
into. The multi year ambition, the
24:50
workflow reinvention and AI first model talent and data.
24:55
And if you had to pick one that leader is consistently like
24:59
literally consistency underestimate, which one would
25:02
it be and why? I think to give the answer
25:05
first, I think the multi year strategic ambition is super
25:09
important. However, everything here is
25:12
important because the five, because some of them become an
25:16
inhibitor later if you don't do this right.
25:18
And as I mentioned also before, the ambition is great, but you
25:21
also need to make sure not to forget technology and data
25:23
because otherwise later you find barriers to scale.
25:26
So everything met us. And if you want to become a
25:30
future built company, everything needs to be be in place.
25:33
Thinking about a playbook. A playbook is something which
25:36
also gives you steps. I mean, if you write a playbook,
25:39
it's boring. If you say, hey, please have
25:41
this thing and things in place, that's not that's maybe a
25:43
checklist, but it's not a playbook.
25:45
Yeah, you need to create a journey.
25:46
So when we also do these studies and know what are the key levels
25:52
to success, of course, we also pave or design A journey also.
25:57
And also depending on your starting point, where you are,
26:00
what is the priority of disrespective stage.
26:02
And definitely in the beginning, you need to have a multi year
26:05
strategic commitment from the top.
26:07
So it's not like her let's do this as a side project.
26:10
It needs to be a North Star because because again, if C
26:13
levels just say this is another tool, another layer on top on
26:16
what we actually do, then you can gain some efficiencies, but
26:19
you will never convert your company to be AI 1st and design
26:24
your entire business model around it.
26:26
But I mean, at the end, I simply see technology being back in,
26:31
in, in the very, very, very, very, very game again.
26:34
I mean, many years, you know, many years ago, like it was
26:37
always IT projects and you had to ask the business.
26:40
Then it turned around to be like, hey, this is a business
26:42
transformation, but please involve it.
26:44
And the best companies, they they combine this.
26:47
Yeah, they really have make this a business and tech
26:50
transformation. Likewise.
26:51
And at the moment, just to give one example, I mean, everyone is
26:55
now having these big SAP transformations.
26:57
Yeah, 2S for Hannah. And this is not a tech
27:01
technology update here. This is so significant because
27:04
you turn around every business unit, every process and
27:08
everything is challenged by its value.
27:10
And it's costly. You need to prove the value of
27:13
it. And at the end, it's also all
27:15
about redesigning your data. Tech is also a business
27:19
transformation, not only business transformation.
27:21
So take it like it's it goes hand in hand.
27:23
And the hand in hand, side by side collaboration is something
27:28
that you've also in like your work.
27:30
So the study said you've published emphasized quite a bit
27:34
because sharing this exact ownership and also combining it
27:39
to mature like as organizations and with the teams is quite hard
27:44
actually. And I also perceive this as one
27:47
of the hindrances because mutual understanding is super hard to
27:51
establish. And could you trade a bit more
27:54
on why that is and also how organizations can practically
27:59
build that shared ownership into their governance set up?
28:03
Yeah, yeah, pretty clear. I mean, shared ownership sounds
28:06
initially logical, but I mean we know the word of reporting
28:10
lines, the word of budgets, the word of politics, and also the
28:16
word of, I must say, key accountabilities.
28:18
I mean the tension. If we think a couple of years
28:21
back, business teams always wanted to have the IT solution
28:24
that best fits the individual problem.
28:26
So they give demands to ITIT builds it and eventually you
28:30
have a super heterogeneic architecture because you
28:32
basically serve 100 different demands.
28:34
So IT at one point was focused to ensure stability and security
28:39
and also scalability. And unfortunately, IT was also
28:42
pretty much squeezed like a like a lemon over years to save more
28:46
and more costs. Yeah, so some of them are really
28:48
burdened by years of of of of, of, of cost savings.
28:53
Yeah. So then IT suddenly got the huge
28:56
mandate. OK, we need to have a digital
28:58
transformation. Can you do this on top?
28:59
Wonderful. So on the one hand, with your
29:01
tiny budget, you need to be the innovator of your organization
29:05
and innovate rapidly. On the other hand, you need to
29:07
make sure that your architecture is not becoming spaghetti or how
29:11
we called it sometimes. So it wasn't easy.
29:13
And then you suddenly created these new roads like there's a
29:16
chief digital officer now and there was often tangents because
29:19
like both had their different targets, but it was not without
29:22
any conflict. Then it was clear everything has
29:25
to be a business transformation. We need to put a business person
29:28
in charge of everything. And this was great exciting
29:30
because suddenly you can you can also have great communication
29:34
how AI and digital turns around your entire business.
29:37
And then you try to experiment a lot, but suddenly you have
29:40
hundreds of great pilots. So we always call this thousand
29:43
flowers blooming. So we basically plant an entire
29:46
garden, but all the roses die at the end.
29:49
So and then there's the moment of truth where you need to
29:51
think, why is this the case? And then you realize that maybe
29:54
tech and digital and data must be part of this entire
29:59
transformation. Then it brings to clearly back
30:01
to what I'm saying in the beginning is that you need to be
30:03
called a sometimes two in the box.
30:05
So all these topics need to have a business and an IT digital
30:10
data counterpart. And if we, if I may share 1
30:13
figure from our report, we ask companies how they organize that
30:17
and the companies which are future bills.
30:19
So the top 5%, they have 50% more.
30:22
This organization is set up of having these two in the box
30:25
ownerships. And I don't mean like just
30:27
collaborating, I mean shared responsibility, shared targets,
30:31
shared accountability, shared budgets.
30:33
So, so really not not just governance on paper, but
30:36
governance established, which is a huge different.
30:38
Actually, I've I've over years of consulting, you can't imagine
30:41
how much governance I saw on paper.
30:43
But the reality was completely different.
30:45
So this time it's for real. Yeah, you need to anchor it to
30:48
make it happen. I would love to dive into that,
30:50
but I do think that then the time would fly even more than it
30:54
already does so. But for now, I do think one of
30:57
the aspects that I want to piggyback on is the
31:00
consciousness of the leaders that they sometimes have when
31:03
either it comes to the shared ownership or the established and
31:07
framework that is required to bring that to life.
31:10
Because there is this huge word that everyone kind of wants to
31:14
avoid because it's regulation. And we all say that it's hard
31:19
for regulation to catch up before it's scaling AI.
31:22
And from your perspectives, what actually are the costs of
31:26
waiting? And what kind of leadership
31:29
mindset do we need instead when we do want to progress?
31:33
I mean, on the one hand, I have this very rational mind.
31:36
I like innovation. I like rapid innovation, speed,
31:40
getting results fast. And on the other hand, I also
31:43
have a half a young daughter and I want to grow up in a in a very
31:46
safe and protected environment and avoid the robotic
31:50
apocalypse. Yeah.
31:51
So I think it's it's it's it's two ways how to look at this
31:54
topic. First, the rational point.
31:56
I mean, if you wait, then you can wait forever.
31:58
You slow down innovation and your competitor US may progress.
32:01
And you know, we also have global competition.
32:04
It's not just your competitor around the corner, it's a
32:06
competitor at the end of the world.
32:07
So you need to you need to also progress and also own your turf
32:11
and keep your competitive advantage.
32:13
The longer you wait, you also create more.
32:16
We call this tech debt. I mean, the big challenge for
32:18
companies nowadays is that they are still adopting these new
32:21
technologies, but they still keep their current tech
32:23
architectures. And this is very, very, very
32:26
costly. And when you wait too long, you
32:28
also don't benefit from learning cycle.
32:30
So at the end, those companies which, which shape guard rates
32:34
now, which tests responsibly, which also take personal
32:37
ownership for the topics and bid readiness, they, they, they jump
32:40
ahead on the curve. So I think you shouldn't wait
32:43
too long. You should bid the capabilities,
32:45
but do it in a responsible bid manner from the heart.
32:48
And the other perspective is, I mean, I hope and that's leaving
32:51
now. Maybe, maybe the, the current
32:53
discussion a bit what is best for humanity?
32:55
I mean, we are all humans now. We are not trans humans yet who
32:59
send their DNA to space. So we are still all together
33:02
here on this planet. And I think we should share a
33:05
common goal and be all aligned that AI should everyone serve as
33:10
a greater good. I mean, the tension is is
33:12
immense. I mean, think about tackling
33:14
climate change, think about curing diseases, education is
33:17
available, quality of life gets better.
33:20
So I recently thought about one word.
33:22
And I mean, no one wants to be remembered as the Oppenheimer of
33:25
AI who created then the beast that was unleashed.
33:28
And years later, you have the big regret.
33:30
So I think both, both perspectives fit together.
33:33
I'm a strong advocate for fast innovation, learning progress,
33:37
but it's important to define red zones.
33:39
So something which like, I mean, we, we were pretty clear with
33:42
GDPR for data and I think wasn't easy, but I think we also get
33:46
something good out of it. And I think we should do the
33:47
same with AI. And one thing which is important
33:50
to share, maybe as a last insight from the study
33:53
responsibly, AI is one topic we survey and assess in our
33:57
studies. And interestingly, in the
33:59
beginning, the topic didn't really pop up that much like you
34:03
should need to focus on other topics first to digitize fast
34:06
and bid AI. But those companies and we do
34:09
this crazy multivariate analysis with cause and models etcetera.
34:13
I spare you the details, but we have seen that from the third
34:17
stage to the fourth stage to become a future bid responsibly
34:21
AI was the most significant inhibitor to become future bid
34:26
if you haven't really designed it until the stage.
34:29
So you can neglect it for, but it will backfire at the end when
34:32
you really want to scale. And it was quite comforting for
34:35
me. So there's clear proven value in
34:37
having responsibility AI. Amazing.
34:40
I think we've also touched upon that in our like book that we've
34:44
published 10 moral questions and there's like the whole
34:47
background on the economical part as well.
34:50
So I'll link the link to the background information on the
34:53
study in the book. Also in the show notes.
34:55
Unfortunately, we already come to close with like today's
34:59
episode and also with our conversation.
35:02
But before we just say goodbye, I do want to share some moments
35:06
with you. They're so-called in between
35:08
moments. I use them with every guest as
35:10
quick, intuitive reflections. And I would love for you to
35:14
answer in like a word or a sentence, but pretty short.
35:17
And I want to start on what are how trust in AI feels like for
35:22
you. Sometimes I wish I could be AI
35:24
for a day. Yeah, it's a bit.
35:28
You know, it would be great if, if, if AI could feel like humans
35:33
feel and humans could feel like AI feel.
35:35
So I don't know if this is achievable, but that would give
35:38
me trust. And what does Europe need more
35:41
of? I think we we can have a bit
35:43
more self esteem and trust our own capabilities, but what we
35:47
really need is pragmatism. And what does AI leadership
35:51
require? Ah, you touched courage before.
35:54
I think courage is important, but courage defined with
35:57
integrity, not not just being radical, but being courage,
36:02
courageous, but in a mindful matter and having also some
36:06
patience I think. And what is the most overlooked
36:08
skill in transformation? I think people driven
36:13
innovation, yeah, I think enable people, make them curious, give
36:17
them cool tools and let them do really use people to innovate.
36:21
Enough that last one. What does future built mean?
36:25
Those future built companies, I have seen a very laser focus or
36:30
not truly matters, but they are also very bold and execution.
36:33
I sometimes envy them. I also sometimes wish I would be
36:36
more focused and what probably I am.
36:38
But you know, it's sometimes stunning.
36:40
It it's, I mean, think about Steve Jobs in the past.
36:43
He had always 100% clear knowledge what he wants to do
36:48
and what matters. That's that's that Champions
36:51
League. Yeah.
36:52
Indeed it is. And I want to close with one
36:55
question that I would love your like view on point of view or
37:01
maybe also outlook, because if trust is a missing currency
37:06
between the ambition that we have and the execution on like
37:10
the clarity and sophistication on implementation, what does it
37:14
mean in 2025 to be a trust builder in AI transformation?
37:20
I. Mean we wrote this book
37:21
together, yeah, about 10 more questions and to test the titles
37:24
like how to design take and AI responsibly.
37:28
And there are a couple of aspects which are really
37:31
important to establish trust in AI.
37:34
And I mean, as we said, I mean make sure it's beneficial to
37:37
everyone. Like it has to be a common
37:39
better good for your customers, stakeholders, your workforce.
37:42
I think that's important. And as we discussed, make sure
37:46
that it's it's humans and AI jointly being in this governance
37:50
models and strive to an intended purpose like give, give agents
37:54
freedom, but also make sure where freedoms end.
37:56
And the last pieces may be coming back to my my being in
37:59
favor of people and humans. Educate people, upscale people
38:03
and power them and use them to reshape their own workforce and
38:07
processes. So then adoption also gets
38:09
higher, etcetera. Yeah.
38:11
And one thing as we as we finish this interview, I had to laugh
38:14
because there's one favorite quote, which I also jumps and
38:19
making fun to my wife that can put on my tombstone.
38:22
Yeah, I hope there's still many years to go, but I like always
38:26
the word do the best and trust the process.
38:28
I think we should simply continue, do our best, act
38:31
responsibly, innovate and maybe be good people.
38:36
That's a big ending. It is.
38:41
It's quite a statement to to leave this conversation, but I
38:44
love it. Thanks so much for spending time
38:47
together and sharing also the background on the report.
38:50
I appreciate your time and like the strong finish and look
38:54
forward to seeing where that brings us.
38:56
But for now, thanks so much to man and we'll speak soon.
38:59
Yeah. Thank you.
39:00
If it was great. Thanks.