Rise of AI Conference 2026 · Berlin · Applied AI Stage

Responsible AI by Design: Embedding Governance into Systems

Christian Guttmann · EQT

At Rise of AI 2026 in Berlin, on the Applied AI Stage, Christian Guttmann argues that the governance models most organizations use for AI were built for an older, predictive-AI paradigm and are now too slow for agentic systems that make decisions in real time. Drawing on work across telecom, finance, healthcare and software, he says the recurring failure modes he observes when AI adoption stalls have nothing to do with the technology itself: misaligned incentives, organizational silos, and threats to the identity of experts whose judgment AI could replace. He illustrates this with a hospital scheduling project in Australia that saved money but was blocked because the hospital's budget rules punished underspending, and a retail bank where product managers were paid to push individual products, undermining an AI built to recommend across the whole portfolio. His central proposal is to move governance out of monthly committee reviews and bolt-on policy documents and build it directly into the AI platform, through evaluation frameworks, gateways, agent identities and observability, so that velocity and trust grow together as systems move toward multi-agent and eventually embodied and reasoning-based AI.

Key takeaways

  • Guttmann says the failure modes he sees in enterprise AI deployments are incentives, structure and identity, not the underlying technology.
  • An AI project that would have saved a hospital money was blocked because its budget rules punished underspending, not because the AI failed.
  • A bank had to change how product managers were paid before an AI that recommended across the whole portfolio could be adopted.
  • He expects multi-agent systems using A2A protocols within 12 to 24 months, followed by embodied and causal/reasoning AI.
  • He recommends leadership teams ask upfront who loses status, what gets rewarded that AI will break, and what to tell the board when it fails.

Detailed notes

Why current governance models are outdated

Guttmann says most AI governance in companies was built for a predictive-AI paradigm of periodic model reviews, which cannot keep pace with agentic systems that now make decisions, such as approving expenses, closing support tickets or placing trades, in the millions every day. Having worked across telecom (predictive churn, customer operations), finance (next-best-action for banking customers), healthcare (an IBM project in Australia optimizing hospital schedules) and software development, he says the mismatch between slow, monthly governance cadences and fast agent decision-making is consistent across industries.

Three failure modes that have nothing to do with the AI

He identifies three recurring reasons AI projects fail to be adopted, none related to the technology: incentives (KPIs still reward the pre-AI way of working), structure (optimizing individual silos instead of the whole business), and identity (professionals fear their judgment being replaced). In the Australian hospital case, an AI solution that would shorten waiting queues and save money was blocked because the hospital's budget structure would have punished it for underspending the following year. In a retail bank's next-best-action rollout, product managers paid on the volume of their own product (credit cards, loans) resisted an AI that recommended across the whole portfolio according to each customer's needs; the bank first had to change how the team was paid before the AI could work as intended.

From bolted-on to built-in governance

Guttmann describes most existing governance as 'bolted on': owned by risk and legal, expressed as policy documents engineers are meant to follow but often do not, reviewed in monthly committee meetings that slow deployment without changing system design. He advocates moving governance inside the platform itself, something he says his current company in Sweden, EQT, has recently released, so that it is owned by engineers and enforced through evaluations, gateways, and agent identities that track individual agent performance. He places companies into four categories along two axes of governance strength and shipping speed: most organizations today sit in the 'tactical' quadrant, shipping fast but reactive on governance, citing a McKinsey case of agents run loose causing a security incident, and, as a separate example he gives, an incident at another company where both the main and backup databases were deleted within about nine seconds. Only about 1% operate in the top-right quadrant where governance is part of the platform, enabling both speed and trust.

What is coming: multi-agent, embodied and causal AI

Looking 12 to 24 months ahead, Guttmann expects multi-agent systems in which individual agents, including ones belonging to different companies or vendors, communicate through MCP servers and A2A protocols and eventually negotiate autonomously toward new actions. Further out he expects embodied AI, physical robots already advancing in Asia and Japan, to require a different governance model, and causal or reasoning AI, since most current large language models only predict the next word and construct a reasoning explanation after the fact rather than reasoning as they act. He places quantum computing as a governance-changing factor more than ten years away.

Five questions for leadership rooms

Guttmann recommends leadership and CEO rooms ask five questions before deploying AI at scale: who in this room loses status if this works; what gets rewarded today that the AI is about to break; which silo did the AI just make invincible; where does the agent stop and the human start (with a concrete metric, such as a euro threshold, for when a human must be involved); and what will be told to the board the first time the system fails. He argues these questions should be asked up front, not after an incident, and that protecting experts' identity means giving them new responsibilities, such as judging agents and owning exception handling, rather than asking them to validate their own replacement.

Q&A highlights

Asked how to get honest answers in boardrooms about who loses status, Guttmann says concrete use cases, discussed openly and mapped to specific business-process changes, tend to produce honest responses; he describes a Swedish offsite where a team mapped AI's effect on different business lines and found the work would not disappear but shift, referencing another speaker's figure of 500% more work alongside more automation. Asked why Germany is slow on digitalization and AI adoption, he points to the smaller size and tighter networks of countries like Estonia and Sweden, a divergence he traces to the 1970s and 1980s when Germany, unlike the Nordic countries, resisted digitizing personal data, and to a cultural tendency to lead with risks and costs rather than upside. On a question about organizational versus technical controls when companies buy AI systems rather than build them, he says the goal is still to concentrate as much control as possible in a platform, referencing a point he says another speaker, Chris Boos, made about building a platform rather than a framework, and comparing this to how Google and Amazon operate their own platforms rather than making individual employees faster.

Quotes

In 10 years, no one will ask how good is your AI, but they will ask, can you still trust it.

Christian Guttmann

Who in this room loses status if this works?

Christian Guttmann

Most governance models that we are employing today are outdated.

Christian Guttmann

Questions answered in this talk

Why does AI adoption fail in large organizations even when the technology works?

According to Christian Guttmann at Rise of AI 2026, the recurring failure modes are misaligned incentives, organizational silos, and threats to the identity of experts, not problems with the AI or models themselves.

How should companies embed governance into AI systems?

According to Christian Guttmann at Rise of AI 2026, governance should be built into the platform itself, through evaluation frameworks, gateways, agent identities and observability, rather than handled through separate monthly committee reviews and policy documents.

What AI developments does Christian Guttmann expect in the next 12 to 24 months?

According to Christian Guttmann at Rise of AI 2026, he expects multi-agent systems in which agents from different companies communicate via MCP servers and A2A protocols, followed later by embodied AI (robots) and causal or reasoning-based AI.

Mentioned

AI governance, agentic AI, A2A protocol, MCP servers, next-best-action, embodied AI, causal/reasoning AI, evaluation frameworks

Keep exploring

More from Christian Guttmann

More from Rise of AI Conference 2026

More on Enterprise AI & Adoption

Transcript

Auto-generated transcript, not edited. · 5,676 words

Show the full transcript

That's it. Last [clears throat] session, lots of last energy. So, thanks everyone to be here. Great congratulations to Veronica and to Fabian Westerheide. I think it's a fantastic achievement, 10 years of Rise of AI. So, very happy to be here and I've been here for the last 7 8 years. So, really good community. Always happy to come to to Berlin from Sweden, where I live. And I think my talk will be a little bit about the front line. So, I'm >> [clears throat] >> um uh I've been working in industry now for many many years across the world and I think one tale of the front line experience is um that now there will be much more focus on having AI that actually works and doesn't break things.

Uh the governance models that you have and the governance structures are becoming very important. Um and it can become a competitive edge. So, it's really velocity and trust that will become a very important part of your of your AI that you're deploying in your company. >> [clears throat] >> So, most governance models that we are employing today are outdated. So, uh they're built for par- paradigm that's no longer exists. Well, predictive AI is one of those areas where you have uh meetings, regular meetings, and try to see if the use cases and the models and the AI you're building is actually applicable and works. And that's just not the time we are living in with having agents that make decisions much faster.

So, I've been exposed to many different industries. I work for several large corporations around the world and I see it across the board. Like, whether it's telecom, uh you have predictive churn or, you know, customer uh customer operations, you know, where you have use cases where you want to optimize where you're interacting with customers, or finance, you know, you look at next next best actions. So, meaning when you're logging into a bank, right? The way the information is presented to you is something that's actually meaningful to you. So, it's very personalized, right? That's another application. In health care, diagnostic and operational AI. IBM in Australia was working there, and we developed a system which [clears throat] was applied in one of the largest hospitals there as an example, and we were optimizing like schedules for for hospitals, or software itself, right?

So, when you look at coding and how you're embedding processes in the software development itself. But, regardless of the industry, regardless of the stacks you have always three failure modes that I have observed, and none of them have anything to do with the technology or the AI. So, um one area that I'm seeing is incentives, right? When I'm going into a business and work with a bank or pharmaceutical company, and we're applying AI, or looking at the AI strategy is that the system optimizes for the wrong thing. So, AI shows a better way but the way the KPIs are set are they are set for the old way. As if AI didn't exist, so prior to the fact, right?

So, the incentives are not not well structured. And then, of course, the structure itself. So, when you are having products for example, and you want to optimize for individual silos, that won't work, and you need to see the entire picture in order to really optimize across the the entire picture. And then, of course, identity. So, whether it's in [clears throat] health care, and investment professionals, or other other jobs, right? It's you need to protect the identity. So, people are very uh, to say, they they they love their job. And you need to see that you can uh, enhance their uh, their their job in some sense. You don't want to go in and say, "Hey, you know, your job will be replaced."

It's really about repointing uh, their attention and responsibility to things that become uh, similarly important. Like, for example, evaluating the agents that do certain tasks and work for them. So, none of the things I mentioned now have anything to do with the models or the AI, but they're often one of the main failure modes of AI not to actually work in companies. Uh, so, to double-click a little bit on that, so, some some experiences um, uh, [snorts] where you see the same pattern, right? I mentioned already that uh, we worked with a large hospital uh, in uh, Australia. And what we saw is we built a pristine solution in AI and optimization a solution, which would make the waiting queues in the hospital shorter.

It would save the hospital a lot of money. And uh, we had discussions, you know, with the C- CEO, with the CFO suggesting that the uh, solutions that we're having would actually save them money. But, the the structure that they had, the KPIs, were such that if they were to save money, it would punish them, right? So, many systems have a very how to say, you have a budget that you need to set. Many of these organizations, hospital authorities, they have a certain budget and if they don't use it, they will be punished, right? They will get less next year. Uh, it was one of those things that again nothing to do with the AI, which worked really well, but it basically prevented the adoption of this technology.

Another one is in financial uh, services, right? So, um, what we noticed there is that often [clears throat and snorts] you have product managers responsible for um, maximizing the revenue for individual products. So, in finance, a bank, they would have uh, credit cards, right? Like platinum, master, and then MasterCard, whatever it might be, or loans, right? And then you have individual managers responsible for driving the revenue in each of these categories. And [snorts] um that's the problem uh because if you have an AI that optimizes across the entire portfolio, every individual product manager will actually be unsatisfied, right? Because the actual AI will choose the product for individual customers as it sees appropriate. And the product managers responsible for individual products will be very unhappy because their their goal will be, "No, no, I want this AI to suggest my product all the time to all the customers."

Right? That that's a that's a huge mismatch here. And uh about the identity part, you know, when it comes to judgements, um when when judgement is the last thing to go, and that you see with lawyers, investment professionals, with healthcare experts, um you cannot go in and suggest that this judgement will be replaced, right? But you need to start looking at alternatives as to how you can make those individuals start re-evaluating the role in the organization. So, you need to protect their identity in some sense, right? So, you cannot go in and suggest that you would like to replace their um their primary activities, and that's actually very possible. But again, none of those things are uh have to do with the way how the technology works.

Um and then when it comes to the uh governance, most of what we are seeing today is like bolted on, right? So, meaning that you have governance uh systems where you built the AI in many of these scenarios that I mentioned before. And that then you are looking at having a monthly committee meeting, right? They look at the models, they look at the use cases, and uh review what has been shipped and whether it works or not, right? So, so that's often owned by risk and legal. It's a policy document that engineers are meant to be following, but they they don't really in in reality, right? And it's not really scalable as it slows the deployment and doesn't change the design.

So, built in, and that's what I'm advocating for, and it's also something we recently released in in my current company in Sweden, is that the governance lives inside of the platform. Meaning, when you have a platform that you're building for a company, you have the ability to build agents, you have have the ability to build AI which is compliant and safe, and the the governance is inside the platform. So, um Yes. So, and it's also owned by the engineers, and it can be enforced through evaluations, gateways. You have an ability to give individual agents identities, meaning that if they are performing well or not perform well, you actually see exactly what what happens. And then the system the platform becomes better and faster over time.

>> [cough and clears throat] >> Yes. So, where most companies sit today, you have basically four categories here. So, one That's no laser, but on the on the bottom part, you see these two dimensions, a stronger governance. So, you have an idea as to how the AI should work, what it should influence, and how the you know, how the AI actually influences the use cases and how it changes the business processes in the organization. Bottom left, [clears throat] basically you say you have governance, but you don't really have, and you also don't have a way of shipping the the tools, right? That's a That's a very bad place to be in, clearly. On the top left part, you have the strong words, but no no slopes.

In other words, you don't really have a platform that allows you to ship. Um The The top right is really where the 1% operate. So, the governance is a part of the platform. It allows you to scale fast, and it also allows you to build trust with the organization and with the main stakeholders in the company. We currently most of the company on this bottom right corner, which is like very tactical. So, you're shipping very fast, but you're very exposed. So, you don't really have You might have many agents or AIs in production, but the governance is very reactive. So, every incident is something that you need to review and can can can freeze the entire development.

And you have seen probably many of these breaches now recently in media. I think McKinsey had a big case uh you know, where they had a lot of agents run loose, and it caused a lot of security uh situation. Pocket iOS, I think yesterday had a big thing where the entire uh backup system was deleted and what like 9 seconds or something, right? So, there was both the uh the the the the database, the main database, and the backup database deleted. That's a tactical That's the These guys fall into the tactical bucket, right? So, [snorts] you want to make sure that you have something that scales and is speedy, but also builds the trust that you need.

So, um >> [clears throat] [snorts] >> as I said before, let's move the governance into the platform. What does that mean? And what have we built, you know, in in in uh at EQT? Ahem. It [clears throat] means that you have evaluation frameworks. So, you know how good your AI actually works and improve it every time you see how it works and it does something. You have gateways, um meaning you know exactly what type of data and what type of actions leave the framework and the agentic platform. So, you have an have a clear view as to what that means. You have an identity, or rather you have something where you approve the um the workflows, the agents before they they go out.

Um, but you only do that once. And then you [snorts] have observability, monitoring, and logging. So, you ensure that you can, if something goes wrong, have full visibility as to what happened. Which is something that also doesn't really happen if you have the situation where you built across the company, uh, different types of AI tools, but you don't really have the oversight as to how they are built or, you know, when when something goes wrong. Then the other thing, the second point, recut incentives before you recut, uh, head counts, right? There's a lot of discussions of, you know, even Meta and other companies letting people go. They might have a slightly different case, but before you consider these types of things in, let's say, a European bank or a European pharmaceutical company, I think you need to really think very carefully about how you're restructuring your incentives.

So, how do you actually give people in the room, um, in the in your different departments in the company, different type of, uh, incentives so that the new AI, which comes in with new structures, that that the, um, the main stakeholders are supporting this type of change in the organization. Um, so that [clears throat] that is a very important part and protect the experts whose identities at at stake. So, don't ask them to validate their own replacement, give them a new craft, judging the agents, shaping their constraint, owning the exception path, right? So, I think these types of things are, um, are very important. They have been working quite well if [clears throat] you do it correctly.

So, the governance that I have been seeing so far, which has been built maybe over the last 10 years in most organizations, is sitting on the left side. So, you might have predictive things, you know, risk scoring and and churn, right? Those types of things that you can see on a dashboard, along the lines of analytics. Now, we're living in this agentic, uh, in this agentic era, if you want to call it, which is like workflows and the agents can start taking some actions. Arguably, most are still in this sort of read stage, so connection to a lot of databases, not necessarily the ability to override databases or or, you know, send emails, but, you know, you go you go towards that direction.

>> [snorts] >> What I'm seeing in the next 12 to 24 months is multi-agent systems. So, companies will start being interested to have actually individual agents, maybe in different companies, vendors, you know, non-portfolio companies that can talk with each other, not just using Now, it's a bit of a technical term, but MCP servers and and, you know, access to data, but A2A protocols. Possibly even more sophisticated protocols for these agents to start negoti- negotiating with each other autonomously in order to arrive at new at new actions. And and then, of course, many more ways will come. I think we are not quite there yet. Embedded, which is when AIs have a physical body, such as robots, right?

We see a big big move now in in in Asia and Japan, where robots have the ability to work in our environment. And, you know, I see more of that happening, but it will again be a different type of governance model that will need to be applied there. And then, of course, causal. I think I saw another talk. There were different words for that, but what it really means is you need a reasoning model, right? Most of the LLMs we have today, they are really just predicting the next word word. And that's very powerful, but what you want to have or what is more powerful at at least is like a reasoning model. So, the AI has the ability to understand and reason its way forward.

So, it doesn't doesn't just make up its model its reasoning model on the go when you ask it, "How did you come to the to this type of conclusion that you shared with me?" Only in that moment, if you don't if you haven't noticed yet, but only in that moment it creates a model, right? Which it didn't have before when it made the first suggestion. And then of course eventually I do see quantum as a as a game changer, but I think this is more like 10 plus years in the future. I think when we are moving all the way down, the governance models will change and I think we can also start talk looking at AI tools which will be having much bigger impact to the in the ecosystem.

So, governance currently is organized still in most organizations by having sort of a review on a monthly basis, you know, often legal, often the engineering teams come together, review the models, review the agents, review the cases, review what has gone wrong or not. And [snorts] that's sort of cadence is out of date because if you see at the bottom part the agent decisions that are made, that's like in the millions every day, right? So, agents make decisions on, you know, approvals of expenses, closing a support ticket, placing a trade, reorder a part, escalate risks, all that stuff happens very very fast with agentic systems. So, there's really a mismatch between the two and the point here is that the governance models that I see in organizations, they are outdated and too slow and they need to be baked into the into the agent platform and the governance system itself so that you can have velocity and trust in the system.

So, >> [clears throat and cough] >> suggestion platform as a product, think of it this way. You have gateways, evaluation frameworks, improving the models on the go, observability, identity of the agents, and the platform will be the moat, right? When if you're building that, you will have a very good starting point in the company to build something that's both compliant, scalable, and builds trust and velocity, right? And [snorts] of course, together with making sure that the humans in the loop are, let's say, um positioned well, uh included in this process, you need to uh make sure you understand what those humans, the experts, can actually do in a very meaningful way in order to sort of reinforce that cycle.

Um Yes, again, interesting enough, very little of what I'm presenting here is actually AI algorithms per se. It's really decisions that leadership needs to make and be aware of. So, um Yes, I wanted to double click on a couple of things here. Uh when the AI worked, so we had a retail bank and it took a long time for next next best action rollout. So, I don't know how you're familiar with the term, but it means like, okay, you have a customer and you want to improve the likelihood for that customer to um do a buy of some sort, so they increase the conversion rate. And um And uh basically, the marketing managers were paid by the volumes of their own product, as I alluded to before, right?

So, you have, let's say, five products. You tell each of them, okay, you sell a lot of Mastercards, another one has a loan preparation, and so on and so on. And the AI that we have been building would would be built in a way that it would optimize it according to the individual that would be logging into the bank's website and present the the product that has a highest likelihood to be sold, right? So, it would be very much against the very much against the incentives of the individuals being paid. So, what we did then is uh we set everyone in a room and we asked question, what if you changed how the team was paid before we change anything as to how the AI worked.

Right? That's an important distinction. It would have been in fact what we have done earlier is we tried [clears throat] to change the AI, but it becomes close to mix it makes no point there's no point of having AI because you start just pushing out the same message and AI would have made no difference in the system. So, we really needed to enforce that. We have been managing to change that which then meant that the that the the leader of that entire section of course had a had a much much higher revenue. So, um So, that's something we have suc- successfully done. Another thing I want to point out here when it comes to the questions that I am asking and I would ask in strategy rooms especially in in CEO rooms of most companies it's those those five questions, you know?

>> [snorts] >> Which is the first one about identity. So, who in this room loses status if this works? So, you have AI. Who in this room and this is an honest discussion you have to have in many leadership rooms really, right? Who loses status? And not something you need to need to wait until it's too late or things are failing, right? Ask them up front and early on. Second question, what gets rewarded today that the AI is about to break? Um third one structure, which silo did the AI just make invincible? All right? So, in other words, if you are employing AI and processes change, there will be a whole lot of tasks which these teams or these individuals will not do anymore.

So, which ones are they, right? Where does the agent stop and the human start? So, here you I I would not leave the room until I see a metrics and see exactly that for example, if the costs of something, right? Is exceeding a certain amount, 10 million euro. That's when a human needs to be involved or not, but you need some type of metrics. And that's also not something that you can um uh wait doing when you have the system already in place. And importantly, and things [clears throat] will go wrong, right? Uh what do we tell the board the first time it fails, right? Prepare that up front and not 9:00 p.m. on a Friday night.

I have uh been many times awake at 2:00 a.m. because things things break. They will break, right? And we had so we had excellent engineers and and everything we could possibly do, but that's um these are some of the questions that are very important to ask, in my opinion. Good. And you know, I've deliberately tried to keep it a little bit shorter because I know we are at the end of this day, >> [clears throat] >> and I finish with this last slide, which also means we have more question the time for questions. Yeah. Last last point I want to make is perhaps you know on uh it is really about now building something where velocity and trust is most important.

In 10 [clears throat] years, no one will ask, you know, how good is your AI, but they will ask, can you still trust it? Hey, voila. >> >> Thank you. >> How nice that you mixed empathy with your tactics. >> Mhm. >> Very empathetic to give us some time for questions. >> Does anybody have a question? I have questions otherwise. You maybe you want to think about it for a minute. Um so, I build strategies for a living, but for communications, and those are very difficult questions. We do this today with like stakeholder mapping, for example, and then AI is one more stakeholder, maybe. But how do you guide companies and boardrooms through these questions so they come to like honest answers?

Is it so that they know you want to fix it and they don't lose status, or do you have these conversations without the people affected or how do you get to the bottom of this? >> When you are in the boardrooms, many that I have been in, or not just the boardrooms, but eventually [clears throat] the executive leadership team, right? So, you can you can and have to have honest discussions about it. I think the difference is that if you have these discussions up front and you have concrete cases, right? You you look at specific use cases and you lay out how the business model and the processes change, you usually get relatively honest answers.

So, it's people are protecting to some extent, but um >> [clears throat] >> you you you just you have to discuss it. People know, you know, the political situation when you are in these types of meetings. So, you get answers. >> I I'm very curious about this because you would have to have these discussions today even with other transformation processes, right? Like who loses status if this works and a lot of the time people have so much of their own agenda that they >> Yes. Yes. >> So, how do you solve this? Do you have an example of, you know, where >> [clears throat] >> where this was solved? >> Look, there were there are many ways.

I think you need to read the room also, but I mean I have one example where I think the the team was keen. They were very keen to adopt AI. We had an offsite, you know, which was a day, I think it was in the Swedish archipelago like archipelago and um >> [clears throat] >> we went through the use cases. We had an assessment and a mapping as to what it means for different business lines. And then we were also suggesting openly, you know, what it what difference it would make in the business. And of course also look at alternatives, right? So, I think maybe it was that was good circumstances because it doesn't mean that competent people would have to leave the organization, but it is about restructuring and refocusing what other responsibilities will come up in this process because as I think I saw some other presentations earlier, right?

You have 10 times more uh what was it? 60% more robots being introduced and 500% more work, right? >> [clears throat] >> So, it's a bit like that. The work won't go away and you will there will be plenty of glory, which becomes often evident in the process when you start looking at it, right? And what type of responsibilities you have. So, I think you need to have a you know, going deep, looking at some use cases and map that specifically to the processes and to the changes and of course what the new responsibilities are that come out of it. So, [clears throat] >> So, if part of the discussion is what you also wrote down, which I also thought was very refreshing to just protect the identity of the people who could be replaced and give them new alleyways to like find their purpose.

That's part of the discussion and then maybe makes it easier. >> I think so. And my my my [clears throat] I'm not thinking that you really need to make such a you know, need to think about moving people out of the organization. I think it's all about making people faster and better with these tools. So, >> Do you not have all of this [clears throat] insight onto the question, why is Germany so awfully slow in digitalization, let alone introducing AI? >> That's a mic drop question, wasn't it? >> >> Also, first of all, I I I live in Sweden, so I have a bit of I can probably I don't know, should I be more careful or less careful, you know, answering this question?

Less careful. Yeah, look, I mean, this is a very long answer in some sense, right? So, I think so, so, why is why is Germany so slow in digitalization? >> [clears throat] >> Yeah, okay. So, one one reason is this, Estonia is much smaller, right? So, I mean, when I when I talk to the president of Estonia, very impressive country, right? But the smaller ships in Europe, the smaller countries can act faster. Sweden similarly, what are we? Nine nine million nine million Swedes? Uh you know, when Ulf, the Prime Minister and others, when we talk, it's much more accessible. We talk the network is tighter, right? So, it's definitely the it's definitely the size is one of the thing things.

Um >> [clears throat] [snorts] >> It was said in many other talks. I'm just wanting to generate something that the other speakers haven't already said now. Uh I would rather like to swi- switch it around. You know, what is it we are doing good? Right? I mean, we still we are very much an engineering nation, right? That's very similar to the Swedes. How can we build something which, you know, comes close to what we can do well and what we enjoy? I think that's important. Um I have given before I gave an interview to journalists, and I think I was saying, "When did it start?" Right? She asked me, "When did it start?" As to why have you fallen behind and why have you become slow?

I think it happened already in 1970s, 1980s, right? When in in Germany, uh in particular, uh you know, after the war, uh there wasn't very much enthusiasm in digitizing, for example, personal data, right? Which, by the way, happened in the Nordic countries very very much. And one of the reasons is people were afraid, right? So, if the government comes in that takes personal data and have everything digitized, not very good. That was the time when Microsoft and and uh Hewlett-Packard in the US had their garage garage businesses, right? Where Bill Gates was building MS-DOS and so on, and we were very slow uh in that process. So, we already then diverged, right? So, I would hope that we have a Curiosity is super important, right?

Re- getting that back, being open about change, and not I notice this a lot in in in Germany is talking about the risks and the costs and the problems. But, it's almost like cognitive behavioral therapy. Like, can you add to such a sentence always what the upside is, right? So, because that would be the answer then partially, right? You think about the problems. And it's part of the culture of the German culture that you like something and by expressing that you like it, you know, right? You you you try to criticize it. That's your That's a German expression of I like I like your idea. I simplified a little bit, but it's a bit like that, right?

>> Looking around the room, people can relate to that statement. We have one more quick question before we wrap up. >> Um yeah, thanks for the presentation. I really like the slides. Um just one question. Um it sounded a bit like you dislike the organizational controls, but you want everything to be embedded in a technical way in the product, but lots of the solutions are adopted um where the company doesn't really it's not controlling the product. It's buying it, but it's not developing it itself. So, it's a bit hard to make technical controls, so organizational controls are more the way forward. Do you have like an opinion or you want to specify um what you think about technical and organizational controls in embedding governance into those systems?

>> So, let's see if I understood your question correctly. >> [clears throat] >> Uh so, it's organizational and technical control and them having having those in the platform. Or >> Well, in my experience, lots of the times you need organizational controls because you cannot build technical controls into a system that you didn't develop yourself. And lots of the systems adopted are not developed by some company itself, but they are bought and they're like software as a service or something. >> Ah, okay. Yeah, okay. [snorts] That's a huge question that we are facing a lot like I guess it's this buy and build question, partially, right? Like do you build something or get a platform from somewhere else or how much >> I guess it's the differentiation also between uh AI in research and development and AI just in enterprise solutions for automation.

>> Mhm. So, first of all, everything I talked about is AI in enterprise, right? So, this is this is not R&D. This is like real real stuff that needs to work. Um I don't know if it answers your question, but I think the main point, and maybe I'm reiterating it now, is that you would want to have as much as possible in a platform. And this goes to a point another speaker made, and I totally underline that. I think it was Chris Booz saying that it's a platform. The Americans think built a platform, built built a framework. So, if you are, for example, at at Google or Facebook, they are not supporting individuals to become better and faster.

They built an operator system, such as, you know, the Google search and the the ads that they have, >> [clears throat] >> and they make that faster. Same with Amazon, right? So, it's like you would like to have a platform that you make better over time, and you want to have as much um, organizational knowledge, uh, and you want to use it as a flywheel, so that you can build and and scale much faster. And you want to have all those types of things in that platform. That that would be Uh, but we can have we can double-click on that after the talk. >> after you. >> Yeah, I think so. I think a round of applause is due as I thank you as well.

>> such

Rise of AI 2027 · Berlin · May 2027

Be in the room for the next one.

Ten years of talks started as conversations in one curated room in Berlin. The next edition brings 300 selected leaders together in May 2027.