»An agent is not accountable. People are«

Bjarne Kock has been close to AI for much of his career. He specialised in AI during his computer science studies, where his work focused on recommendation algorithms and predictive models. Since joining Netcompany in 2015, he has kept one foot in the technical engine room while taking on more responsibility across projects and teams.

That combination has shaped his current role in Netcompany’s Agentic AI work. He researches new developments, follows industry trends, tests emerging tools, and helps translate the technology into practical use across the organisation. 

»My role has been both to go deep into the technical details and code, and to help give executive leadership an overview of what we are seeing in the industry,« he says. 

It is a role that requires speed, curiosity and caution in equal measure. Agentic AI is moving fast, but for Bjarne, the challenge is to make sure Netcompany moves with it without losing control. 

»If I look back at what I could do two years ago, a much cheaper model can do the same today in a fraction of the time. When you are close to it every day, the progress feels gradual. But when you look back, it is quite impressive how quickly things have changed.« 

From early interest
to everyday use

Bjarne’s interest in AI began with computer games and the question of how to build systems that could make decisions, react, and play alongside humans. Later, at university, he worked with machine learning and recommendation systems, exploring how existing data could be used to predict what might be relevant to a user. 

»Back then, I worked on what we’d call predictive machine learning«, he says. »You looked at existing data and tried to predict what would happen next. Generative AI is a different branch of the same field, and it’s what’s driven most of the recent leap « 

That early foundation gave him a way to understand the new wave of AI tools when they began to appear. Before Agentic AI became a formal focus area, he was already experimenting with it in his own work and spare time. 

At first, the experiments were practical and personal. Bjarne is an avid cyclist and runner, and he built small tools to connect training data from different apps. Over time, that grew into his own AI-based training coach. 

»I have made my own coach using Agentic AI. It can look at my training and sleep data and tell me whether I should do an 80-kilometre ride or take it easier with a zone 2 workout. I like cycling, running, and numbers, so it fits very well with my interests.« 

The same mindset now shapes his daily work. Bjarne uses agents to track research, summarise developments, and help him stay updated in a field where weeks can bring major changes. 

»I have agents in the cloud that scrape research sites and create podcasts for me, which I listen to in the morning. That is one way I try to keep up.« 

Turning speed into value

One of the clearest examples of Agentic AI’s potential came when Bjarne and a colleague were asked to build an application within a week. The task was to extract data and compare it with another format. Others had tried to solve it before without success. 

»We used Agentic AI to code it quickly, and after a week, it was good enough for the project to continue using it,« he says. »Without Agentic AI, it would have taken significantly longer.« 

For Bjarne, that is where the technology becomes genuinely interesting. Not because it writes code for the sake of writing code, but because it changes where people can spend their energy. 

»What I find exciting is solving problems and building things that people can actually use. With Agentic AI, that is accelerated.« 

He sees the future developer role shifting away from some of the repetitive, manual parts of implementation and towards review, validation, higher-level thinking, design, and responsibility. 

»Instead of sitting and moving things around in a file until everything fits, you can discuss the solution with your agent. You can ask it to challenge the design and help find the best way to build something. Your role becomes more about steering and designing the system.« 

Responsibility cannot
be automated

For Bjarne, the lesson is clear. Agentic AI can be powerful, but it should never be treated as a blank cheque for execution. 

»I am sure that somewhere right now, a vendor is putting a public sector project at risk by handing Agentic AI too much control and bypassing human oversight,« he says. »We are putting in the work to build a strict governance framework, so our use of Agentic AI is never that reckless.« 

Even with all the possibilities, Bjarne is clear that speed cannot come at the expense of quality, governance, or accountability. 

»An agent is not accountable. People are.« 

For Bjarne, this is what “Agentic with control” means in practice: using AI to accelerate delivery while keeping ownership, review and decision-making with people. 

That belief sits at the centre of Netcompany’s approach to Agentic AI. For Bjarne, responsible rollout means testing carefully, measuring impact, staying close to users, and making sure people understand the tools they are using. 

»You should not roll something out that you have not tried yourself,« he says. »If we want to use Agentic AI for testing, we first need to know whether it can actually write a test case. Then we look at how to make sure the people using it do not lose time because the tool is new.« 

Pilots play an important role. Bjarne explains how different users, both experienced and less experienced, are invited to test the tools. Their experience is measured, followed up on, and used to improve the next rollout. 

»We support people closely in the beginning, but the learning quickly becomes a two-way process. Some become superusers who help onboard the next group, and at the same time, we learn from the ways our colleagues work with and challenge the technology in practice. That is how knowledge spreads across the organisation.« 

Responsible use also means knowing when to stop, measure again and avoid scaling something before it has proven its value. 

»If something does not work, you should not be afraid to put it in a drawer and look at it again in three months. You have to be willing to kill your darling.« 

No black boxes

Transparency is another central concern. Bjarne is cautious about relying on systems that cannot be understood, inspected, or controlled. 

»We do not want to lock ourselves into something we cannot control,« he says. »People should understand how things fit together.« 

That is one reason he sees governance and methodology as even more important, not less, when Agentic AI enters the development process. 

»We still need our documentation and our guides. In many ways, AI amplifies the need to hold on to Netcompany’s methodology. The last 20 years have shown that our methodology creates value, and that does not disappear just because things can move faster.« 

The same applies to reviewing AI-generated work. Bjarne warns against what he calls rubber-stamping, where people simply accept what an agent produces because it looks polished. 

»You have to be critical of what comes out.« 

The risk, he explains, is that AI-generated output can appear convincing even when something is wrong. 

»If a five-year-old writes a novel, you can see from the spelling mistakes that something is off. But an agent can write beautiful code with the biggest error in the universe hidden inside it. That makes it harder to spot.« 

Optimistic, but not uncritical

Bjarne believes many tasks will be automated, but that responsibility, judgment, consulting, and control will remain essential. 

»There is still a lot of consultative work. There still needs to be control and accountability. People need to take responsibility for what is being built.« 

That is also what motivates him personally. His work gives decision-makers better insight into what is happening, what is possible, and what direction Netcompany should take. 

»What motivates me is helping the people who make decisions have the best possible foundation for those decisions. It is motivating to see that the work you do can influence the direction the company takes.« 

Moving forward, responsibly

The next step, Bjarne says, is to expand the use of Agentic AI across the organisation where it creates proven value, while keeping the rollout responsible, tested and grounded in Netcompany’s way of working. 

»We are not chasing this technology. We are deciding where it fits, testing it where it matters, and scaling it where it works. Twenty years of methodology is what lets us move this fast without losing control. That combination is rare, and it is where the real advantage sits.« 

For Bjarne, that is the challenge and the opportunity. Agentic AI can accelerate work, remove repetitive tasks, and help people solve problems faster. But only if it is used with judgement. 

»You should not rely on it blindly. You still need to understand what you are doing. You still need to be responsible for the result.« 

And that, to him, is where the real value lies: not in replacing people, but in helping them spend more time on the work that matters.