Computerworld: When AI outpaces the organization

Many companies have adopted AI. However, studies show that this has not yet clearly translated into productivity gains. This is often referred to as the AI productivity paradox. There may be several explanations for this, and Malthe Kirkhoff Stougaard, Managing Director of Custom Solutions at Context&, has one of them.

“AI hit like a meteor. Not as a gradual change that organizations can adapt to over time, but as an impact that permanently altered the landscape. Right now, we are in the middle of the dust cloud. The organizations that win are not those waiting for the dust to settle - they are the ones that start building in the new landscape while it’s still swirling,” he says.

This is a realization many have not yet reached - the moment when AI goes from being interesting to fundamentally reshaping the strategic map of what an organization can do.

“Everyone will have that moment. The moment when you see what the technology can actually do - not in a demo, but in reality. And it’s not just for developers. It affects everyone.”

It is a moment many have already experienced at the individual level. The developer who sees an AI agent complete a task in 20 minutes instead of half a day. The sales manager who sees an entire customer segment analyzed in hours instead of weeks. The project manager who sees deliverables take shape at an almost impossible pace. And the executive who suddenly realizes this is not just another IT project, but a fundamental shift in what the organization can achieve.

“We stopped asking ‘how can AI make this faster?’ and started asking ‘what can we do now that we couldn’t before?’ It sounds like a small shift, but it changed everything,” says Malthe Kirkhoff Stougaard about Context&’s own journey.

That journey really accelerated in early 2025, when it became clear that AI works in practice and can be trusted.

“After the summer of 2025, we took the first major step in Custom Solutions with the ambition that now we build everything with AI agents. All employees get access to all tools so they can truly start experimenting, exploring, and figuring out what works and what doesn’t.”

This process can deliver more than efficiency gains in individual tasks. If AI is simply layered on top of existing processes, companies may achieve up to around a 20% productivity increase.

A study by Faros AI of 1,255 teams shows that employees using AI complete 21% more tasks individually, while the improvement at the team level was 0%. This illustrates that organizational impact does not happen automatically.

At the same time, DORA - one of the world’s largest studies on software delivery with over 39,000 respondents - indicates that AI adoption can be associated with a decline in delivery stability. Not necessarily because the technology is flawed, but because it enables more changes at a faster pace than organizations and processes can absorb.

Technology accelerates, but organizations do not necessarily move at the same speed. This is where the gap emerges - and why AI does not automatically create organizational transformation.

The real value of AI only comes when organizations dare to change their processes, emphasizes Malthe Kirkhoff Stougaard. Many current processes were developed under very different conditions - where knowledge was concentrated among specialists, analysis took time, and work required coding, approvals, and coordination between people. These conditions shaped workflows, roles, and responsibilities - and set natural limits on speed and capacity.

“This is where it really starts to get interesting. Once we map out the possibilities AI and AI agents offer, we can see that many of those limits disappear. We gain access to far more knowledge within processes and can pull in much more data from many more sources. That allows us to organize work in entirely new ways - and rethink the organization itself. Some processes almost collapse because they can be done completely differently than before,” he says.

This phase is not equally welcomed by all employees, as their roles begin to change. As technology takes over more sub-tasks, the human role shifts toward prioritization, quality assurance, and decision-making. This does not mean humans disappear - but it does mean the nature of work changes.

“The technology is ready. That’s the hardest realization. Because it means the bottleneck is us - our ability to change, our willingness to think differently, and our patience in bringing the organization along,” says Malthe Kirkhoff Stougaard.

He emphasizes that leadership carries responsibility for ensuring the transformation succeeds while maintaining employee engagement and motivation. He also encourages leaders to use AI gains offensively rather than treating AI as a cost-cutting initiative.

However, many companies still make the same mistakes. They send employees to prompting courses or provide access to tools and expect transformation to happen on its own.

“AI is not a training program. It is a transformation that starts at the top and runs through the entire organization,” he stresses.

Studies from McKinsey show that 62% of organizations are stuck in AI pilots that never scale. Malthe Kirkhoff Stougaard recognizes this - Context& was there too, with tools and enthusiasm but an organization that did not change fast enough.

Today, the transformation has come a long way. AI agents work alongside teams throughout delivery - from requirements and architecture to implementation and review. Employees’ work has shifted - less routine, more of what requires judgment, creativity, and understanding. The work that actually creates value.

But Malthe Kirkhoff Stougaard does not claim that Context& has “solved it.” AI integration is never finished. There is no end state. It is about continuous movement, learning, adaptation, and change.

“We move as fast as we can - but not faster than our people can keep up. That’s the balance. Technology has no speed limit, but organizations do,” he says.

Transforming an organization by changing processes and functions is not easy. It also creates uncertainty among employees. Not everyone wants things turned upside down, which means leadership must work hard to bring people along and create a sense of security.

The point is not that everyone must master everything from day one. The point is to set up the organization to learn. Malthe Kirkhoff Stougaard emphasizes that it is not possible to create a detailed multi-year AI transformation plan, because the technology evolves so rapidly. What companies can do is build feedback loops, knowledge sharing, and a culture where everyone experiments, fails, and learns together.

“We are not there yet. No one is. We are on a journey - and we will continue to be. That’s not a weakness - that’s the point,” he emphasizes.

This is increasingly becoming a key discussion in leadership teams: what does it take organizationally when AI agents become part of the work, and how does it change roles, responsibilities, and decision-making processes?

That is why Context& is bringing together a small group of leaders in early June for a roundtable discussion on what it takes organizationally when development departments move from traditional development toward increasingly building software in collaboration with AI agents and agentic processes.

Despite the excitement around AI’s potential, many still have a concern: how much can you trust AI and AI agents when the output looks convincing but may still contain errors?

“You should not blindly trust the output. Humans make mistakes - and so does AI. That’s why companies need to build processes with clear control steps and accountability, ensuring that AI not only produces faster but does so within frameworks that can be documented and quality assured,” says Malthe Kirkhoff Stougaard.

The technology is already here, and AI agents are entering many work processes. Companies can gain both speed and quality - but AI only creates value when the organization changes more than just its tools.

“The question is not whether your organization will have its ‘meteor moment’ - when it truly sees the potential of AI. The question is whether you are ready to act, and whether you can change fast enough to take advantage of it,” he says.

This article was originally published in Computerworld