Skip to content
Article

How much change can your organization handle?

-By Andreas Rask Lundsgaard, Industry Sales Lead Government

 

AI can transform work far faster than organizations can adapt. So, what is the real bottleneck? 

Imagine a caseworker on a Tuesday morning. An AI system has drafted a decision. It looks reasonable. But she doesn’t know whether she can trust it. Her manager doesn’t know who is accountable if it’s wrong. And nobody has told her what her role is now: writing the decision or assessing it. So she writes it herself, the way she always has. 

The technology works, but the organization hasn’t kept up. When I talk to our public sector customers, I see that we spend too much time discussing what AI can do, and too little time considering how much change the organization can actually handle. 

Let me put it bluntly. If AI can change 30 percent of your organization’s work processes, but you only have the capacity to change 5 percent, what is the real bottleneck? The numbers are a thought experiment, not a measurement. But the answer changes the conversation. Buying more licenses or launching more pilots won’t help. The bottleneck is your capacity for change. That makes AI a leadership challenge rather than an IT challenge. 

I believe we need to approach the conversation about AI in the public sector differently if we want to get better at using it. At Twoday, we combine AI and digital engineering in our work with public sector organizations. And what determines the outcome isn’t technology itself, but everything around it. We deliver the solutions, but it’s your organization that has to change how the work gets done. 

More than a hunch  

Nordic studies point the same way. In Norway, well over half of public sector organizations say they use AI, but only one in four has reduced costs or resource use. The OECD finds that 32 of 36 countries offer AI training, but only 13 provide training on how to use AI in public services. And the National Audit Office of Finland notes that the benefits of AI do not materialize on their own. Realizing these benefits requires strong ownership and change management. 

None of these studies directly measures an organization's capacity for change. But they all highlight the need to focus on the point where technology ends and the organization has to take over. So how do we do that when it's so much easier to talk about systems than processes?


Andreas Rask 4 1920 x 1280

 

What change capacity really means 

Change capacity is not something you develop through a course.  

It is what is left once day-to-day operations are covered. It's the capacity that remains once day-to-day operations are taken care of. That's why employees need time to explore. They can't learn to use AI if nothing else is taken off their plate. New tasks are added to existing ones, and change becomes an additional burden rather than a redistribution of work. Meanwhile, employees are left to figure out AI on their own.  

Decision-making authority is just as important. Someone needs the authority to change a workflow, even when it affects several departments. AI Sweden, Sweden's national center for applied AI, calls this organizational room to act. Without it, AI becomes an assistant in an unchanged organization, and the benefits remain limited. 

Then there’s learning: someone follows up, shares what works, and is willing to stop what doesn't work. The caseworker from earlier needs all three: time to learn, a manager with the authority to clarify and change responsibilities, and an organization that learns from her experience of using AI in practice.  

This perspective on change capacity is based on my own experience from many years of working with public sector digitalization.

 

What would I do on Monday morning as a public sector leader? 

If I were part of a public sector leadership team, I would start by putting a number on it. How many workflows can we realistically change each year using AI without disrupting day-to-day operations? The answer doesn’t need to be exact, but we need to say it out loud so we can set priorities accordingly. 

Then I would focus on a few workflows rather than many. Specifically, those where the change is significant, and the benefits are visible to employees themselves. I would give employees time to adapt and take one task off their plate for every new one added. And I would assess managers based on whether the new workflows are actually being used, not simply whether the pilot was completed. If a pilot doesn't change how people work, it often delivers little value. 

Technology doesn’t pause while the organization catches up. Too often, that means we simply try to move faster, resulting in poor quality and frustrated employees. If we instead focus on how much change the organization can handle, and use that capacity wisely, I believe digitalization projects will deliver value sooner. 



Want to check the numbers? They come from the OECD’s Digital Government Outlook 2026, the Norwegian survey IT i praksis 2025, the National Audit Office of Finland’s report on AI in government (10/2026), and AI Sweden and Vinnova’s report Lärdomar från Kraftsamlingen (2025). All four can be found online.

You might also like

No related content