At every executive session, in every MAT, at almost every cabinet, the same question falls today: "What are we going to do with AI?"
The urgency is palpable. The Flemish government is investing tens of millions of euros a year in its AI Policy Plan. Microsoft Copilot is being rolled out to 10,000 civil servants. The AI Radar already counts more than 240 ongoing AI projects at Flemish and local administrations. AI is suddenly on the agenda everywhere.
AI is not a break from what we have been doing until now. It is the logical next stop on the digitisation train that the public sector has been on for years. The same journey that has taken us from paper forms to e-counters, from separate databases to links via MAGDA, from physical counters to digital services.
And as with any digitisation step, the technology can only be as good as the foundation on which it runs.
The absolute minimum: data quality
The extent to which AI can do something for you today is determined by what you have done with your data in the past few years. In our conversations with Flemish agencies and local governments, we see the same patterns every time. Files living in six different systems. Services each defining the same data in their own way. The same citizen known under the same information in three databases. A file called "handled" in one service and "closed" in another.
This is not a criticism. It is the reality of an industry built for decades on faithful service delivery, not data architecture. But it has a consequence: an AI application you put on such a foundation is going to disappoint. Not because the technology fails, but because it can only excel on the data that we give the technology. Rolling out ten thousand Copilot licenses to employees who are working with unreliable data mainly results in the same problem ten thousand times over.
So those who invest in better data are not investing in a prerequisite for AI alone. You are investing in an organisation that makes sharper decisions, shifts faster and reports more reliably. You achieve that return anyway whether an algorithm comes on top or not.
Process optimization with AI as an ally
Those who put their data in order inevitably discover that the same question arises for the processes around it. Because data do not come out of nowhere, they are created, transmitted and processed by people and systems working together every day.
An example from our work in a Belgian city. The city was struggling with a classic problem: their complaints process around GAS fines was bogged down under the increasing volume. The question they knocked on our door with was clearly can artificial intelligence provide a solution here?
After an initial analysis, the answer turned out to be more nuanced than expected. AI could certainly contribute, but only to a limited part of the problem. What the city needed above all was a procedure that ran smoothly from start to finish. That's what we worked on together. Incoming complaints and forms today are automatically read and sorted by type of objection (by AI). Then they are checked against the relevant sources, such as the accounting system, the database for people with disabilities, and the appropriate response is prepared on that basis. A staff member does the final check.
Of that complete solution, only that first sorting step is really artificial intelligence. All the rest is digitisation and smartly linking together systems that the city already had in place. And that's exactly where the biggest gains are: faster processing for citizens, less repetitive research for employees, and an approach that you can explain to a city council member or citizen without falling into technical jargon.
That's the right relationship between technology and organisation. Not: we buy an AI tool and adapt our operation. Well: we take an honest look at what could be done better, and deploy technology where it adds value.
The shop floor makes or breaks AI
Because most of this preparatory work is not about data or technology. It's about people.
About colleagues who dare to open up their own process. About departments that dare to have the difficult conversation about who does what. About managers who make room for reflection alongside the daily hustle and bustle. A tool that works perfectly technically but that your employees haven't helped shape quickly ends up in the drawer. Change takes time to get people on board, to take resistance seriously, to rethink processes together with the shop floor.
And it is also immediately where the real nature of an AI project becomes apparent: It is not a technology project that brings organisational change. It is an organisational change, enabled and accelerated by technology.
Where do you start?
You don't have to wait until everything is in place before getting started with AI. But you can start where the difference is really being made: with your data, your processes and your people.
The right order is simple, even if it is not always easy. First the honest mirror: what's going wrong today? Then get your data right and rethink your processes together with the people who work with them every day. And only then choose the technology that fits. Sometimes that's AI. Sometimes a well-designed process turns out to be at least as effective, and significantly cheaper.
Interested in a sparring session?
At Möbius we don't start from technology, but from your organisation. From the processes that get stuck, the data that don't make sense, the employees that need to be included, and the objectives that you as a local government or agency are trying to achieve. Only from that starting point does the question of what role AI can play in this come into play. We have walked that exact path with countless Flemish and federal public organisations in recent years. Sometimes it ends in a concrete AI application. Sometimes in a redesigned process without an algorithm. Does this question resonate with you? Then schedule a no-obligation conversation with us.
