AI in the public sector: How to get started safely and legally

As a player in the public sector, we know that you too are curious about AI. But many are at an early stage. How do you move forward without compromising on security, legal matters or the trust of the citizens?

How high is the AI maturity in the public sector today?

According to our survey, about 80% of organisations in the public sector either have an immature AI organisation or a basic AI maturity where initiatives are driven at department level without clear connection to the business strategy. Only 1.5% have reached a transformative level where AI is integrated into the strategy and drives operational and strategic decisions.

The willingness and curiosity exist. But there are also clear barriers that hold back progress.

What are the biggest barriers to AI in the public sector?

• Competence. Almost half see a lack of competence as the biggest obstacle. This makes it difficult to know where to start.

• Security and privacy. Uncertainty about how AI handles sensitive data creates hesitation. Questions about information security and personal privacy need to be answered before initiatives are scaled up.
• Data quality and technical debt. Poor data quality and difficulties integrating with older systems make it hard to get AI to work in practice.
• Governance and regulatory compliance. Uncertainty about risk assessment and regulations creates hurdles.
• Change management. AI projects without proper anchoring risk not being used, even when the technology is in place.

The goal is not high AI maturity for its own sake. AI is a tool to free up time, offer better service and make legally secure decisions. A means to fulfil the core mission of serving the citizens.

How can AI create value in municipalities and agencies?

There are several ways AI can do just that. Here are some examples:

Processing support that frees up time. AI can summarise lengthy case histories and present relevant legislation. Experienced case officers focus on complex assessments, while new staff enter the work faster.

Chatbots that provide round-the-clock service. Smart chatbots answer common questions about, for example, waste collection, building permits or opening hours. Increased accessibility for citizens, relief for customer service.

Decision support that ensures equality. AI can analyse historical decisions, identify deviations and help ensure that similar cases are assessed equally, which can lead to increased legal certainty and transparency.

The Three-Step Rocket: Get Started with AI Safely and Legally

Step 1: Choose the Right Problem and Risk Level

Classify your initiative according to the AI Regulation's risk model. Start with a low-risk problem, such as an internal efficiency project. You minimise legal risk but gain space to learn for future endeavours.

Step 2: Build on a Responsible Foundation

Privacy, transparency, and regulatory compliance must always be included. Therefore, you should have a plan for:

• Transparency. How do you explain to employees and citizens how the AI tool works?

• Fairness. How do you ensure the tool does not discriminate or reinforce injustices based on historical data?

• Data Governance. Is data used correctly and securely? Do you have the right to use it for the purpose?

Step 3: Choose the Right Technical Platform

There are three main routes here. What you should choose depends on your needs:

• Buy a ready service (SaaS). Fastest to get started, least control. Suitable for standard tools like chatbots.

• Build on a platform (PaaS). Flexible and customisable but requires more expertise. Suitable for tailored solutions.
• Run locally. Full control but most complex. Suitable for solutions with high risk and sensitive data.

Start Small, Think Big

AI in the public sector does not need to start with a big investment. Choose a defined problem, ensure you have governance and a responsible foundation in place, and learn along the way. The organisations that succeed best are those that start with a concrete need rather than the technology itself, and have management support to scale up what works. It does not need to be perfect from day one. But well thought-out, transparent, and anchored in the organisation, that is a must.