All visitors

AI for the public sector

Public organisations face a specific challenge: AI must be not only useful but publicly accountable. Every output needs to be documented, traceable, and explainable to citizens, auditors, and oversight bodies. I help teams get there.

In action
What this looks like for your team

A benefits eligibility case: one of the most common and consequential decisions in the public sector, where AI can do the regulatory analysis and a human makes the call.

Missed prenatal appointment: benefit eligibility review
1
Input
A citizen file: a pregnant woman has missed a required prenatal medical appointment. A case review is triggered to determine whether she loses the associated benefit under the applicable regulation.
2
Agent
Retrieves the applicable regulation, reads the citizen file, identifies whether the absence qualifies as justified under the relevant criteria, and drafts a cited recommendation with every finding linked to the specific article and condition it relies on.
3
Human checkpoint
The case officer reviews the recommendation and all cited articles before any benefit decision is made or communicated to the citizen. The officer's decision is the official act. The AI produces the analysis.
4
Output
A documented case recommendation with full audit trail. Every finding is traceable to its regulatory source. The decision is defensible to the citizen, an appeals body, or an external auditor.
Training programmes
Three courses, one method

All courses use nothing more than a standard chatbot. All include a certificate of completion and cover regulatory AI literacy obligations.

Flagship course

Build it yourself

5 hours instructor-led · 4 hours self-study

Participants work through the full four-pillar method on a public sector workflow they will recognise. By the end, each participant has built a working, auditable AI agent and leaves with it deployed. Certificate issued per participant.

  • ✓ Adapted to public sector processes and accountability requirements
  • ✓ Simulated data, no citizen data used
  • ✓ No prior AI experience required
Workshop

Build your own problem

3 hours face-to-face

Each participant brings a real administrative task from their own work. The session is a supervised build: they identify the workflow, apply the four-pillar method, build the agent, and demonstrate it to the group.

  • ✓ Works on the participant's own real process
  • ✓ Any workflow, any department
  • ✓ Recommended as a follow-on to the flagship course
AI for managers

Understand without building

5 hours face-to-face

For heads of department, directors, and senior decision-makers. Covers what each type of AI tool can and cannot do, what to commission, what to scrutinise, and how to maintain public accountability and auditability.

  • ✓ Designed for decision-makers, not builders
  • ✓ Covers governance, oversight, and public accountability
  • ✓ Certificate of completion included
How I can help
Two ways to work together

Training gives your teams the skills to build and supervise AI tools themselves. The transition engagement goes further: I come in, build it with your team, and leave them running it independently.

Training

Your team builds it themselves

Three courses available: a flagship where participants build the public sector example shown above, a workshop where they build a tool for their own daily process, and a management course for decision-makers who need to understand AI without building it. See the courses section above for full details.

  • All courses use a standard chatbot, no engineering required
  • Certificate of completion per participant
  • Covers regulatory AI literacy obligations
Transition

I come in and build it with you

A defined engagement where I embed with your team, identify the workflows where AI can have the most impact, build and deploy the tools alongside your people, and leave them fully autonomous.

  • Scoped to a defined period and set of workflows
  • Built with your team, not for them
  • All outputs documented and explainable
  • Governance and oversight built in from the start
  • Starts with a free 30-minute intro call
Credentials
Why this works

I bring 25 years of experience in software verification and AI, including work with CEA France on formally verified operating systems for high-stakes environments. The same rigour that makes software safe is what makes AI outputs trustworthy and accountable.

I have delivered AI training at the House of Training and Digital Learning Hub in Luxembourg, and held academic positions at the American University of Paris, University of Luxembourg, and École Polytechnique.

Selected media
Uni Luxembourg AI legal assistant for patent law launched at the University of Luxembourg science.lu Comment l'intelligence artificielle transforme le système juridique Research Luxembourg In conversation with our sciencepreneurs: supporting the legal profession with AI → All media and speaking
9.13/10
learner satisfaction across 50+ courses
25 yrs
in software verification and AI

Start with a 30-minute call

No form, no pitch. From first call to proposal in three steps, and the first two cost you nothing.

01
Intro call
30 min · remote · free
You describe your situation. I tell you honestly whether I can help, and what I would look at first.
02
Working session
1 hour · in person or remote · free
We go deep on one concrete workflow. You leave with a specific recommendation, whether or not we continue.
03
Proposal
Only if the session showed value
A defined scope, timeline, and price. No obligation at any earlier step.
Book a 30-minute call

Or write directly: tomer@libal.info