- Home
- Industries
- Public sector
AI agents for the public sector
Public bodies carry the AI Act’s widest deployer duties. They need agents that are explainable, logged and stoppable, and a supplier who can show it. We start with the administrative work where the gain is clear and the risk is low.
The problems we hear first
Citizen and business questions
The answers are in published rules and guidance, but people still call and write to ask.
Casework files
Officers assemble the same documents for every case before they can decide anything.
Tender evaluations
Every bid is checked against the mandatory requirements by hand, line by line.
Duties that come with AI
Impact assessments, registration and transparency apply to public bodies first, and the evidence has to be ready.
Each one prepares. A person decides.
Examples of what we build in this sector, not client work. For each one: where a person approves, and what it never does.
Citizen and business questions
Answers from published rules and guidance only, says it is an AI, and routes cases to the right desk.
- A person approves
- Officers handle any individual case.
- It never
- Decide on an application or a benefit.
Casework file preparation
Collects the documents a case needs, checks them against the checklist and lists the gaps.
- A person approves
- An officer decides the case.
- It never
- Evaluate eligibility for benefits or services.
Tender submission check
Checks each bid against the tender’s mandatory requirements and prepares the evaluation file.
- A person approves
- The evaluation committee evaluates and decides.
- It never
- Score the quality of a bid or exclude a bidder.
Correspondence drafting
Drafts replies from the case file and your templates, in plain language.
- A person approves
- An officer signs every letter.
- It never
- Send a letter on its own.
One job, step by step, with the person inside the flow.
Tender submission check
A public body receives nine bids for a services tender (sample). The agent checks each against the mandatory requirements; the committee evaluates.
- Sector
- Public sector
- Risk class
- Not listed in Annex III; reasons kept
- Who decides
- The evaluation committee
- Bids receivedNine bids, each with its forms, certificates and declarations (sample).Done
- Checked against the mandatory listEighteen mandatory requirements per bid, each with the page it was found on.Done
- Two bids incompleteOne missing insurance certificate; one signed by a person not on the register extract.Flagged
- Evaluation file preparedA table per bid: requirement, found or not, page. No scoring of quality.Drafted
- Waiting for the evaluation committeeThe committee decides on clarifications and every exclusion.Accept the fileAsk for clarificationDiscuss in committeeAI Act Art. 14Waiting for a person
- Clarification letters sentOnly after the committee decides.Locked
- Run recordedEvery check with its page, the flags and the committee’s decision.AI Act Art. 12On record
A sample, scripted on invented data: no model is called and nothing here is client work.
Where the law draws the line for you
Each rule with the day we read it. An engineer’s reading for your counsel to confirm, not legal advice. How we read the AI Act.
| Rule | What it says (our summary; quotes are verbatim) | What it means for an agent |
|---|---|---|
| AI Act Annex III 5(a)Read 12 November 2026 | Lists AI used by or for public authorities “to evaluate the eligibility of natural persons for essential public assistance benefits and services”. | We don’t build agents that decide eligibility. One that prepares such a decision is designed and documented as high-risk from the start. |
| AI Act Art. 27Read 12 November 2026 | Bodies governed by public law, and private entities providing public services, carry out a fundamental rights impact assessment before deploying a high-risk system. | We prepare the inputs for the assessment: the process, the people affected, the oversight and the complaint route. |
| AI Act Art. 26(8)Read 12 November 2026 | Public-authority deployers of high-risk systems comply with the Art. 49 registration duties, and don’t use a system that isn’t registered in the EU database. | We give you the registration facts and check the database entry before go-live. |
| NIS2 Annex I, Art. 21(2)(d)Read 12 November 2026 | Public administration entities of central governments and at regional level are in a sector of high criticality, and manage supply-chain security. | We answer your supplier security review layer by layer, including what is partial. |
| GDPR Art. 22Read 12 November 2026 | A person has the right not to be subject to “a decision based solely on automated processing, including profiling” with legal or similarly significant effects. | The agent prepares; a named person decides anything that affects a customer. |
| AI Act Art. 50(1)Read 12 November 2026 | Providers make sure people are told they are interacting with an AI system, unless that is obvious. Applies from 2 August 2026. | Customer-facing agents open with an AI notice in their first message. |
Nine layers around every agent, each with what is partial.
- Access and identityGoogle sign-in, revocable keys, an agent’s rights narrowed by its spec.Partial No MFA of our own and no enterprise SSO; open sign-up is on.
- Separation between clientsEvery record and search index filtered by the signed-in identity; a live agent gets its own containers and database.Partial Logical separation in one shared database, not yet tested independently.
- Hosting, network, encryptionHetzner, Nuremberg; HTTPS with HSTS; no inbound web ports; credentials sealed.Partial Stored data in general is not encrypted at rest.
- Only the data the task needsThe agent reaches only the systems its approved spec declares; secrets are scrubbed.Partial Personal data is not masked before the model.
- The model stepProvider set per deployment; each request admitted and recorded before it is sent.Partial The region is named in your contract, not promised EU-only.
- Untrusted content and testingDocuments are data, never instructions; a safety pack that must all pass.Partial No independent penetration test yet.
- A person decidesWrites wait for a person; a person approves every release; Stop and spend caps.Partial A live agent’s held actions are approved by our operator, not on your screen.
- Every run on recordEvery run recorded step by step, with approvals and who decided.Partial Append-only by convention, not tamper-evident; no client export screen.
- Deletion, export, backupsErasure with a receipt per step; a backup before every release.Partial Export and account deletion by hand, on request; backups not encrypted.
A timeline up front, a fixed price after one call.
Most teams start with a readiness review of the process they have in mind, or go straight to a pilot when the use case is clearly outside the high-risk list. We don’t publish prices: the proposal after a short scoping call gives a fixed price for your scope, in EUR, excluding VAT.
All engagements and what moves the priceBook a callPrice on request
AI Act and GDPR readiness for one use case.
What’s in itBook a callPrice on request
One agent on one process, tested and approved.
What’s in itQuestions from public sector
Do public bodies have extra duties under the AI Act?
Yes. Bodies governed by public law carry out a fundamental rights impact assessment before deploying a high-risk system (Art. 27), comply with the registration duties and don’t use an unregistered high-risk system (Art. 26(8)). Most administrative agents are not high-risk, but the classification and its reasons should be written down.
Will citizens know they are talking to an AI?
Yes. Every customer-facing agent says so in its first message (Art. 50(1)), and hands over to a person when the question is about an individual case.
Can you take part in a public tender?
Yes. We answer security and data questions with the same facts as on our trust page, layer by layer, including what is still partial.
Will the agent decide on benefits?
No. Annex III 5(a) lists AI that evaluates eligibility for public assistance benefits and services as high-risk. Our agents prepare files; an officer decides.
Related pages
Posts on these rules
Bring one process. Leave with a risk class, a pass bar and a plan.
A 45-minute scoping call with the people who would build it. You get a first view of the AI Act risk class, the controls it needs, and whether an agent is the right tool at all.