AI in public offices and public administration

Assistant for letters and correspondence

Process automation

Problem: Preparing replies to letters and applications takes too long because the case file is scattered across the EZD system

How it works: The assistant connects to EZD RP / EZD PUW and e-Doręczenia (e-Delivery), reads the case file and prepares a draft reply with sources indicated. The civil servant checks, corrects and signs

What you gain: A draft letter is ready in minutes instead of hours, and every sentence points to its source in the case file — it is easier to meet administrative procedure (KPA) deadlines without hiring more staff

Verifiable budget and reporting analyses

Analytics and reporting

Problem: Budget analyses and reports need results that the treasurer can reproduce and check

How it works: The model understands the treasurer’s question, but the calculation is done by deterministic Python/SQL code on reporting data — every result comes back with its source, formula and limitations

What you gain: Comparisons with peer local governments, long-term financial forecast (WPF) analysis and cost-pressure scenarios in a natural-language conversation — with results the treasurer can repeat and present to the council

Procedure assistant and faster onboarding for civil servants

Adoption and skills

Problem: New civil servants do not know which procedures apply, and some are not written down anywhere

How it works: The Knowledge Gap Register points to missing or outdated procedures instead of guessing, and the knowledge audit (CDF-F0.K) puts them in order before the system goes live

What you gain: New employees get answers only from procedures in force, and management gets a list of procedures nobody has written down

Safe AI work with citizens’ data

Security and sovereignty

Problem: Citizens’ data must not reach an external AI provider

How it works: PESEL, NIP, REGON and other identifiers are detected and masked before a query is sent, and an “EU only” switch blocks routing outside the EU for the whole office

What you gain: Staff can use the best models, and the DPO has a technical — not contractual — guarantee that personal data will not leave the office

A secure AI environment instead of public chatbots

Security and sovereignty

Problem: Staff use public ChatGPT to work with the office’s documents

How it works: A single gateway controls the use of models and tools and records activity, and a scan for unauthorised AI tools detects them and offers a path to legalise them

What you gain: The office knows who uses AI and for what; staff get a legitimate tool that is better than the one they used in secret

Compliance of high-risk AI systems

Compliance and oversight

Problem: A public body must register a high-risk AI system and demonstrate an impact assessment

How it works: The methodology guides you through the impact assessment (ASIA/FRIA) and conformity assessment, and the system exports material for audit and registration in the EU database (Article 49 of the AI Act)

What you gain: Registration and impact assessment carried out along a ready path, with documentation you can show an inspection instead of creating it after the fact

Can civil servants use publicly available AI tools?

Only if the office has approved them: it has a contract with the provider, knows where the data goes and has trained its staff. A public chatbot on a private account does not meet these conditions, and the office is responsible for the data entered. It is safer to give civil servants a tool that runs in the office’s infrastructure or in a controlled cloud in Poland or the EU

How can you use AI without breaching GDPR?

Personal data may go only to tools covered by a data processing agreement and control over data location. PROXY:AI masks personal data before a query reaches the model and records every query in a log. TWIN:DESK can run entirely in the office’s infrastructure

Who is responsible for errors in content prepared with AI?

The civil servant who used it, and the office as the deployer of the AI system. That is why every TWIN:DESK answer includes a citation from the source, and the AI use policy sets out who checks the output before a letter is sent

Where should a public office start?

With an inventory of tools, an AI use policy and training for civil servants. Then a pilot on one process, e.g. draft replies in EZD (electronic document management system)