AI webinar for public administration

WEBINAR FOR PUBLIC ADMINISTRATION

WEBINAR SCRIPT (Duration: 90 minutes)

Title: AI in e-government: From concerns to safe practice. How can an authority implement artificial intelligence wisely?

• 90 minutes — of practical discussion
• 3 theses — from needs to tools
• Q&A — questions and answers
• 87% — want to participate in training
• 70% — point to missing procedures
• 46% — have already used AI

Using specific examples, we will show how to use AI safely in a public authority, protect data and prepare the organisation for a structured implementation

Webinar dates

Date 1

25 September 2026 10:00-11:30 90 minutes

Date 2

2 October 2026 10:00-11:30 90 minutes

Date 3

9 October 2026 10:00-11:30 90 minutes

Date 4

17 October 2026 10:00-11:30 90 minutes

Webinar programme

10 min — Introduction – Welcome and warm-up

Welcoming participants: Introducing the presenters, the purpose of the meeting and the agenda.

Warm-up (a short chat poll): "Have you used AI tools at work this week?".

Introducing the topic: Emphasising that the webinar is based on data – the latest report diagnosing the attitudes of public-authority employees.

25 min — Part I: Thesis 1 – Authorities urgently need training. Findings from the NASK report

Starting fact: Generative artificial intelligence is already working in public authorities. During the past 6 months, as many as 46% of public officials used it for professional purposes. Most often (59%), they use it to search for facts and information.

Officials’ hopes: 81% of respondents believe AI will support the automation of routine tasks. They see enormous potential, among other things, in searching for information in documents using natural language (67% of responses).

Main barriers and concerns:

• Security concerns: 71% of the youngest officials fear for client privacy and unauthorised data processing.
• Implementation barriers: 73% point to unclear regulations or their absence, and 70% to a lack of internal procedures.

Conclusion (Thesis 1): The findings demonstrate the need to organise the rules, competences and method of implementing AI in public administration. As many as 87% of officials declare an interest in training. Training in technical skills (expected by 88% of those interested in training) and security (expected by 78%) is essential. Training is the foundation of safe e-government.

20 min — Part II: Thesis 2 – How should public administration be trained? The “AI in public administration” training catalogue

Presenting the solution: Presentation of a comprehensive Competence Development Programme with 8 dedicated modules.

A structure adapted to an authority: Each module consists of 4 hours of learning divided into 3 blocks of 80 minutes (each on a different day), preventing participant overload.

Practice instead of theory alone:

• Each training course ends with the development of a practical document or working material for use and application in the authority (e.g. a checklist, oversight map or pilot card).
• The modules cover, among other things, creating a secure knowledge base (RAG), legal issues (GDPR, AI Act) and implementing AI agents.

Conclusion (Thesis 2): Effective training for public administration must be structured, adapted to the realities of an authority (e.g. a shared format for small municipalities that reduces costs) and end with a concrete result that closes the gap caused by the lack of internal procedures.

20 min — Part III: Thesis 3 – Training as a gateway to secure tools: TWIN:DESK and PROXY:AI

From theory to a sovereign ecosystem: Training participants are ready to work with the target tools. It is time to present SAIE – the Sovereign Artificial Intelligence Ecosystem.

Presenting TWIN:DESK:

• Instead of public, risky chats, an official receives a secure web application (TWIN:DESK) for everyday work with documents.
• The system is based on RAG technology, allowing work with legal acts and procedures so that every AI answer includes a source reference (quotation and location).
• It guarantees confidentiality – files can be isolated and processed locally.

Presenting PROXY:AI:

• It addresses concerns about sensitive-data leakage. PROXY:AI acts as a secure gateway supervising information exchanged between the user and AI services.
• The system analyses prompts and has a built-in firewall (AI Firewall) that blocks or masks sensitive data (e.g. PESEL and NIP numbers) in real time before they leave the internal network.
• The tool allows administrators to manage access by assigning profiles and provides a complete audit trail of activities.

Conclusion (Thesis 3): Only the combination of competence (acquired during training) and suitable, controlled tools (TWIN:DESK for everyday substantive work and PROXY:AI for supervising security and costs) ensures the full success of an AI implementation in an authority.

15 min — Conclusion – Questions and answers (Q&A) and next steps

Q&A session: Answering questions submitted by participants in the chat.

Summary of the 3 theses: 1. Authorities feel both a need and concerns. 2. The remedy is dedicated training that concludes with documentation. 3. The objective is safe work using highly controlled TWIN:DESK and PROXY:AI tools.

Let us discuss a webinar for your institution

We will adapt the meeting to your team’s experience and the challenges faced by your authority