What if Polish civil servants had British tools?
Author: Andrzej Albera, Responsible for strategy and the company's direction
The UK government gave an AI assistant to more than 20,000 civil servants. In the three-month pilot, participants saved an average of 26 minutes a day, which the UK government puts at almost two weeks of work a year. We calculated what this would mean in Poland. The calculation assumes that a Polish official with a good tool will achieve the same as a British one. There is no reason why it should be otherwise.
How we calculated it
| Step | Value |
|---|---|
| Time saved per day (United Kingdom) | 26 minutes |
| Working days per year | approx. 220 |
| Time saved per person per year | approx. 95 hours |
| Local government officials in Poland (2024) | approx. 261,000 |
| Total | approx. 24.9 million hours a year |
| Working hours per FTE | approx. 1,760 (220 days of 8 hours) |
| Result | approx. 14,100 FTEs |
This covers only municipalities, counties and regions. With central government administration the result would be considerably higher.
Poland and the world
| Country | What it did | Effect |
|---|---|---|
| United Kingdom | AI assistant for more than 20,000 civil servants | 26 minutes a day per person |
| United Kingdom | Consult, a tool that analyses responses to public consultations | an estimated 75,000 days of work and GBP 20 million a year |
| Denmark | agreement between the government, municipalities and regions | target: at least 30,000 FTEs freed up by 2035 |
| Singapore | Pair assistant for public officers, results in an open quarterly report | an average of 81 minutes saved per task, almost 100,000 active users a month |
In Poland, almost half of officials (46%) already use generative AI at work, although the vast majority of respondents work in offices that have not deployed such tools (NASK, 2026). We wrote about it in AI at work in public administration: out of the grey zone. The will is already there; what is missing is an official tool and a way to measure the effects.
How to read this number
- Jobs do not disappear. The time won back goes to serving residents, shorter queues and backlogged cases. The Danes wrote explicitly that the aim is work closer to citizens.
- 26 minutes is what participants reported in the British pilot. That is why it pays to measure the effect from day one, as Singapore does.
- Simple tasks work fastest: answers based on public information bulletins and resolutions, draft letters, meeting notes, analysis of consultation comments.
What this means for institutions
- Give officials an official assistant that knows the office's regulations and documents. TWIN:DESK, our Digital AI Desk, runs in the office's own server room, on Polish models such as Bielik and PLLuM. Read more about deploying AI in public offices on our page for public administration.
- Measure from day one. PROXY:AI records who uses AI and for what. With such a log, an office can count its own minutes, not British ones.
- Deploy methodically. The CDF methodology starts with an initial assessment and leads to oversight, and our training for civil servants teaches people how to make use of it.
The timing is right. We have the law, we have a Polish AI stack and we have officials who want to work better. 14,100 FTEs are waiting for a decision. We will talk about it at the 2nd Forum on AI in Public Administration in Wrocław, 4-6 November.
Sources
- GOV.UK, 2.06.2025: Landmark government trial shows AI could save civil servants nearly 2 weeks a year
- GOV.UK, 14.05.2025: Government-built Humphrey AI tool reviews responses to consultation for first time
- Altinget, 10.06.2025: AI to free up at least 30,000 FTEs (in Danish)
- Pair public report (Q2 2026)
- NIST, Analyses and Opinions 8(76)/2026: number of local government officials, GUS data (in Polish)
- NASK, 17.03.2026: AI in public e-administration. The perspective of officials and institutions (in Polish)