Work · Case study

CERCAT, a public platform that understands what you ask

TechNLPPublic sector 2025
CERCAT, a public platform that understands what you ask
CERCAT · prototype · GovTech Catalunya
The brief

Public information in Catalonia is out there… scattered across a thousand websites and forms nobody understands. CERCAT was born to fix that: bring public services together in one place and make them usable by anyone.

Nobody commissioned this one: it was an entry. GovTech Catalunya, the Catalan government’s programme for finding technological answers to public challenges, and a team of four with one fixed idea: tear down barriers with technology. Not the first time either — in 2020 the same team won Hackovid with Ajudem.cat.

What we did

We built a platform that centralises public information and translates it into plain language. With natural language processing, you ask «what do I need to renew my DNI?» or «how do I register as self-employed?» and get a clear answer in seconds. No hopping between tabs.

We modelled procedures as data, not as pages. Every procedure has an authority, requirements, documents, deadlines and an official link; the answer is built from that and always cites its source. A search tool that makes things up is no use in public administration: here every answer shows where it came from.

We built the prototype and the entry: product, brand, demo and pitch to the jury. The project reached the GovTech Catalunya final, a nod to its potential to change the public sector and bring the government closer to people. If it moves to the next stage, it comes with a €41,000 budget for a real pilot.

And we left it at prototype stage — plainly said. CERCAT is not a public service in production; it is proof that it can be done. We put it that way because this studio doesn’t sell pilots as launches.

The key

The government already has all the information. What it doesn’t have is someone to understand it for you. The key was treating the citizen’s question as the starting point, in their words and with their doubts, and making the technology fit that instead of the other way round.

4
founders, the same ones who won Hackovid 2020
€41,000
budget for the pilot if it moves on
0
forms to decipher before you get an answer
How it works

Three steps and no forms. You write what is going on, the way you’d say it at the counter. CERCAT works out the intent: which procedure it is, which authority handles it and what situation you are in (first time? renewal? expired?). And it answers with steps, documents and the link to the official procedure, citing where each part came from.

From question to answer
01 An ordinary question “What do I need to renew my ID card?” free text ↓ 02 CERCAT reads the intent Natural language: procedure, body, situation answer in seconds ↓ 03 A clear answer, with a source Steps, documents and a link to the official form Finalist prototype at GovTech Catalunya. No forms to decipher.

What we learned, and now apply to any public product: the natural language part is the easy one. The hard part is making the answer reliable, and that is solved with well-modelled data and visible sources, not with more artificial intelligence.

How it went
01ChallengePublic information, scattered and opaque
02DataProcedures modelled with their source
03PrototypePlain-language question, clear answer
04GovTechEntry and the programme final
"Fewer headaches, more answers. This is how public services should talk to you."The project team
Credits · who took the shot
Product and technologyHumbert Blanco
Founding teamPau Riba
Founding teamMarc Celeiro
Founding teamRoger Bach

Want a result like this for your project?