Calango LabAI-powered Public Management Assessment
A questionnaire that becomes an AI-generated digital maturity assessment, with a PDF and a public ranking of municipalities used at industry events.
Softplan serves municipalities and public bodies, and publishes the report "Panorama da Gestão Pública Municipal". Anyone reading a report like that learns the national average, but not where their own municipality stands — and it’s the second question that starts a commercial conversation.
The problem
Turning a market report into a conversation. A municipal manager can’t benchmark themselves by reading a PDF, and the sales team has no way of knowing who read it or what stage that municipality is at. What was missing was a route that answered the manager and qualified the contact in the same pass.
What was built
A landing page where the manager signs up, answers around ten questions in an embedded chat and receives, within seconds, a digital maturity assessment of their municipality: a score from 0 to 100, a level, a comparison against the national benchmark, three priorities and a contextual insight — with a PDF to download. Every assessment enters a live public ranking of municipalities, used at sector events.
- Landing page with sign-up, chat questionnaire and an assessment in seconds
- Assessment as a PDF, generated from a template and stored for permanent download
- Live public ranking of municipalities, deduplicated by municipality
- Results segmented by event and by landing page
- Commercial variant with a shorter form and a filterable ranking
- CRM integration for leads and campaign tracking
Decisions worth knowing
What separates a portfolio from an album of screenshots is the reasoning behind each choice.
The AI output is a contract, not free text
The assessment comes out through a tool call with a fixed JSON Schema. Score, level, benchmark and priorities always arrive in the same fields, typed. Reading free text and extracting a number from it works in the demo and breaks in production, the day the model decides to phrase its answer differently.
The AI answers from the report, not from memory
Semantic search runs over the Panorama da Gestão Pública Municipal, indexed for retrieval. The assessment stays anchored to sector data rather than to the model’s general impression — which is the difference between a defensible comparison and a pleasant-sounding number.
One codebase, two landing pages
The commercial variant reuses the entire home page through parameters, with a shorter form, its own ranking and a CRM submission that doesn’t block the flow. Duplicating the page would have been faster on day one and more expensive on every other day, because each change would then count twice.
Open to read, closed to write
The ranking is public, and that’s exactly why the page can’t write anything: it queries, and every record goes in from behind, through the server. Separating the two is a design decision, taken before the first screen — once the page is live, reversing it costs rewriting the whole path the data takes.
Outcome
In August 2026 the public ranking held 149 participants, with an average score of 46 and a maximum of 89. The three figures are open on the ranking page itself, which updates live with every new assessment.
Got a problem in the same family?
This case study is about custom software. If what you need looks like it, the first conversation costs nothing and commits you to nothing.