Replica Torino: A gateway to privacy-safe urban intelligence

What if understanding a city were as simple as asking it a question?

That's the idea we set out to test with Replica Torino, our experiment under the ToMove "Call4Testing" programme, run alongside the City of Turin, LINKS Foundation, and 5T. The goal: put rich, privacy-safe urban intelligence directly into the hands of the people who shape public decisions: no SQL, no dashboards, no technical background required.

A synthetic city, built to be queried

At the heart of the platform sits a synthetic population of 2.19 million Turin residents, generated with our privacy-preserving models and validated against ISFORT (Audimob) mobility benchmarks. Around it, we harmonised more than twenty open and proprietary datasets, from ISTAT and AperTO to OMI and Overture Maps, into a single, queryable knowledge base on BigQuery.

Beneath the surface, those sources are organised into three macro categories: demographic (population, households, age structure), socio-economic (income, employment and social-hardship indicators), and mobility & infrastructure (public-transport stops, cycle lanes, parking and points of interest). Because they all live in one place, a single question can cut across all three at once.

The result is a conversational interface, powered by an MCP server and Claude Sonnet 4.5, where a public officer can simply ask: about demographics, mobility, accessibility, or what-if scenarios, and get an answer in plain language, complete with the data sources behind it.

Tested the way we always do: user-first

A tool only matters if people actually use it. So across two sessions, eleven stakeholders from the City of Turin, LINKS, and Italy's Department for Digital Transformation put the platform through its paces. Their feedback drove a rapid iteration sprint: more concise answers, automatic source citations, CSV and Excel export, and a richer data foundation.

The verdict? Every target met or exceeded, statistical accuracy within 6% of real benchmarks, and, most telling, every single participant said they'd use Replica Torino again in their daily work.

What it shows, and what's next

Replica Torino is a working example of what trustworthy, human-centred AI looks like in practice: complex, sensitive data made accessible, transparent, and genuinely useful, even for users without a technical background. And the approach isn't limited to cities. The same privacy-safe synthetic-data methodology already powers synthetic users for commercial teams, in marketing, energy management and beyond, so it travels naturally from urban intelligence to industry.

This is just the beginning. We'll keep leveraging this approach, in cities and beyond. 🚀