Compare the Market cut the time to deliver a new report from six weeks to under one week, according to Emanuela Ciotti, the price comparison company's head of enterprise analytics, speaking at dbt Summit 2026.

Ciotti, who drew on seven years of working with dbt across three organizations, warned that AI trusts the data it consumes rather than validating it, so a weak data foundation gets amplified rather than exposed by an AI layer on top.

She described the problem as a "reporting factory": a BI team run like a service desk, SQL logic duplicated across reports and tools, and metrics defined differently from one dashboard to the next. An internal audit of the company's Power BI estate found 433 reports, 41% of them (179) already abandoned, each a one-off answer with no shared foundation.

Her framework has three stages. Discovering shifts analysts into analytics engineers, introducing version control, modular design and testing before any new tooling. Scaling tackles a single monorepo with more than 4,000 interdependent models, including a staging layer dependent on Google Sheets seeds that occasionally broke financial pipelines before month-end reporting; the fix was a multi-repo architecture aligned to domain ownership, drawing on dbt Mesh principles so each business tribe owns its models. The team is now transforming, redefining what analytics produces: tested, documented data models instead of dashboards, positioned as the foundation AI needs to be trustworthy.

Alongside the drop in delivery time, test coverage passed 75% and model reuse climbed to 3.0x. Ciotti framed the shift as an operating-model change, not just a stack change: a team moves from service desk to platform only when what it hands the business changes from a dashboard to a governed model, which requires designing ownership in from the start.

Source: dbt Labs Blog