# AI/BI & Genie dashboards

> TechFabric builds AI/BI and Genie experiences on Databricks: the semantic layer and metric definitions underneath them, benchmark question suites that measure answer accuracy over time, dashboards reading those same definitions, and Unity Catalog permissions so a Genie answer respects who asked it. The two to four week Genie Accuracy engagement is the named way to start.

For: Analytics leads whose stakeholders still export everything to a spreadsheet
Canonical: https://www.techfabric.com/services/ai-bi-genie

---

Genie will answer any question you ask it. Making the answers right often enough that a VP stops checking them is a data modelling problem well before it is a prompt problem. Genie reads your tables, your column names and whatever instructions it was given, so an ambiguous grain or two columns both called revenue produce a confident wrong number and no warning. We build the layer underneath: metric definitions agreed with the people who own the number, joins and grain written down rather than inferred, and a set of benchmark questions with known answers that Genie has to keep passing before anyone trusts it in a board pack. Dashboards read the same definitions, so the figure on the slide and the figure Genie gives are the same figure.

## What it includes

- Metric definitions agreed with whoever owns the number, written once and read by both Genie and the dashboards
- Benchmark questions with known answers, run as a suite, so accuracy is measured rather than felt
- Grain and joins written down, instead of Genie inferring them from column names
- Row and column filters inherited from Unity Catalog, so an answer respects who is asking

## Questions

### Why does Genie give confident answers that are wrong?

Usually because the model underneath is ambiguous rather than because the language model is bad. Two tables both hold something called revenue, the grain changes halfway through a table nobody documented, or an instruction says one thing and a column name implies another. Genie resolves that ambiguity silently and moves on. The fix is upstream of the prompt every time.

### How do you measure whether Genie is accurate?

With a suite of benchmark questions that have known answers, run repeatedly rather than once at launch. That gives you a number that moves when someone renames a column, which is the only way to know accuracy is holding. The Genie Accuracy engagement at /databricks/genie-accuracy exists to build that suite and the definitions under it.

### Can Genie show someone data they should not see?

Not if the permissions are done properly, and this is the question worth asking early. Genie queries under the identity of the person asking, so row and column filters in Unity Catalog apply to the answer the same way they apply to a query. Where that boundary has not been drawn yet, it is governance work before it is Genie work, and it is at /services/data-ai-governance.

### If we have Genie, do we still need dashboards?

Yes, and they should read the same definitions. A dashboard answers the question everybody asks every Monday, and it should look identical every Monday. Genie answers the question somebody thought of in the meeting. Building them on separate definitions is how you end up with two numbers and an argument.