It understands · AI-Powered Search
Ask your data in plain language — and get the same number every time
Ask in plain business language
"Text-to-SQL" demos are easy; trusting the number is hard. Veracy separates the two: an AI helps you phrase the question in your organisation's own vocabulary, and a deterministic engine compiles it into SQL over governed data. The model never touches the number.
Today
Chat-with-your-data tools guess at table names, invent joins and return a confident figure that changes when you ask again. Analysts end up checking every answer by hand, which is slower than writing the query themselves.
What Veracy does
- 01
Your question is matched to the semantic layer — the metrics, dimensions and terms defined once from the contract — not to raw table names.
- 02
The engine compiles a deterministic query plan. Same question, same plan, same number, whoever asks and whatever tool they use.
- 03
The answer ships with a receipt: which metric definition, which contract version, which rows of evidence. Anyone can verify it, offline.
What you get
No hallucinated numbers
the language model is never in the calculation path; it can only choose among governed definitions.
One answer across tools
dashboards, spreadsheets and AI agents asking the same question get the same figure.
Governed by default
access rules apply inside the engine, so a question can only see what its asker is allowed to see.
Questions
- Does an LLM generate the SQL?
- No. The language model helps map your words to governed definitions in the semantic layer. The SQL is compiled deterministically from those definitions, so the same question always produces the same query and the same number.
- What if my question is ambiguous?
- Veracy says so and asks — for example, which of two "revenue" metrics you mean — rather than guessing. Ambiguity is surfaced, never resolved silently.
- Can AI agents use it, not only people?
- Yes. The same governed path is exposed to agents (including over MCP), with the same access enforcement and the same verifiable receipt on every answer.
In the same job · it understands
Stop building your data platform. Just sign what matters.