Before a wrong answer reaches a client, measure your retrieval.
Independent evaluation of search, Q&A and RAG over contracts, case law, regulations, filings, policies and research.
Opens our booking page on Calendly, a US service; it asks for your name and email. Prefer email? info@tensoropt.ai
Why the stakes are different
A missed or wrong document, clause or figure can have legal or financial consequences, and nobody notices at the time, because people only see what search returned.
Legal
- Tested AI legal research tools “each hallucinate between 17% and 33% of the time” (Magesh et al., 2025, Journal of Empirical Legal Studies).
- A public database counted 1,058 US court cases involving improper generative AI use in filings as of the start of June 2026 (Charlotin database, as summarised by the UNC Law Library).
Finance
- ESMA, to investment firms that use AI: “firms’ decisions remain the responsibility of management bodies”, and quality assurance “should include thorough testing of algorithms and their outcomes for accuracy” (ESMA public statement, 30 May 2024).
- FINRA reminds its member firms that its rules and the securities laws “continue to apply when member firms use Gen AI or similar technologies” (FINRA Regulatory Notice 24-09, June 2024).
These are context. TensorOpt does not claim that its work satisfies any regulator.
How it starts
A Readout. It works from your log, which needs no labelling, or from 30 to 100 questions your experts write, each with the document that answers it. It tells you which layer to measure first, how to label, and gives a fixed quote for the Evaluation. Packages and prices
Where accuracy is lost
Four layers, measured in this order:
- Extraction. For scanned or OCR-processed documents: is the text read correctly?
- Retrieval. Is the passage that answers the question among the results?
- Citation. Does the cited passage support the answer?
- Generation. Does the answer stay within what the retrieved passages say?
Measure the earliest failing layer first.
What this is not
Not a validation package, not a compliance certificate, not legal or financial advice, and not implementation. A report states what was measured, on what sample, with what uncertainty.
Book a 30-minute call
A call to decide whether measuring your retrieval is worth it, and what the smallest useful first step is. The booking form asks what your search or RAG system runs on, and what the measurement should help you decide, and by when. Please do not send documents, queries or personal data before the call.
Opens our booking page on Calendly, a US service; it asks for your name and email. Prefer email? info@tensoropt.ai
Measure the search you can't see
Learn how to tell whether your search works, and how to measure a change before you make it. Five short emails over two weeks:
- Why misses stay invisible: users judge what is on the screen, never what is missing.
- The measurement ladder: nine rungs, and where your system sits.
- The first rung to climb: a log of each query and the results it returned.
- Getting started safely: the free plugin through security review on Liferay DXP, or, on other stacks, where accuracy is lost first.
- Two averages are not a decision: what to measure before you switch on semantic search or change a parser, embedding or model.
Then at most two short follow-ups. You also get practical files that are not on the site: a plugin rollout checklist (Liferay DXP) or a logging schema (other stacks), and a labelling kit.