Analect pilot

Test Analect against your own documents and questions.

Agree a representative document set and the questions that matter to your team. The pilot compares answer quality, token use and storage requirements, then gives you the evidence to decide whether to proceed. We apply the same measurement approach to your material; we do not promise your numbers will match ours.

What you need to bring

Three things, nothing more

  • A representative slice of documents — a few hundred pages is plenty, and it should include the awkward formats, not only the clean ones
  • The questions your teams actually ask of them, chosen before you see any fragments
  • Known-correct answers where you have them, so the report can score correctness rather than only agreement
  • A view of current AI spend, or a provider bill export, for Forecast

Requesting a conversation commits you to nothing. Scope and fee are agreed in writing before any work starts.

Request a scoping conversation

By submitting, you agree to be contacted about your enquiry. See our Privacy Policy.

How a pilot runs

Scope, test, review, decide

Step 1

Scope

We agree the slice in writing: which documents, how many, which questions, which models, and the baseline to compare against — including your current retrieval workflow where you have one.

Step 2

Test

We decompose the slice, run the questions against every agreed route, and weigh the storage under both retain-source and replace-source scenarios.

Step 3

Review

A written report: every number with its unit, the method, the acceptance criteria we agreed at the start, and the questions where fragments did no better. It is yours whatever you decide.

Step 4

Decide

You decide whether to proceed to production. Choosing not to is a normal outcome and costs you nothing further.

Terms

What a pilot commits either side to

The specifics — document count, question count, models, duration and fee — are set in the scoping document for your pilot. What does not vary is below.

FeeFixed, quoted and agreed in writing before any work starts. The fee is set from the scope: number of documents, number of questions and number of models tested.
Scope limitsDocument count, question count and models are capped in the scoping document. Anything beyond the cap is a change we agree and price, not a surprise on the invoice.
Your dataHandled under a written agreement covering where it is processed, who can access it and how long we keep it. Pilot material and anything derived from it is deleted on the schedule set in that agreement.
The baselineAgreed before testing. Where you have an existing retrieval workflow, it is one of the baselines — comparing only against a model reading whole documents is the easier test and we do not rely on it alone.
Acceptance criteriaSet with you at scoping, so the result is judged against a standard chosen before anyone saw an outcome.
What you receiveA written report covering answer quality against each baseline, token use with its categories stated, storage under both retention scenarios, and the questions where the approach did not help.
The decisionYours. If you proceed, we scope production implementation separately. If you do not, the report is still yours and there is nothing further owed.

Region & currency

Changes spelling, terminology, the data-protection regime named in our notices, and the currency used in indicative figures. It does not change where we are or quote you a price in your currency. ETT is headquartered in Dallas, with offices in London and Vancouver.