BLUEHARBOURD3
Book an intro call

Depth 10.1 · AI engineering and automation · Level 3

Assistants and copilots

An AI assistant, sometimes called a copilot, is a tool that answers questions or drafts text by combining a language model with a company's own information. A good one quotes the document it relied on, so the person reading can check it. It does not know anything outside the material it has been given, and it can still misread that material.

Specification

Input
The document set the assistant should answer from, the questions people ask today, and the access rules that already apply to those documents.
Method
We index the documents with their permissions, build retrieval so each answer cites the passages it used, and test the assistant against questions your staff have already answered by hand.
Deliverable
Retrieval assistant over internal docs: a working assistant, the test question set and a page of known limits written for its users.
Measured by
The share of test questions answered correctly with the right source cited, and how often the assistant correctly says it does not know.

Most often bought inB2B software, Professional services, Ecommerce

The artefact, before and after

Retrieval assistant over internal docs: a working assistant, the test question set and a page of known limits written for its users.

Before: a blank assistants and copilots record with a row for each of document ingestion that keeps access permissions, retrieval with cited sources, refusal when no source is found, feedback capture from users, a written list of known limits, and no entries. After: every row carries a finding and a checked status, which is the state the work hands over.

Inside it

  • 01Document ingestion that keeps access permissions
  • 02Retrieval with cited sources
  • 03Refusal when no source is found
  • 04Feedback capture from users
  • 05A written list of known limits
Deliverable · Level 4→ Retrieval assistant over internal docsA retrieval assistant over internal docs answers staff or customer questions from a defined set of your own documents, and shows the passage each answer came from. When the documents do not hold an answer, it says so instead of guessing.

Typical first engagement

The usual order of a first piece of work. Timing and price are set at scoping, once we have seen the stack.

  1. 01Use case and sourcesYou get: One task, the documents it may draw on and who may ask it.
  2. 02Evaluation setYou get: Questions with expected answers and source passages, written with your team.
  3. 03PrototypeYou get: A working assistant that cites passages, scored against the set.
  4. 04DecisionYou get: Scores, failure cases and a recommendation to extend, change or stop.

AI engineering and automation

Next in this line

Ask about this part of the stack

Write one question about assistants and copilots in your own systems. We answer it in writing and name what we would need to see to answer it properly.

Send one question

A written reply, no call needed.