Retriever is built around making a company’s knowledge searchable for AI products and internal teams. In a signed-in Retriever tab, Strawberry can inspect the sources and results you are testing, compare answers with their evidence, identify coverage gaps, and prepare a retrieval-quality review that makes it clear what needs fixing.
The work stays in the Retriever workspace you already use, and anything that sends, publishes, changes a record, or spends money remains for you to approve.
01
Treat a weak answer as a retrieval problem to investigate.
When an AI answer misses the mark, the fault may be missing source material, an ambiguous question, stale documentation, or the result selection itself. Strawberry can organise the visible evidence into a diagnostic brief instead of labelling every bad answer a model problem.
02
See which documents actually support the response.
A source link is useful only when it backs the specific claim being made.
Strawberry can compare the answer, retrieved passages, and source documents you open to point out where support is strong, partial, or absent.
03
Turn knowledge gaps into an owned documentation queue.
The right fix for a recurring unanswered question is usually not another prompt tweak.
Strawberry can draft a backlog of missing documents, contradictory guidance, and owners to involve, with the test question that revealed each gap.
04
Re-test the questions that matter after the knowledge base changes.
A release-based check makes sense for Retriever because new documents, chunking changes, or source updates can shift answer quality overnight. Save a small set of critical retrieval tests as a skill and schedule a routine after each content or configuration release to prepare the evidence for review.
No. Use Strawberry in the signed-in Retriever tab where you already access the workspace.
It can prepare context, drafts, and proposed next steps from the visible workspace. You review the target and outcome before a change is made in Retriever.
Yes. Open the relevant Retriever view and supporting context, then ask for a focused internal review that keeps the visible evidence attached.
Yes. Save a defined set of critical search tests as a Retriever skill, then run it after a source or configuration release to prepare the evidence for the owner.
Use the signed-in Retriever workspace you already control, start with read-and-prepare work, and verify the particular downstream consequence before approving an action.
Retriever Strawberry