Own products · systems analysis
AnalystCraft: how I made AI-assisted requirements traceable to their sources
How Innokenty Bodrov created AnalystCraft Coworker to trace AI-assisted requirements back to the source documents.
← Cases & experienceThe problem
A draft specification can look convincing until an engineer asks, “Why is this field required?” If the answer is spread across API documentation, an old Confluence page, and a provider email, the analyst needs to show where the requirement came from and who confirmed it.
I created AnalystCraft Coworker, an AI assistant for systems analysts, to support a workflow where teams can trace a requirement back to the source that supports it.
The product decision
The analyst starts with a source map that records when each document was updated and what status it has. They then assemble the context for the task: confirmed facts, terms, constraints, and open questions. Coworker helps surface disagreements and draft a specification with source references. When two documents prescribe different rules, the disagreement stays visible until the analyst or process owner resolves it.
What this looks like in practice
In a public teaching example, a team is preparing to move to a new version of a payments API. It has the current contract, an old Confluence page, a prototype built for the previous version, and an email from the provider. The new contract requires an Idempotency-Key header that the prototype does not send.
The draft shows where the requirement came from, which implementation it conflicts with, and what needs checking in the new integration. If the sources leave an open question, the team can take it to the process owner before development starts.
What you can inspect
The published walkthrough includes the source materials, the source map, and the draft specification. You can follow a requirement back to its evidence and see which decisions the team still owns.
Read the walkthrough and source-map template (in Russian)