Your first AI workflow as a BA — steal this exact stack.

Welcome to Issue #01.
Every Saturday I ship one field-tested lesson from live BA × AI delivery — the stuff I actually use on client work this week, not repackaged theory.

The 30-minute setup

On day one of any new engagement, I stand up the same three surfaces before I touch a single requirement. It takes about half an hour and it earns back days across the project.

1. A living project brief in Claude

One long document — objectives, stakeholders, in/out of scope, open questions, glossary. I paste it into a Claude Project so every prompt in that project inherits the same context. When I ask for a user story two weeks in, Claude already knows the domain, the actors, and the terms.

2. A prompt library, not one-off prompts

Discovery, requirements, stakeholder updates, test scenarios, risk logs — one saved prompt per artifact, tuned to the client's voice. The free BA Prompt Stack is the starter set; I extend it per project.

3. A "guardrail" prompt at the top of every session

Two lines: "You are working on {project}. Never invent data. If unsure, ask a clarifying question or say 'not confirmed'." This one paragraph kills 80% of the hallucinations that make AI output un-shippable.

What NOT to do

• Don't drop raw AI output in front of stakeholders. Always run one human pass — you are still the analyst on record.
• Don't put client PII in a public model. Anonymize or use an enterprise tier. This is a contract-level rule, not a preference.
• Don't skip the traceability. Every AI- drafted story still needs a source — the transcript, the workshop note, the ticket. If you can't cite it, don't ship it.

Try this week

Pick one artifact you write every week — status updates, meeting summaries, story drafts. Build one really good prompt for it. Save it. Reuse it Monday. That's the whole game: one workflow at a time.

See you next Saturday.
— Dibya Mishra, Senior Business Analyst