Less retyping, more analysis
The craft of business analysis is asking the right questions and controlling scope. Most of the day goes on neither: it goes on writing the same documents from an old file, and then editing all of them when something changes.
Where the hours actually go
Nothing here is a productivity claim. It is a description of the work, and where the product sits inside it.
Discovery
The first document exists to expose what nobody has answered yet — which is what makes every later document honest.
- You
- Read the brief, list what is missing.
- The system
- Writes the technical questions first, with assumptions made explicit and the risk of each being wrong.
Requirement analysis
Feature lists are easy to write and impossible to defend. Outcomes with measures survive a change-control conversation.
- You
- Decide what the product actually has to do.
- The system
- Produces a PRD in outcomes and journeys, with exclusions given the same weight as inclusions.
Scoping
Scope disputes come from what was never written down. The generated document has a place for it.
- You
- Set the commercial boundaries.
- The system
- Writes the SOW from the questions and the PRD, so it cannot contradict them.
The rest of the pack
The estimate is written against the architecture and the SOW, not alongside them.
- You
- Sanity-check numbers and risks.
- The system
- Derives architecture, milestones, estimate, risks and proposal in dependency order.
Revision
The version you sent last week is still there, whatever you have done since.
- You
- Change what is wrong, in the document.
- The system
- Keeps every version, diffs any two, restores any one.
You stay the author
Generated text is a draft with your name going on it, so it is editable everywhere and nothing is ever overwritten. A hand edit is recorded exactly like a generation, attributed to you.
- Regenerate one section’s worth of thinking by basing a run on an earlier version.
- Give a run an instruction — a tone, a constraint, a client’s preference.
- Set milestone counts and weightings explicitly rather than accepting an inferred split.
- Compare any version against the one before it before you send anything.
- v4Regeneratedminimax · 10 min ago
- v3Hand-editedyou · 2 h ago
- v2Restored from v1you · 4 h ago
Every document looks like your firm’s
Page size, margins, typeface, logo placement, footer and confidentiality note come from the workspace, not from whichever template you started with. Two analysts on the same account cannot send differently formatted documents.
What you approve in the paper view is what the client receives — the preview is driven by the same format the exporters render from.
The engagement replaces the current storefront with a headless build. Success is measured against checkout conversion, not delivery dates.
- Lift checkout conversion by 15% within two quarters
- Cut publish time from days to minutes
| Item | Included | Notes |
| Catalogue migration | Yes | 4,200 SKUs |
| Payments | No | Existing gateway retained |
Read the client’s material once
Uploads are indexed against the project, so the audit you read in week one is still answering questions in week six — and so is anything a colleague added.
Ask instead of skim
The assistant answers from the engagement's own documents, with citations.
Ground the writing
Retrieved passages feed generation, so deliverables quote the client's constraints.
Nothing lost between people
Material belongs to the workspace, not to the person who uploaded it.
Try it on the engagement currently on your desk
Paste the brief, upload what the client sent, and read the questions it comes back with. That is the honest test.