An AI Co-Pilot for Business Analysts: Inside the BA Portal

Writing user stories, success criteria and test cases is some of the most repetitive, highest-leverage work a Business Analyst does — and some of the easiest to get subtly wrong under deadline pressure. We built the BA Portal to take the drafting off a BA's plate without ever taking the judgement off it.

It's a web portal where a Business Analyst captures business requirements, and AI drafts the user stories, success criteria and test cases underneath them — with every single AI suggestion sitting in a "pending approval" queue until a human accepts it.

The problem it solves

A BA's day is full of translation work: turning a business requirement into a properly-scoped user story ("As a [persona], I want to [action] so that [value]"), breaking that into success criteria, then making sure every success criterion actually has a test case behind it. Done by hand, it's slow, and coverage gaps are easy to miss — a success criterion nobody wrote a test for, an edge case nobody thought to raise. The BA Portal is built to catch exactly those gaps, at every stage, without doing the BA's thinking for them.

What it actually does

• Projects, requirements, and AI-drafted problem statements — set up a project with a name, description and problem statement, and generate that problem statement with AI from the description whenever it's useful.
• Create user stories with AI from a business requirement, in standard notation, with success criteria nested directly underneath — or write them manually if that's faster for a given case.
• Create test cases with AI — one traceable test case per success criterion that isn't already covered, with steps drawn from the user story and an expected result tied to that specific criterion.
• Four independent AI quality checks — of the business requirement set, a single user story, a story's success criteria, and test coverage — each run manually by a button click, each flagging specific gaps rather than rewriting anything.
• A "needs recheck" nudge whenever something has been edited since its last quality check, so stale AI feedback never sits there looking current.

The part we think matters most: nothing ships without a human

Every quality-check finding gives you exactly two options: "+ Add as draft", which creates a real record tagged Pending approval sitting inline with everything else, or "✕ Dismiss", which discards just that suggestion (it won't reappear until the check is re-run). A pending item can be approved, edited, or rejected like anything else — the AI never adds anything to the project automatically in the background. That single design decision is the difference between a tool a BA trusts and one they have to double-check line by line.

Built to keep working even when the AI provider hiccups

Every AI-calling route is wrapped so a single failed request can't take down the server or affect any other request in flight — and swapping to a different AI provider later touches exactly one file. Password reset and email verification are handled the same pragmatic way: reset tokens are stored as hashes only, so a database leak alone can't be used to reset someone's account, and if no email provider is configured, codes and reset links log straight to the console so the whole flow still works end-to-end in development.

The takeaway

The BA Portal isn't trying to replace a Business Analyst's judgement — it's trying to remove the drafting and gap-hunting grind around it, while keeping every decision about what actually goes in the project firmly in human hands. That's the same principle behind every AI tool we build: speed on the repetitive part, a human still holding the pen on everything that matters.