🧭 Discovery & Planning
From Business Intent to Machine-Ready Requirements
Overview
Discovery has always been the highest-leverage, lowest-visibility phase of the SDLC — the place where ambiguity either gets resolved or gets baked into the backlog. AI changes the economics of this phase by compressing the distance between a stakeholder conversation and a structured, testable requirement.
Instead of a Business Analyst manually transcribing workshop notes into epics, an AI-augmented discovery process ingests transcripts, tickets, support logs and market research, then proposes user stories, acceptance criteria and a first-pass estimate — all of which a human strategist reviews, challenges and finalises. The BA role shifts from scribe to curator.
What AI Actually Does in This Phase
Automated Story Generation
Convert meeting transcripts and PRDs into epics, user stories and INVEST-aligned acceptance criteria.
Requirement Gap Detection
Flag ambiguous, conflicting or untestable requirements before they reach a sprint planning session.
AI-Assisted Estimation
Pattern-match new stories against historical velocity data for defensible, data-backed sizing.
Story Mapping
Auto-generate journey-based story maps that visualise scope, MVP cut-lines and release slices.
Market & Competitor Synthesis
Summarise competitor features, analyst reports and user feedback into prioritised opportunity briefs.
Traceability Seeding
Establish requirement-to-test-case linkage at the point of creation, not retrofitted later.
Enterprise Use Cases
- ✅Turning a 90-minute stakeholder workshop recording into a structured backlog within the same day
- ✅Mining support tickets and NPS verbatims to surface unmet requirements before a discovery sprint
- ✅Auto-drafting acceptance criteria in Gherkin format so QA can start test design in parallel with development
- ✅Generating three scope options (MVP / phased / full) with relative effort so sponsors can make a funding decision fast
- ✅Flagging requirement drift by diffing the current backlog against the original business case each sprint
How This Phase Fits the Framework
Precision-Led
Precision starts here — AI forces explicit, testable acceptance criteria instead of vague intent, closing the ambiguity gap before it compounds downstream.
AI-Augmented
AI augments the BA/PM, not replaces them: synthesis and pattern-matching are automated, judgement and prioritisation stay human.
Transformational
Transforms discovery from a weeks-long documentation exercise into a same-day structured backlog.
Execution
Execution-ready output: stories arrive with acceptance criteria and estimates attached, not just a title.
Lifecycle
Feeds the AADV™ engine from day one — baseline velocity and quality targets are set against real historical data, not guesswork.
Where to Start Looking
A starting shortlist, not an endorsement of any single vendor — the right tool depends on your existing stack and governance maturity.
| Tool / Category | Type | Best For | Notes |
|---|---|---|---|
| Jira + Atlassian Rovo | Backlog & requirements | Teams already on Atlassian | AI agents draft stories and summarise epics directly inside Jira/Confluence. |
| ChatGPT / Claude (enterprise) | Synthesis & drafting | Workshops, PRD drafting | Best for turning unstructured notes into structured requirement drafts. |
| Productboard AI | Prioritisation | Product-led orgs | Synthesises customer feedback into ranked opportunity themes. |
| Otter.ai / Fireflies | Transcription | Discovery workshops | Feeds clean transcripts into downstream story-generation prompts. |
| n8n / Zapier | Workflow glue | Connecting tools without code | Automates the ticket → transcript → draft-story pipeline end to end. |
What to Watch For
- ⚠️Hallucinated requirements. AI can confidently invent acceptance criteria that were never discussed — every generated story needs a named human owner before it enters the backlog.
- ⚠️Context loss at scale. Long transcripts and legacy docs exceed context windows; chunking strategy matters more than model choice.
- ⚠️Stakeholder over-trust. Sponsors may treat AI-drafted scope as final rather than a first draft — set that expectation explicitly.
- ⚠️Data sensitivity. Workshop transcripts often contain commercially sensitive detail; route through enterprise, not consumer, AI tools.
Pradeep’s Verdict
Discovery is the cheapest place in the SDLC to fix a mistake and the most expensive place to miss one. AI’s real value here isn’t speed for its own sake — it’s forcing precision earlier, so the requirement that reaches a developer in week three is the same one the sponsor actually meant in week one.
Want this mapped to your org’s actual SDLC?
I work with delivery leaders to translate this framework into a phased, governed rollout plan — starting with the phase that moves your AADV™ the most.
