How to Run a 3-Hour Discovery Sprint with AI
You don't need two weeks and a dozen user interviews to validate an idea. Here's a repeatable process for running a fast, evidence-backed discovery sprint using Specky.
The problem with slow discovery
Traditional discovery cycles take weeks. Schedule interviews, conduct them, transcribe, analyse, synthesise, present. By the time you have findings, the team has already moved on or made the decision without you.
AI changes this. Here's a process we use internally at Specky that consistently produces strong signal in under three hours.
The 3-hour discovery sprint
Hour 1: Signal gathering (60 min)
- Open the Discovery Workbench and search for the topic you're exploring
- Review the auto-surfaced themes from the last 90 days of Gong calls and Slack threads
- Open Alex Research and start a session: "Help me understand what customers are saying about [topic]"
- Use Alex to challenge your initial hypotheses — it will push back if the data doesn't support your assumption
By the end of this hour, you should have a shortlist of 3–5 signal clusters worth exploring further.
Hour 2: Opportunity mapping (60 min)
- For each signal cluster, create an opportunity node in the Opportunity Tree
- Use the AI to score each opportunity against your strategy filters
- Deploy an Alex campaign to 10–20 users in the relevant segment — a short 5-question interview
- While the campaign runs, draft a lightweight opportunity brief in the PRD editor
Hour 3: Decision prep (60 min)
- Review the Alex campaign responses as they come in (most complete within the hour)
- Update your opportunity scores with the new signal
Keep reading
Most "AI Features" Shouldn't Be AI Features. Here's the Test That Tells You.
Most "AI features" shipped in 2026 are rules engines wearing an AI costume. Here's the four-question test to run before you spec one — and what it looks like when a feature fails all four.
The Product Velocity Illusion: Why Shipping Faster Hasn't Made Your Roadmap Better
AI collapsed the time it takes to build a feature, but not the time it takes to decide what to build or validate that it worked. The Three Clocks framework shows product teams where their real bottleneck moved — and how to stop mistaking build speed for progress.
Your Next User Might Not Be Human: A PM's Framework for Agent-Ready Products
Software is quietly picking up a second audience — AI agents acting on a human's behalf. Here's a framework for auditing whether your product actually works for both.