AI Startup Strategy

An enterprise AI startup building a decision platform for organisations in the middle of a transformation. It was pre-MVP: a lot of strategy and no code yet. In the room for the business alignment workshop were the founder and CEO, a co-founder, two more people from the client side, a facilitator and me. I led the follow-up product and brainstorming sessions and owned the demo crib sheet.

Pre-work worksheets from three people on the client side, each with a "future headline" for the company. A stress test of the strategy found that the two founders were not yet telling the same story. The workshop left three questions open: work beside consulting firms or replace them, which market first, and what the words in the pitch actually meant.

Decide in the room, not in the deck: nine open decisions, each with an owner.

The alignment workshop ended with four product calls and a follow-up. Anchor the first product on a digital map of how an organisation is structured today. Start with a narrower segment where the return is clearer and buying is faster. Position the tool beside consulting firms, so the client keeps its own knowledge. Tighten two key words so the messaging holds.

We then planned a 90-minute strategy workshop around those nine open decisions, each with a deadline: three before the partner meeting, two before investors, four before the build. We also corrected our own method. Early drafts assumed the client only had to approve our research; the workshop became a co-creation session instead.

A presenter-led demo of seven screens, about eight minutes long, with a 30-beat guided tour. Every request from the review rounds was logged and mapped to a change. The strategy went through five versions. The PRD that followed has stable IDs, P0 to P2 priorities and eight falsifiable hypotheses.

The first strategy set came from a chain of six research agents, covering discovery synthesis, competitive research, vision and personas. A separate QA agent reviewed it and found three critical issues before anything reached the client. A scoring script with no model in it graded the strategy documents.

Time saved: not measured.

• Alignment workshop, pre-work packs and decisions list with owners
• Strategic synthesis
• Personas: two primary, one channel
• Jobs-to-be-done analysis, competitor map, stress test
• Product strategy, five versions
• Wireframes
• Seven-screen interactive demo
• PRD