Why summaries are a waste of tokens and how we encoded the Shape Up methodology into Gemini to automate product strategy.
The High Cost of “Alignment”
In the early stages of a startup, communication is a double-edged sword. You need syncs, but every hour spent in a Zoom call is an hour stolen from execution. At CultLab, we hit a wall: the “Founder-to-Engineer” translation gap. We were burning 32 hours a month just documenting decisions.
Most people solve this with an AI summarizer. That is a rookie mistake.
A summary is passive. It’s noise. As a Growth Hacker, my goal wasn’t to “record” meetings; it was to automate the transition from talk to tech-spec.
The Hack: Shape Up as an LLM Constraint
We didn’t just prompt Gemini; we rewired its reasoning using the Shape Up framework. Why? Because LLMs love to hallucinate “big picture” fluff. By forcing the agent to think in terms of Appetite and Rabbit Holes, we turned a chatty bot into a ruthless Product Strategist.
The “Shaper” Architecture:
We built a pipeline that treats a Zoom transcript like raw data to be mined, not a story to be told.
- Fixed Appetite vs. Variable Scope: We instructed the model to categorize projects into “Small Batches” (2 weeks) or “Big Batches” (6 weeks). If the transcript didn’t have enough data for a 6-week “bet,” the AI was programmed to flag it as a risk.
- The “Rabbit Hole” Filter: Most AI assistants miss the “No-Gos.” Our agent was engineered to identify what not to build, preventing engineering drift before it started.
Scaling Without Hiring (The Growth ROI)
This isn’t just about productivity; it’s about resource leverage. By automating the documentation and initial “shaping” of projects:
-
Founder Liquidity: We effectively “cloned” the founder’s strategic thinking, freeing up 4 full working days per month for high-level fundraising and growth hacking.
-
Engineering Velocity: Briefs are now generated in seconds. Designers and front-end teams receive Context, Problem, and Success Metrics instantly, reducing back-and-forth by ~70%.
-
The Cost of Zero: We eliminated the need for junior PMs or technical writers. The system scales with the volume of calls, not the size of the payroll.
Conclusion: Methodology > Model
The real competitive advantage in 2026 isn’t the model you use (we used Gemini 2.5 Flash for its speed and context window). The advantage is the methodology you encode within it.
We didn’t automate a task; we automated a cognitive workflow. That is how you scale a lean team to compete with incumbents.