AI Moves from Execution to Ideation
Summary
AI is crossing a threshold from a tool that executes instructions to one that generates ideas. At Anthropic, Claude is already looking through feedback, bug reports, and telemetry to suggest features to ship — not just implement ones humans have specified. This is a qualitative shift: AI is beginning to participate in the upstream problem space, not just the downstream delivery space.
The implication for PMs and adjacent disciplines: the boundary between “what AI does” and “what humans do” is moving upstream. For most of software’s history, humans owned ideation and AI (or automation) owned execution. That boundary is dissolving. Engineering is evolving from implementation as the core discipline to judgment — knowing which AI-generated ideas to pursue, which to discard, and which to escalate.
How to Apply
- For PMs: Treat AI as a prospective ideation partner, not just an execution accelerator. Experiment with prompting AI to analyze your own product’s telemetry, support tickets, or user feedback for feature hypotheses. The AI won’t always be right — but it will surface patterns that manual review misses.
- For engineering relationships: Recognize that your engineers are shifting from implementers to evaluators. The skills to invest in (theirs and yours): taste, judgment, specification quality, evaluation rigor. These are the human skills that survive the execution-to-ideation shift.
- For product strategy: If AI-generated feature ideas are real (and they are), the strategic question shifts from “how do we generate enough good ideas?” to “how do we evaluate and prioritize AI-generated ideas effectively?” Evaluation infrastructure — evals, decision frameworks, taste — becomes a strategic asset.
- Timeframe: This is happening now at frontier companies, but will reach most organizations in 2–3 years. PMs who develop AI-collaboration-for-ideation skills before it’s standard will have a significant early advantage.
Sources
From: 2026-02-19 Head of Claude Code: What Happens After Coding Is Solved
Key quote: “Claude is starting to come up with ideas. It’s looking through feedback, bug reports, and telemetry, then suggesting features to ship.” Attribution: Boris Cherny What this source adds: First-person observation from the head of Claude Code, describing what’s already happening at Anthropic — not a prediction but a current behavior. The specificity of “looking through feedback, bug reports, and telemetry” makes this concrete: AI is doing lightweight product analysis, not just generating abstract ideas. Links: Original | Archive
Related
- AI Disrupts Strategic PM Skills Most — Consistent with: AI disrupting strategy/ideation is precisely what “AI moving to ideation” means at the engineering level
- Generalists Outperform Specialists in AI Era — As AI handles more execution and ideation, cross-domain judgment becomes the differentiating human skill
- Latent Demand as Product Signal — AI scanning telemetry is a systematized version of the latent demand discovery pattern