Idea-to-Design vs. Code-to-Launch (AI Investment Gap)

Cohen splits the product lifecycle into two macro stages when auditing where AI agent investment has actually gone:

Code-to-launch — coding agents, review agents, maintenance/on-call agents. This stage has already absorbed the bulk of agent investment industry-wide.
Idea-to-design — research, ideation, design exploration. Comparatively under-invested, despite being earlier and arguably harder to automate (see Design as the Hardest Craft for AI to Replicate — and Designers' Cross-Training Advantage).

Cohen treats this asymmetry as a gap to close deliberately rather than a natural order of things — LinkedIn's Narrow, Single-Purpose Internal Agent Suite (LinkedIn) includes agents (growth prioritization, trust review) aimed specifically at the idea-to-design half of the pipeline.

Apply: audit AI tooling spend across the full lifecycle, not just the coding stage, and fund the earlier stages on purpose.

Из тем: AI and the Changing Shape of Product Work