ai-product-death-cycle¶
Insight: Companies are creating a new, more dangerous product death cycle by using AI to replace user research rather than augment it. The original cycle (no users → ask customers → build features → repeat) at least involved humans; the new version (no users → ask AI → build AI features → repeat) cuts humans out entirely, producing technically perfect features that solve problems nobody has.
Detail: Tisza identifies the pattern where companies are under pressure to ship AI features, so they use AI to decide what AI features to ship — a self-referential loop. Successful AI projects she observed share common traits: they use AI to enhance user research (e.g., feeding actual interview transcripts to spot patterns) rather than replace it, start with user problems rather than AI capabilities, and treat AI as a tool rather than a strategy. The key test: when you ask "where's the data backing this?" and the answer is crickets, you're in the death cycle.
Sources
Related: existing entry "AI is commodifying knowledge work" in external/ai-assisted-design.md — COMPLEMENTS