The Death of Amazon's Writing Culture

A former Amazon product manager argues that Amazon’s once-formidable writing culture - six-page narrative memos read silently in “study hall” to force clear thinking - has been hollowed out by AI. Where full-sentence memos once guaranteed that authors had done the necessary analysis before convening decision meetings, generative models now produce readable six-pagers without the author having thought through the plan. Reported consequences include senior presenters who haven’t read their own memos, widespread tagging of AI-written docs as “AI;DR” (AI; didn’t read), and performance reviews and promotion write-ups being auto-generated and rolled up by LLMs. That shift risks two bad outcomes: teams wasting senior time reworking unvetted plans, or executing poorly thought-through initiatives; both erode trust between colleagues and degrade long-term decision quality.

The remedy proposed is pragmatic: treat AI as appropriate for engineered, reversible “two-way door” work but preserve human authorship and review for irreversible “one-way door” decisions. Concrete rules recommended include owning every word (regardless of whether AI produced it), labeling quoted generated text, and always human-writing critical documents - shareholder letters, planning OPs, PR-FAQs, major rearchitectures, performance reviews, and promotion materials. A parallel policy from an engineering lead at Clay underpins these points: writing is thinking, you must stand behind each sentence, and authors should spend more time writing than readers spend reading. Use AI for brainstorming, editing, and consistency, but keep the human in the driver’s seat for consequential decisions.