AI agents have become the primary consumers of company documentation, accounting for two-thirds of measured doc traffic in July 2026 and outnumbering human page loads nearly 2:1 in August. That shift turns documentation errors into operational risk: a 2024 case held an airline liable for misleading chatbot information, and agents can amplify incorrect API parameters, policies, or recommendations across thousands of interactions. Because agents expect exhaustive, structured inputs while humans need mental models and judgment, knowledge management must move from static pages to systems that keep information authoritative, retrievable, and consistently updated for both audiences.
The emerging role of knowledge engineer builds that system: establishing authoritative sources and owners; connecting releases, code changes, support tickets, and feedback to content updates; publishing structured artifacts like Markdown and llms.txt; defining terminology, review, permissions, and freshness standards; and measuring agent queries and retrieval failures. Automation can speed updates and reduce engineering burden - Coinbase cut documentation update time from over 20 minutes to under 60 seconds and HubSpot halved documentation infrastructure effort - while subject-matter experts review key changes. Assigning clear ownership and operational processes gives companies an advantage in turning change into trusted, usable knowledge for people and agents.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.