The AI-Native Technical Writer

Confirmed Speaker Ane Tröger, PhDLead Technical Writer at BayerTalk: Building a Diátaxis Classification Skill for Docs-as-CodeDescription: In internal docs-as-code repositories documentation tends to grow organically, rather than by design. Pages get added where it's convenient rather than where they belong, and the result is a flat, untyped structure that nobody fully understands. Platform migrations make it worse: whatever hierarchy existed before partially collapses, and content arrives as a batch. The cost lands on everyone — readers lose time in the wrong mode, writers lose confidence about where their contribution goes, and AI tools lose accuracy because mixed signals produce mixed outputs. One possible solution is to restructure the repo so that is easier to find information, easier to maintain, and potentially easier for AI to consume.This talk walks through building an AI skill that classifies documentation pages against the Diátaxis framework — tutorial, how-to, explanation, and reference — plus a fifth type, wayfinding, added to handle the navigation hubs every real repository contains. We'll look at how the skill is structured across three files itself, the four-step workflow it follows (read, classify, split, navigate), and why it produces a confidence level from signal distribution instead of a pass/fail verdict — with low confidence acting as the mixed-content flag. Suitable for technical writers at any level; sections on skill design and decision-making go deepest for experienced practitioners. The foundations of the Diátaxis framework and examples will be covered briefly, and if you want to get the most of this talk, I suggest taking a look at the Diátaxis framework ahead of time:[ https://diataxis.fr/start-here/](https://diataxis.fr/start-here/) Frances LiuCo-founder, Promptless Talk: What AI agents get wrong when writing docs Description: We regularly test frontier agents (Claude Code, Codex, and OpenCode with open models) on documentation tasks. The drafts usually come back polished and coherent, but their failures happened before drafting: deciding whether a change needed docs at all, working out who the docs were for, choosing which page the change belonged in, and answering what the reader needed to do rather than describing what the feature does. We'll go through which agents hold up on which kinds of task, what context raises their odds of success, and how to review their output so the quiet failures don't reach your readers. You'll leave with a checklist for deciding what to hand to an agent, and what to check when it hands work back.Helena JerneyHead of Information Experience at StreetLight Data Talk: Information architecture for the AI audienceDescription: What it takes for content to be consumed by AI to enable optimal user experience. Where content and context architecture intersect and how to manage the information pipeline to address the needs and limitations of LLMs. More details to come.