You Can’t Prompt Your Way to Trust
- Sponsored Session
- 24. September
- 11:00 AM (CEST) - 11:30 AM (CEST)
- Aluminium
-
Stefan Gentz
- Adobe Systems GmbH
Contents
Generative AI can produce convincing answers, but no prompt can repair outdated sources, restore missing context, resolve conflicting versions, or create provenance. As people and AI agents increasingly rely on conversational answers, trust must come from the governed knowledge foundation beneath them – especially when an incorrect version, overlooked prerequisite, or inapplicable rule can lead to costly decisions.
Building on DITA, metadata, taxonomies, knowledge graphs, and guardrails, Stefan Gentz explores how structured content becomes an operational knowledge system. He examines why production environments must preserve relationships, versions, context, and access rules, following the anatomy of a trusted answer from governed source content through retrieval and reasoning to evidence, attribution, and controlled access.
A trustworthy system must also recognize when evidence is insufficient and refuse to invent an answer. Stefan shows how unanswered questions, content-quality signals, human feedback, analytics, and evaluations create a continuous improvement loop – and why Technical Communication plays a critical role in building reliable knowledge for both humans and AI agents.
Takeaways
- Trust starts below the prompt: Understand why prompting cannot compensate for missing structure, context, provenance, or governance.
- Structure becomes knowledge: Explore how metadata, versions, taxonomies, and relationships enable contextual, multi-step reasoning.
- Trust by design: Learn how evidence, attribution, filtering, access control, and refusal reduce the risk of plausible but incorrect answers.
- One foundation, two audiences: See how governed knowledge can serve humans and AI agents through conversational experiences, APIs, and MCP.
- Content gets a feedback loop: Discover how unanswered questions, analytics, feedback, and evaluations expose knowledge gaps and drive continuous improvement.