Operational knowledge often lives with the people doing the work rather than in the documentation. Asking an AI system to “write a runbook” can produce a polished document while leaving critical assumptions, exceptions, and recovery steps unexplored.

This session examines an interview-led approach in which an AI assistant asks structured questions about one bounded operational procedure. We’ll explore how it can help surface prerequisites, decision points, expected outcomes, failure conditions, and escalation paths before assembling a draft runbook.

The focus is on assisting knowledge capture—not claiming that an agent can independently discover how an organisation operates. We’ll distinguish information supplied by a subject-matter expert from assumptions introduced by the model, and examine how unresolved questions should remain visible rather than being filled with plausible instructions.

We’ll also cover the human review needed before a draft becomes an approved procedure, including sensitive-information handling, ownership, versioning, and validation by someone other than its author.

Attendees will leave able to:
- Structure an AI-assisted interview around a bounded operational procedure.
- Identify unsupported assumptions and missing decision or recovery steps.
- Define a review and validation process before publishing an AI-assisted runbook.

Audience: Operations engineers, platform teams, architects, and technical process owners.
Level: Introductory to intermediate.
Duration: 45 minutes.
Prerequisites: Familiarity with operational procedures; no coding or machine-learning knowledge required.
Format: Slides and a worked interview-to-draft example.
Scope: Knowledge capture and reviewed documentation—not autonomous workflow observation, procedure execution, or continuous documentation maintenance.