Outcome before tool
Each playbook begins with a real job to be done. We define the desired output, accountable user, approved inputs, and review standard before introducing prompts or selecting an AI tool.
Workflow, not prompt pile
A useful AI playbook includes preparation, source boundaries, prompts, example inputs and outputs, quality checks, escalation points, and a clear human approval gate.
Evidence stays visible
Material observations should remain traceable to approved sources. Playbooks instruct users to expose missing information, conflicting evidence, and uncertainty rather than letting polished language hide weak support.
Human decisions remain human
AI can assist with extraction, organization, comparison, drafting, and synthesis. It should not silently assume accountability for awards, legal conclusions, compliance determinations, employment decisions, medical decisions, or other consequential judgments.
Tool-neutral by design
Most workflows can be adapted for ChatGPT, Microsoft Copilot, Claude, Gemini, or another organization-approved model. Exact behavior varies by model, configuration, connected data, and organizational controls, so users must validate outputs in their own environment.
Responsible-use pattern
Our standard pattern is: define the task; approve the sources; constrain the model; require citations; surface uncertainty; review the output; document corrections; and retain human accountability.
Original and independently developed
Published materials are created as original, generic educational resources. They are not copied from an employer, client, vendor, or confidential operating process.