AI Decision Checklist
A pre-adoption checklist for AI decisions: accountability, failure modes, assumptions, and when to pause instead of automating.
This checklist is for teams considering AI automation. It’s designed to support judgment and defensibility—especially in real operations where exceptions, accountability, and trust matter.
Pre-Adoption Checklist
If you can’t answer these, don’t automate yet.
- What breaks if this fails?
- Who is accountable for outcomes (a person, not “the system”)?
- What assumptions does this make about inputs, categories, and definitions?
- What happens if the data is wrong or incomplete?
- How will exceptions be handled and escalated?
- What’s the human review path for low-confidence cases?
- How will we know if it’s working (one or two weekly metrics)?
- What is explicitly out of scope (what we will not automate)?
Operational guidance
If the workflow is judgment-heavy, relationship-based, or carries moral/accountability weight, treat AI as decision support. Use AI to summarize, extract fields, and highlight exceptions—then keep the final call with a person.
If you can’t answer these, don’t automate yet.
Pause and clarify the workflow. Document the decision points, define required fields, and assign ownership. Automation becomes calmer and more reliable when the human process is clear.
Soft CTA
This checklist is part of our AI Automation Audit process. If you want help applying it to your workflows, request an audit and we’ll scope what’s worth automating—and what isn’t.