Checklist · 15 min
AI operations audit checklist
The 20 questions we ask in the first week of an Assess, so you can run one yourself.
In short
What is an AI operations audit?
A structured look at how work actually flows through a function: volumes, cycle times, the systems it touches, and where human attention is spent. Its output is a shortlist of functions that could run AI-native.
Volume and flow
- How many items does this function handle per month?
- What is the cycle time from trigger to done, median and worst case?
- Where does work wait, and for whom?
- What share of items are routine versus exceptions?
Inputs and systems
- What documents, messages or records start the work?
- Which systems are read and written, and do they have APIs?
- Where does the same data get re-typed?
- Who owns access to each system?
Decisions and risk
- Which steps are judgement calls, and by whom?
- What must never happen without a person?
- What is the cost of a mistake, and how is it caught today?
- Which steps are externally visible to customers or regulators?
Readiness
- Who would supervise an agent doing this work?
- Is there a written procedure, or does it live in people's heads?
- Can we get 50 real past examples to build an evaluation set?
- Who on the leadership team owns the decision to change this function?