How to audit your team's workflows for AI readiness.
Before you bolt AI onto your business, audit the work underneath. Most teams don't have an AI problem — they have a clarity problem. This is the audit we run before any business workflow automation project at Keel.
Why audit before you automate
AI and automation tools amplify whatever process they sit on top of. If a workflow lives in someone's head — the steps, the exceptions, the "actually, we do it this way now" — automating it just hides the chaos behind a faster interface.
The shift you're after is from tribal knowledge (it works because Sarah knows) to system knowledge (it works because the system knows). The audit is how you get there.
Step 1 — Pick one real workflow
Don't try to map the whole company. Pick one workflow that is high-volume, painful, or about to be handed off. Good candidates: client onboarding, content production, support triage, billing, hiring, lead handoff between sales and ops.
Step 2 — Map the owners
For every step in the workflow, answer:
- Who is accountable for this step happening?
- Who actually does the work?
- Who has to be informed when it changes?
- What happens if that person is out for a week?
Any step where the answer to the last question is "it stops" is a tribal-knowledge step. Flag it.
Step 3 — Map the handoffs
Handoffs are where work goes to die. For each handoff between people, teams, or tools, write down:
- What gets passed (a doc, a Slack message, an unspoken nudge)?
- Where does it live afterwards?
- How does the next person know it's their turn?
- How would an AI agent know it's its turn?
If the answer involves "they just know," the handoff isn't ready for automation yet.
Step 4 — Map the decision points
Most workflows have two or three real decisions hidden inside twenty steps. Find them. For each one, capture:
- What inputs the decision depends on
- What the rule actually is (not the rule people quote)
- Who has authority to override it
- What an obvious wrong answer would look like
Decision points are where AI helps most — but only if the rule is written down somewhere other than a person's head.
Step 5 — Score AI readiness
For each step, ask three questions:
- Is the input structured? Can the next step reliably find what it needs, in a known place, in a known shape?
- Is the rule explicit? Could a new hire — or a model — follow it without asking?
- Is the output observable? Can you tell, after the fact, whether the step happened correctly?
Three yeses means the step is ready for automation. Anything less means the work needs shape before the tool goes in.
Step 6 — Fix the shape, then add the AI
The output of the audit isn't a roadmap of AI features. It's a shortlist of the workflow's weakest joints — the handoffs, decisions, and ownership gaps that need to be made explicit first. Fix those, then layer in automation and AI where it actually compounds.
This is what we mean by making the work ready before you make the AI powerful. The audit is the first step.
Want us to run this with you?
Keel runs this audit as a focused engagement for growing teams. You leave with a mapped workflow, a readiness score, and a concrete plan for what to automate first — and what to leave to humans for now.