AI readiness checklist: six checks to run on one workflow before a pilot

Pick one workflow and check six things before any money goes into an AI pilot: who owns it, where its data lives, whether permissions are enforced, whether changes are tracked, what is already automated, and which AI tools people already use. Two weak answers are enough to wait.

Written by Ben Friedman, fractional forward-deployed AI engineerPublished · 2 min read

Key takeaways

  • Readiness belongs to a workflow. A company is never ready or unready as a whole.
  • Each check has a pass you can verify in the system, not in a meeting.
  • Ownership and permissions are the two that block a build outright.

The six checks and what a pass looks like

1. Ownership
One named person can approve a change to the workflow this week, without convening anyone.
2. Data location
Every field the task needs is in a system with an API. Nothing essential sits in a personal spreadsheet or an inbox.
3. Permissions
Access is enforced by the platform. You can state what a service account would be able to read.
4. Change tracking
You can find out who changed a workflow, a field or a rule last month.
5. Existing automation
Each rule has an owner and a known purpose, and failures notify someone.
6. AI already in use
You know which assistants staff use on this work today, and with what data.

Scoring yourself without flattery

Test each answer in the tool. For ownership, ask the named person to approve a small real change and see how long it takes. For permissions, log in as the least privileged user and try to open something sensitive. A check that passes in conversation and fails in the system is a fail.

If you want numbers, the scorecard asks two questions per area and returns a score for each.

What to fix first when several checks fail

Start with the owner, since every other fix needs someone to decide it. Permissions come second, because they are the gap that turns into an incident. Data location is usually the largest piece of work and can be narrowed: scope the first build to what the system already holds and move the spreadsheet step later.

When a checklist is not enough

A checklist tells you whether to proceed. It does not tell you what to build, in what order, or what it will cost. Those answers need someone reading the configuration itself, which is the job of the AI Readiness Audit.

Frequently asked questions

How long does the checklist take?
An hour for one workflow if you test the answers in the system, ten minutes if you only discuss them. The hour is the version worth doing.
Our workflow fails three checks. Is AI off the table?
For that workflow, for now. Either fix the three, or choose a different workflow that passes more of them and start there.

Prefer to have the six checks done inside your systems, with a written result? That is the audit.

About the author

Ben Friedman runs Viter, a forward-deployed AI engineering service. He has spent over ten years running Atlassian and operations tooling, including as internal Jira lead at Aroundtown, and holds the ACP-610, ACP-620 and ACP-120 certifications. He builds JAMES, the AI Jira administrator.

Next in this topic · 1 of 2Why AI pilots stall at the process, not the model AI pilots stall on unowned processes, scattered data and untracked changes. How to spot the gap and what to fix first.Read next →