How to check whether a process is ready to be enhanced with AI
A process is ready to be enhanced with AI when five things are true: it is documented as it really runs, it has an owner who keeps it current, its decision rules and exceptions are explicit, the data each step needs is available with clear access, and you know how it performs today. Check those before you choose an AI tool. Where the answer is no, the first investment is capturing the knowledge — AI on an undocumented process repeats the confusion faster.
Before you start
Prerequisites: one process you are considering for AI, its owner (or whoever acts as one), one or two people who do the work, and a recent case to test your answers against. Our full AI-readiness checklist goes further than this guide — ask for it through the button below.
Use this guide before you buy or build an AI solution for a process, when a pilot delivers answers people do not trust, or when you need to choose which process to try AI on first. The same checks help to compare candidate processes side by side.
The steps
Step 1 — Check the documentation against a real case
Take a recent case and follow it through whatever documentation exists. Note every step, workaround or exception the documentation does not show.
Pitfall: Accepting that a procedure exists. The question is whether it matches the work.
Step 2 — Check the ownership
Is there one person accountable for the process who keeps its knowledge current? If not, name one before anything else.
Pitfall: Assuming “the department” owns it. AI projects on ownerless processes stall at the first disagreement.
Step 3 — Check the decision rules and exceptions
List the decisions in the process. For each one: is the rule written down, are the exceptions known, and is it clear who decides when in doubt?
Pitfall: Expecting the AI to learn the rules from examples. It learns the common case and misses the exceptions that matter.
Step 4 — Check the data and the access
For each step: which information is needed, where it lives, whether it is complete, and who may see it.
Pitfall: Discovering access problems after the pilot has started.
Step 5 — Measure the baseline
Pick two or three measures the owner already trusts — lead time, rework, questions to the expert — and record today's level.
Pitfall: No baseline. Without it, nobody can tell whether AI helped.
Step 6 — Decide where AI helps and where people decide
Write down which tasks AI may assist with, which it may prepare for review, and which stay with people.
Pitfall: Letting the tool define the boundary.
An example
Take a distributor that wants AI to draft answers to customer questions about delivery terms. The team runs the checks on that one process.
The documentation dates from before a change of carriers, so it no longer matches the work. Nobody owns the delivery terms. The exceptions for key accounts live with two account managers. The data, on the other hand, is in good shape: the order system holds what each answer needs, and access is clear. For a baseline, the team records how long a first answer takes today and how often a colleague has to correct it.
The decision follows from the gaps. First, name an owner and capture the key-account exceptions from the two account managers. Then let AI draft answers for review by a person, and keep key accounts with people until the drafts prove reliable against the baseline.
Where MentX comes in
MentX captures how your company really works: the knowledge of your key people and the processes they run. We offer both a hands-on service approach — mapping your processes and capturing the knowledge behind them, from meetings, workshops and a guided intake — and our own personality-aware knowledge base, with full source attribution, in your own Knowledge Space. We build on the tools you already use and add best-practice solutions where they fit.
When the checks show gaps, capture your operations closes them: owners, decision rules, exceptions and data, with their sources, in one process model. Then operational excellence puts AI to work where it helps. For the why behind it, read why AI needs your business context.
Checklist
- Documented as it really runs
- One owner who keeps it current
- Decision rules and exceptions explicit, with a decider
- Data available, access clear
- A baseline the owner trusts
- A written boundary between AI and people
Questions people ask
What is in the full AI-readiness checklist?
It goes deeper than the five checks in this guide: ten criteria, including how much a process varies, whether every rule can be traced to a source, and who approved each change — plus a simple ladder to see where a process stands today. Ask for it through the LinkedIn button below.
Does a process need to be perfect before we use AI?
No. It needs to be known, owned and honest about its exceptions. Gaps are fine as long as they are visible and someone is closing them.
