How to run an AI readiness assessment with the people who do the work
Most AI plans start with a tool and look for a place to put it. The ones that last start with the work: what people do every week, where it waits, and which parts they would gladly give away. Here is a plain way to find that out.
01Pick one area of work, and say what you want to find out
Start narrow. Order processing, month-end, client onboarding: one piece of work that passes through several people is enough for a first pass. Write down the question you want answered, such as “where could AI take work off our plates in order processing?”, and what you would do with the answer.
02Hear from everyone who touches it
A workshop hears from the people with time to attend. A survey hears from the people with opinions. The work itself is often done by the people in neither. List everyone who starts, passes on, checks or finishes the work, across every team it touches, and plan to hear from each of them one to one.
03Ask about the last real case, not opinions
“What would you automate?” gets guesses. “Walk me through the last order you keyed in” gets the steps, the systems and the waits. Ask about the last time, the exception, and what happened while something waited. Ask one question at a time, and follow up on the example the person gives.
QUESTIONS THAT WORK
- “Walk me through the last time you copied something from one system into another.”
- “How long did the last batch take, start to finish?”
- “What's one bit of that you'd happily never do again?”
04Listen for repeated, digital work
The best first candidates have the same signs: they come round every day or every order, they move information from one system to another, someone checks the copy, and the work waits on another team while it happens. Note how often each one comes round, how long it takes and how many hands it passes through, in the person's own estimate.
05Sort what to hand over, what to fix, and what to keep
Not everything slow should be automated. Some steps wait because of a rule or a handoff that a tool will not fix. Some work people want to keep, because it needs judgment or a relationship. Put each piece of work in one of three lists: hand to a machine, fix the process first, or keep with people. Let the people who do it have a say in the third list.
06Choose one first step, then ask again
Pick the piece of work that most people raised, that comes round most often, and that nobody wants to keep. Hand it over with a person checking, and agree who owns it. A few weeks later, ask the same people what changed and what broke. That is how you know it worked, and what to do next.
WHERE COMPLEXIN FITS
Complexin runs this for you: it drafts the questions from your goal, interviews everyone one to one in their own time, and ranks what people would hand over beside what they want to keep, with the interviews behind every finding.
See ai adoption →