Free tool · No signup · 2 minutes
AI readiness assessment. Is the workflow you have in mind ready for AI?
Think of one workflow: invoices, customer requests, a monthly report. Answer eight questions about it. You get a score, what to fix first and where to start. Nothing you select leaves your browser.
Your result
0 of 8 answered
Answer the eight questions to see where this workflow stands.
Where it stands
- Frequency
- Definition of done
- Inputs
- Authority
- Checking
- Exceptions
- Cost of a mistake
The AI readiness checklist
The seven things the assessment looks at, and what ready looks like for each. They are what a system needs from the business around it, whatever the model.
- 01
Frequency
Why it matters: A system costs much the same to build whether it runs ten times a month or ten thousand. Volume is what pays it back.
Ready looks like: It comes round every day, or many times a day.
- 02
Definition of done
Why it matters: If people disagree on what good looks like, the system will too, and checking its work turns into an argument.
Ready looks like: It’s written down: a template, a checklist, rules.
- 03
Inputs
Why it matters: A system can only work with what it can read. Information that lives in people’s heads never reaches it.
Ready looks like: Systems you control, in a consistent format.
- 04
Authority
Why it matters: What a system may do alone, and what needs approval, is a business decision. Left undecided, everything waits for someone.
Ready looks like: Rules or limits already exist: amounts, clients, categories.
- 05
Checking
Why it matters: Every result has to be checked before it counts, and the time spent checking comes straight off the time saved.
Ready looks like: Clear checks: totals match, fields are complete, rules pass.
- 06
Exceptions
Why it matters: Every workflow has cases that don’t fit. What matters is whether they land with someone who can decide.
Ready looks like: A named person, with a known route.
- 07
Cost of a mistake
Why it matters: It sets how much a person needs to see before anything leaves the building.
Ready looks like: Quick to spot and cheap to fix.
How the score works: seven questions score from 0 to 3, and the score is their total out of 21, as a percentage. A workflow is ready to build at 70% or more with nothing at zero. The eighth question, where you are today, does not score: it decides where to start.
Questions about AI readiness
What is an AI readiness assessment?
A short check of whether a company, or here one workflow, has what an AI system needs to work in daily use: enough volume to be worth it, a clear definition of done, inputs a system can read, rules for what it may do alone, a way to check results, and someone who owns the exceptions. Most assessments score the whole company. This one scores the workflow you have in mind, because that is where a system succeeds or fails.
Why assess one workflow instead of the whole company?
Because a company-wide maturity score rarely tells you what to do on Monday. Readiness differs from one workflow to the next inside the same company: supplier invoices may be ready while proposals are not. Start with one, get it into daily use, then do the next.
What makes a workflow ready for AI?
Seven things: it comes round often, people agree on what done looks like, its inputs sit somewhere a system can read, someone has decided what the system may do alone, there is a way to check a result, a named person owns the exceptions, and a mistake is cheap to catch. None of them is about the model.
How long does it take, and is it really free?
About two minutes: eight questions, one tap each. The result appears as soon as you answer the last one. There is no signup and no email to give.
Do you store my answers?
No. The assessment runs in your browser and nothing you select is sent to us. Your answers sit in the page address after the # sign, which browsers never send to a server, so you can share the result with a colleague.
What if my workflow is not ready?
That is the common case, and it is worth knowing before paying for a build. The result names what to fix first. Some fixes take a week: writing down what done looks like, naming who owns the exceptions. Others need a closer look at how the work really runs, which is what Diagnose is for.
My workflow is ready. What next?
Put a number on it: the AI ROI calculator shows what the workflow costs today and what a system could give back. Then build it. Build puts a system into production in two to four weeks, and the deployment milestone only invoices if it ships.
Does a high score mean AI can do the whole job?
No. It means a system can do the routine part under rules you set, and that a person sees what matters: the exceptions, the risky actions, the results that fail a check. Below the line is the work a system can generate. Above it are the objective, the authority to act, the evidence and the exceptions, and those stay with your people.
Further reading
- AI ROI calculatorReady? Put a number on it: what the workflow costs today and what AI could give back.
- Which agent actions actually need your approval?We ran two rules across 28 real action classes. Reversibility alone missed eight boundaries, including sensitive reads, deployments and permission changes.
- Everything That Failed Was Written DownThe UK's AI evaluation institute published an incident report on its own tests. Every control that failed existed on paper. The one that held was a person.
- An Agent Does a Task. A Loop Runs the Work.Most AI pilots produce an answer. A loop observes what happened, decides what to do next, and continues until the work is finished or judgment is required.
- Work: Diagnose, Build, EmbedThe three ways in, and what each one leaves you with.