Air Canada once tried to blame its own chatbot in front of a tribunal.
The bot had told a grieving passenger, Jake Moffatt, to book his flight to the funeral and claim the bereavement discount afterward. Air Canada didn’t offer that.
When he asked for the money, the airline’s lawyers argued the chatbot was “a separate legal entity that is responsible for its own actions.” That’s the corporate version of the dog eating your homework.
The tribunal ruled against them on Valentine’s Day 2024. Air Canada paid C$812.02.
An AI readiness assessment exists to find the gap that chatbot fell through, before a customer finds it for you. The bot did what every chatbot does when nobody has written down what it isn’t allowed to say.
The one on this page is 24 questions my team works through with every consulting client before we build a thing, scored out of 72.
Air Canada fails at least three of them. Keep count as we go.
What is an AI readiness assessment
An AI readiness assessment is a scored check of whether your business can hand real work to AI without it going sideways a few weeks after launch. You finish with a number out of 72 and a short list of the gaps worth fixing first.
Ours is the self-serve version of the operations audit that opens every build we do, cut down to the questions that decide most outcomes.
They fall into six areas: direction, data and content, processes, people, tools, and measurement and risk.
Is my business ready for AI
If most of what your business knows lives in your head, it isn’t ready yet. That’s the short answer to “is my business ready for AI.”
I learned it on myself. When I started First Movers I was working 90-hour weeks and selling $50,000 integrations in the first month.
Everything I knew about doing the work sat in one place: my head. A machine can’t read that place.
I paid for that the hard way: 18 months of Claude projects that wrote like a polite stranger, because the brain wasn’t written down yet.
So I started writing. I kept going until it came to over 100,000 words of my own thinking. We call it Voice DNA now.
It’s the reason my AI clone has presented over 750 videos I never filmed, for a YouTube channel with 300,000+ subscribers as of September 2026. The clone books out at $4,000 per 10-minute appearance.
At 14 I was fixing computers for $40 an hour and thought I’d made it.
How the 24-question AI readiness assessment is scored
Give each question a score from 0 to 3.
Zero is no. One is “sort of,” which usually means it happened once or it lives in somebody’s head. Two is mostly there, with gaps you could name if pushed.
Three is written down and working this week, so be a little stingy with those.
Plans for next quarter score a 0 today, since everybody has a plan for next quarter. Six areas with four questions each gives you 72 points.
| Area | Questions | What 12 out of 12 looks like |
|---|---|---|
| Direction | 1 to 4 | One sentence on what AI is for, a plan with an owner, one chosen use case, leaders using the tools weekly |
| Data and content | 5 to 8 | Knowledge in text a model can read, a map of customer data, 50+ work examples in one place, content checked |
| Processes | 9 to 12 | Top five tasks documented, biggest time sink named in hours, clear inputs and outputs, one process handed off |
| People | 13 to 16 | One named owner, the team told what changes, structured training completed, a separate AI budget line |
| Tools | 17 to 20 | Core systems reachable by automation, paid business accounts, one automation running unattended, every subscription listed |
| Measurement and risk | 21 to 24 | Hourly cost of the target work known, a dated baseline, a written data rule, a weekly check |
Grab a pen. The AI readiness checklist below takes each question in turn, with what a low and a high score look like in a real business.
Direction (questions 1 to 4)
A team with no stated goal will happily automate the wrong thing for a whole quarter and call it progress. These four come first for that reason.
1. Can you say in one sentence what AI is for in your business
Write the sentence down. “Cut proposal time from six hours to one” gets a 3.
“Stay competitive” gets a 0, since nobody in history has built a workflow out of a mood.
If you ask three people on your leadership team and get three different sentences, that’s a 1. It’s also the most common answer.
2. Is there a written AI plan with an owner and a date
A real plan fits on a page. It names an owner, a first milestone inside 90 days and a budget.
The slide from last year’s offsite with the rocket on it doesn’t count, however nice the rocket was. Give that one a 1.
3. Have you picked your first use case and said no to the others
Pick one workflow and park the others in writing for later. That’s a 3.
Five pilots at once looks great in the Monday update.
Five half-finished experiments still teach you very little about whether the sixth will hold up, so that’s a 1.
4. Does leadership use AI tools themselves every week
If the people approving the AI budget haven’t used Claude or ChatGPT on their own work this week, score a 0. Teams copy whatever the boss does.
Data and content (questions 5 to 8)
Most founders I talk to swear they don’t have enough material to train a model.
Then we go looking: old proposals, a few hundred sent emails, recorded sales calls and a shared drive folder called something like “misc” that nobody has opened since 2022.
At the Clone Mastermind we pull all of it together on day one, because a model can only learn from what it can find.
5. Is your best knowledge written down somewhere a model could read it
Processes, pricing logic, positioning and past work, all sitting as text a model can read, gets you a 3. Videos count once they’re transcribed.
Knowledge that lives only in the founder’s head gets a 0, which is also the answer I hear most on strategy calls.
I tested what a 0 costs on my own YouTube channel.
In spring 2026 I ran it on AI with no training data, on purpose, and monthly organic views fell from about 2 million toward half that. I left the failed videos up as evidence.
Then the training data landed. Same channel, same audience, same week: 3.2K views in 3 weeks without trained skills, 25.2K views in 3 days with them.
AI is pretty horrible without a brain. You have to have a brain.
6. Do you know where your customer data lives and who can access it
Say the systems out loud. CRM, email, invoicing, the support inbox and the spreadsheet everyone promises to retire next year.
Full marks means each one has an owner and a rule for what a model is allowed to touch. If you had to go ask around, you already know it’s a 1.
7. Do you have at least 50 examples of your own work in one place
Fifty proposals, emails, scripts or posts in one folder is about the minimum for teaching a model what good looks like in your business.
Mine had 100,000 words to work from, which is overkill for most companies and about right for someone who has written 10+ books. Fifty scattered across five drives is a 1.
8. Has anyone checked your content for things a model shouldn’t repeat
Old prices, former clients and promises you quietly stopped making in 2023 all need pulling out before a model learns from them.
Air Canada’s bot was repeating a refund rule the airline didn’t have. Going by the tribunal’s decision, nobody had caught it.
That’s a 0. Strike one for the airline.
Processes (questions 9 to 12)
Automation copies whatever process you give it. Hand it a habit and it copies the habit, loose ends included.
The fix is in my guide to AI workflow automation, and it starts with a pen.
9. Are your top five recurring tasks documented step by step
Your team’s five most repeated jobs, written as steps a new hire could follow without tapping anyone on the shoulder.
All five is a 3. Three of the five is a 2.
10. Can you name the one workflow that eats the most hours each week
Mine was content. I was making almost everything myself at close to 90 hours a week, until the clone and a handful of automations brought it down to about 3.
You want that kind of number for your own business, the workflow and the hours.
If you’d have to guess, give yourself a 1 and time the work for two weeks.
11. Do your tasks have clear inputs and outputs
“Inquiry form comes in, qualified lead lands in the CRM with a score” has a clear start and a clear finish. That’s a 3.
“We sort of deal with inquiries as they arrive” scores a 0 and, I’d bet, loses a few leads a month.
12. Has any process been handed off to a new hire in the past year
A new hire learning a job from your documents faces the same test an AI system does, minus the coffee breaks.
If someone picked up a process that way in the last 12 months, that’s a 3. If every handover took a month of shadowing, it’s a 1.
People (questions 13 to 16)
We were two people when First Movers hit its first big revenue milestone. We’re about 15 now.
Every automation that survived the growth in between had somebody on the team whose job it was to keep it running. The ones that quietly died had a group chat.
13. Does one named person own the AI rollout
Name one person, ideally not the founder, since a founder who is the single point of failure is the exact problem First Movers was set up to fix, with hours in their week set aside for it.
The founder doing it at 11pm is a 1. “The team” is a 0.
We run every AI employee we build on the same rule, with a named human who owns the result.
14. Has the team been told what AI will change about their jobs
Each affected role should hear, in writing, which tasks are moving to a system and what they’ll do with the time back.
A passing mention at the all-hands is a 1. Silence is a 0.
Quiet resistance usually starts right there.
15. Has anyone completed structured AI training in the past six months
Watching YouTube doesn’t count, even my channel. You want a course with exercises, finished by at least one person who now uses it every week.
AI Labs has 80+ courses and live workshops twice a week.
16. Is there a budget line for AI, separate from general software
An AI line in the budget with a name beside it is a 3. ChatGPT on somebody’s personal credit card is a 1.
Finance finds those in December.
Tools (questions 17 to 20)
Buying software is the part everyone’s already good at. This section scores whether your tools can talk to each other.
17. Do your core systems have an API or a Zapier or Make connection
Check the CRM, email, calendar, invoicing and project tool. If an automation platform can reach all five, you’ve got a 3.
Two or fewer is a 0. Not knowing is a 1 and a 10-minute job for whoever manages your logins.
18. Is your team on paid business accounts instead of personal free plans
Business or team plans with admin control score a 3. A mix of free personal accounts scores a 1.
The Claude plan our consulting clients run their finished systems on costs about $100 a month, which makes this the cheapest point on the page.
19. Do you already have one automation running without a human touching it
Any automation counts, even a small one. Form to CRM, meeting notes to a summary, invoice reminders on day 30.
One that has run for a month without anyone checking on it is a 3. One that runs while somebody hovers over it is a 2.
20. Can you list every AI tool the business pays for right now
Shadow spend adds up quietly. If your finance lead could find three AI subscriptions you’d forgotten about, you’re a 1.
Measurement and risk (questions 21 to 24)
Air Canada would like this section back. Two of these four are about what happens after launch, when the AI is talking to real customers and nobody is reading over its shoulder.
21. Do you know the hourly cost of the work you want to automate
Loaded cost per hour, times the hours per week, for the workflow you named in question 10.
Most of our consulting clients see payback within 3 to 6 months. I walk through the math in AI consulting ROI.
Without the hourly figure you can’t tell whether you’ll be one of them.
22. Is there a baseline measurement from before AI
Hours, turnaround time, cost per lead and output per week, written down with a date before anything changes. Measuring after launch is a 1, because there’s nothing left to compare against.
23. Is there a written rule for what AI can and can’t do with customers
One page, read by the team, covering what can go into a model, what can’t, and who signs off anything a customer will see.
Air Canada’s bot was quoting policy to the public. Going by the tribunal’s decision, nothing stopped it quoting the wrong one.
Strike two.
24. Would you know within a week if an automation broke
A weekly check or an alert with someone’s name on it is a 3. Finding out from a customer is a 0.
Finding out from a tribunal is, I’d argue, a negative number, though the scale stops at 0.
Strike three.
The AI maturity model behind the score
Add up your 24 answers. You’ll land in one of four bands, the same AI maturity model my team uses in the operations audit that opens every build.
Treat the band as your next step, since nobody has ever improved anything by staring at a grade.
0 to 24, Watching
Don’t buy anything yet.
Write down the workflow from question 10 step by step. Then teach one person on your team to run it with AI, start to finish.
The AI readiness checklist above doubles as your to-do list. AI Labs is built for this stage.
25 to 42, Experimenting
The tools are in, a couple of people are excited and nothing connects yet. Most pilots die quietly in this band.
Fix questions 9 to 12 first, then get one automation running for a month. My AI marketing automation post shows the first loop I’d build.
43 to 58, Building
Direction is clear and the work is written down. Your gaps sit in tools or measurement, which are the easiest ones to fix with money.
You’re ready for a systems build, done with you or done for you.
Week one of a consulting engagement audits your tools, goals and content before the build plan gets locked.
59 to 72, Running
Your AI does its job without you watching every step.
The next job is keeping it that way next quarter, when the person who built it has moved on to something else. Push your two lowest areas to 12 and start looking at agents.
This band is where the first mover advantage lives, since you’ll be building on the next tool while competitors are still reading about it.
Your first AI result costs nothing
Start with your lowest area. A 50 with a 2 in measurement is shakier than a 44 spread evenly, because the build will ship and nobody will be able to prove it did anything.
As for Air Canada, a tribunal ruling can’t score all 24 questions, so I won’t pretend it can.
Three strikes on questions 8, 23 and 24 is 9 points gone before anyone looks at the tools. The tools were probably fine.
The airline got its lesson for C$812.02, plus a few years as the favorite example in posts like this one.
Write your score down with today’s date and run the assessment again in 90 days. Most questions move a point or two in two weeks through work that costs nothing, like naming an owner or writing down your top five tasks.
That band change is your first measurable AI result.
It arrives before you’ve bought a single tool.
Rank Math note. Visible FAQ block starts here.
Questions people ask about the AI readiness assessment
How long does the AI readiness assessment take
About 10 minutes if you answer from what exists in your business today. Questions 10 and 21 take longer if you stop to check the real numbers, which is worth doing. Question 10 asks which workflow eats the most hours each week, and question 21 asks what an hour of that work costs you. Those two figures are the baseline your first AI result gets measured against.
What is a good AI readiness assessment score
43 or above out of 72. From there a systems build tends to hold, because direction and process are already in place and your remaining gaps sit in tools or measurement, the easiest ones to fix with money. Score under 25 and your business isn’t ready for AI yet. Spend on training and documentation first, then run the assessment again in 90 days.
Is this AI readiness checklist meant for small businesses
Yes. This AI readiness checklist is written for businesses doing roughly $250K to $10M a year, which is who we build for. The questions are sized for a company where the founder still knows everyone’s name. They weigh whether your onboarding steps are written down, whether one named person owns the rollout, and whether you’d know within a week if an automation broke.
How is an AI maturity model different from a readiness assessment
The AI readiness assessment gives you a score out of 72. The AI maturity model turns that score into one of four stages, Watching, Experimenting, Building or Running, with a next step attached to each. The stage tells you what to do on Monday: write down one workflow if you’re Watching, fix your processes if you’re Experimenting, start a systems build if you’re Building and look at agents if you’re Running.
Can I go through my result with someone
Yes. The strategy call on the First Movers consulting page is free, and it works best if you bring your AI readiness assessment score and your lowest area. We use the call to map a build together, starting from that lowest area. It’s most useful for businesses scoring 43 or higher, since that’s where a systems build tends to hold.
Your lowest area is where your first real AI result lives.
Take it into AI Labs and start there.
