Is My Business Ready for AI? Signs to Read First

A business is ready for AI when one named job can move to a tool without anyone inventing the process along the way. Ready is not a mood, and it is not a purchase. It describes the data, the process, the people, and the decision around that one job. If any one of those four is still a shrug, then the job is not ready yet.

Cited in this note

13%

The Readiness Check scores this as four bands. Early is 0–39, Forming is 40–59, Ready is 60–79, and Running is 80–100.¹ Each sentence on the check describes a state of the business. The check does not diagnose anything, and we like that limit. A diagnosis invites shame, but a state invites the next piece of work. Early is allowed to be the true state. Plenty of good businesses are Early on the job they just tried to automate, and they are fine at the rest of their work.

The Early band and the Ready band spell out those states in full sentences. The assessment note compares the public tools side by side. Read the band pages when you want the exact wording. Read this note when you want to know which signs to trust before a vendor tells you what the signs are.

What a large-firm index is counting

Cisco’s 2025 AI Readiness Index surveyed 8,000 senior leaders at organizations with more than 500 employees, across 30 markets and 26 industries.² The firms it called pacesetters were near 13% of that sample, and they were 4 times more likely to move a pilot into production. That sample is not a census of small businesses. So a small firm can be Ready on our check and still look nothing like a pacesetter.

We say that last part slowly, because owners often miss it. Ready on our check means the named job can carry a tool. A pacesetter in Cisco’s index is a large organization in a sample of senior leaders, and the 13% describes those organizations. It is not a report card for a shop, a clinic, or a trades desk. Suppose a salesperson says you should buy now to catch up with the pacesetters. Then they have glued a large-firm statistic to a small-firm fear. The 4 times gap in moving pilots into production is real inside that sample. It does not mean your service writer is late.

Use the index as a picture of a different size of firm. The useful part is that moving a pilot into production follows a written process. The useless move is to decide you are behind a business that staffs a whole crew for its pilot. You have a bay, a desk, and a person who knows the customers. Readiness is whether that person can hand the job to a tool without the job falling apart.

The OECD SME AI adoption tool is part of 2026 Group of Seven work. It is a short self-check for smaller firms.³ It takes about five minutes, keeps the answers in the browser, and does not publish an adoption rate. We’d use it as a second set of questions and not as a national share. A second set helps because our own questions can go stale in our own mouths. Someone else’s short check, with no stored data and no pretend rate, is a good way to see whether you dodged a theme. It will not tell you whether small firms in your country are ready. If someone finishes it and announces a countrywide rate, they have added a sentence the tool refused to say.

We look for four signs that a job can move to a tool. The one place the facts live has a name, the steps are written down, a person owns the output, and someone can stop the experiment. We also look for four signs that it cannot move. The knowledge sits in one head, the exceptions are unwritten, nobody may send the output to a customer, and success has no definition.

The signs are easy to see when the person who does the job is in the room. “Named” means you can point at one record, and the other records lose the argument. “Written” means a new person could follow the exceptions and not only the sunny path. “Owns the output” means a human is accountable for what a customer receives. “Can stop” means someone may halt a bad send without asking the vendor.

The bad signs are the same facts turned around. Knowledge in one head feels like loyalty to a great employee, but it also means the job stops when that person is out. Unwritten exceptions feel like flexibility, but they are where a tool will start to improvise. Nobody being allowed to send feels like caution, until you notice that someone is already sending from a personal account. And success with no definition is how a pilot turns into a subscription.

A small firm can show the good signs on one job and the bad signs on another. That is normal, and it is why “is my business ready?” is the wrong question. The honest question is which job. The check’s bands describe the state you are in, based on how you answered, and they are not a tattoo on the business. Early and Forming are 0–39 and 40–59, and they are descriptions. Ready and Running are 60–79 and 80–100, and they are descriptions too. None of them means you have bought a product, and none of them means you have failed as an owner.

A worked example

Say we are in an independent auto shop on a slow afternoon. One person writes the service tickets, and another is the lead tech. This is an example we walk through, so we are not reporting their revenue, how often cars come back, or any other number. We did not measure those, and we will not fake them.

The owner asked after a vendor showed a tool that drafts customer updates from the ticket. We split the question and asked which job she meant. The job on offer is the status text. The lead tech finds what is wrong, the customer hears it and says yes or no, and then the lead tech goes ahead or does not.

Walk the good signs. Is there one place the facts live, and does it have a name? The ticket writer says it is the shop system. The lead tech says she writes the real finding on the paper that stays on the dash. The system ticket holds only what the ticket writer had time to type, and he is often still with the last customer. The owner says she trusts whatever the ticket writer texts her. That makes three candidates, so the source is not named, and that is the first bad sign already. The true finding sits on the lead tech’s paper and in her head, which is the “one head” sign, even though the ticket writer has a system. A system that nobody believes is not the one place the facts live.

Are the steps written down? The sunny path is written, and it goes car in, inspect, call, approve, repair, car out. The exceptions are not written. What if the customer does not answer and the bay is needed? What if the lead tech finds a second problem after the yes? What if the ticket writer is off and the other writer does not know the lead tech’s shorthand? Those are unwritten exceptions, and that is the second bad sign.

Does a person own the output? The vendor’s tool would send a text, and the ticket writer assumes he would glance at it. The lead tech says she would not see it until the customer called angry. The owner says she wants to approve “the important ones,” which means there is no rule. So nobody owns the output in a way that survives a busy morning, and that is the third gap.

Can someone stop the experiment? The owner says she could cancel the subscription. That stops the bill, but it does not stop the day’s work. A real stop sounds like this: this text does not go out today because it is wrong. The person who sees that may halt it without a meeting. The shop does not have that person yet, and that is the fourth gap.

So the shop is not ready for this job, and that is not an insult. The owner runs a respected bay. The signs say the status text would invent a process that the ticket writer, the lead tech, and the owner do not share today. If we put the same shop on the check, we would not be surprised to see Early or Forming. We would be surprised, and a little suspicious, if the answers came back Ready while the dash paper still beats the ticket.

What would ready look like, without pretending we watched it happen? The ticket writer and the lead tech agree that the ticket is the record. The dash paper becomes a scratch pad that gets typed in before anyone is texted. The no-answer case and the second-problem case are written where the other writer can find them. The ticket writer owns the text. The lead tech can reject a finding that is not what she saw, and the owner can halt a wrong text the same day. Success means the customer hears the finding the lead tech actually made. That is Ready in the ordinary sense, and it does not look like a pacesetter. It looks like a shop that stopped inventing the process inside the tool.

How we’d run it

First, we refuse the wide question. “Is my business ready?” becomes “is this job ready?” The owner does not get to mean the whole shop. She gets the status text, because that is what the vendor is selling. A yes about a different job is a way of changing the subject.

Second, we read the large-firm index for what it counted, and then we say out loud that the shop is not in that sample. The fear of being behind the pacesetters does not get a vote.

Third, we walk the four good signs and the four bad signs as pairs. Is the source named, or does it sit in one head? Are the steps written, or are the exceptions loose? Does a person own the output, or is nobody allowed to send? Can someone stop the trial, or does success have no definition? The ticket writer, the lead tech, and the owner each answer, and we write down the conflicts instead of smoothing them over. The lead tech’s paper against the ticket writer’s ticket is the conflict that matters.

Fourth, we open the band pages if a label will help. Early and Ready are states with sentences, and they are not prizes. If the signs are bad, we expect Early or Forming. We treat a Ready result as a reason to reread the answers, not as a reason to celebrate. The check does not diagnose, and neither do we.

Fifth, we use the OECD self-check as a second pass. It takes about five minutes, the answers stay in the browser, and there is no adoption rate to misuse. If it shows a theme we skipped, we add that theme. If someone tries to turn it into “small firms are behind,” we stop them, because the tool did not say that.

Sixth, we talk about the vendor only after the signs are honest. If the signs are bad, the meeting ends without a quote. If the signs are good for this one job, we still name who stops the trial before anyone creates a login. Ready includes the brake.

Common mistakes

The mistake we see most is hearing “ready” as a personality. The owner is decisive, the lead tech is skilled, and the ticket writer is careful. None of that answers the signs. A careful writer with two records is not ready for a tool that needs one, because character is not a band.

Another mistake is using Cisco’s 13% as a small-business failure rate. It is the share of pacesetters in a sample of large organizations. Quoting it to scare a shop owner mixes up two different kinds of firm. The sample is not a census of bays and storefronts, and we will not nod along when a pitch says otherwise.

A third mistake is counting the cancel button as the ability to stop. The owner can always call the vendor later. What we want is someone on the floor who can block a bad customer message today. Later is how a wrong text sits in a customer’s phone all afternoon.

We also see owners collect the good signs as adjectives and skip the bad ones. They can name a system, so they ignore that the lead tech does not believe it. Naming a system you do not trust is the one-head problem in a nicer outfit. So look at the bad list on purpose, because it is ruder and it is clearer.

The last mistake is treating the OECD tool’s missing rate as bad news. It does not publish an adoption rate, and that is restraint. Do not fill the silence with a guess and then feel behind the guess.

What we’d do

We would answer the owner with the signs and not with a pep talk. Most shops that ask “are we ready?” are really asking for permission to buy. The honest answer is often that one job is Forming and the rest of the question is too big. We would rather shrink the question than inflate the readiness.

If we were the owner, we would make the ticket the only record the lead tech is willing to have texted. We would write the no-answer case where the other writer can see it. We would also give the lead tech a plain way to kill a message that does not match her finding. Then we would take the check and read the band as a state. If it says Ready, we would still start with one job and a person who can stop it. If it says Early, we would believe the dash paper over the demo.