AI Readiness Assessment: What It Measures and How to Score

An AI readiness assessment asks whether a business can put a new tool to work. Public indexes and checklists answer that question for different groups of people. The groups are easy to mix up. This note compares them. Then it explains what the Readiness Check guide scores for the firm that is actually in the room.

Cited in this note

13%

We get asked for “our number,” as if readiness were a credit score you could borrow from a headline. A figure about leaders at big companies cannot tell a clinic whether the front desk and the clinician use the same record.

What the public tools measure

Cisco put out its third yearly AI Readiness Index on 14 October 2025. It surveyed 8,000 senior leaders at companies with more than 500 employees. Those leaders came from 30 markets and 26 industries.¹ The leaders it calls pacesetters stayed near 13% of the sample for three years. Those companies were 4 times more likely to turn a first trial into real use. They were also 50% more likely to see real value. Of all the people surveyed, 83% plan to use AI agents. Nearly 40% expect agents to work beside staff within a year. That sample is not a small-business survey.

Owners hear the Cisco line on a podcast and wonder if they are behind. Behind what, and compared with whom? They would be behind a company big enough to put a senior leader on a survey. The 13% figure is a fair check for large firms. It is a poor mirror for a practice with a front desk and two treatment rooms.

One part still fits a small firm, and that is the gap between a trial and daily work. The companies that cross that gap have the job written down. The demo was never the hard part. So if you use Cisco at all in a small firm, use it as a warning against mistaking a trial for a habit. Please do not use it as a ranking that you are supposed to join.

Gartner said on 26 February 2025 that 63% of the people it asked do not have the data habits that AI needs. Some of them were also not sure they had those habits.² The same note predicts that through 2026, companies will give up on 60% of AI projects that lack data made ready for AI. We point to that line when someone wants to skip the record and start with a login. The project stops when the data was never ready. That story is not hard to guess.

The OECD tool for small and midsize firms is a short self-check about AI use. It is part of the work of the 2026 French Group of Seven Presidency.³ It takes about five minutes and stores no data. Your answers stay in the browser. As of May 2026 it is a test version, and Group of Seven governments are still checking it. It does not publish an adoption rate. We like it as a set of questions you can answer without making an account. We do not like it as a stand-in for a national share, because it never claims to be one.

AWS posted an AI readiness checklist in September 2025.⁴ It walks through five steps and sorts a business into four stages. The stages run from Not yet ready, through Emerging and Operational, to a last stage about using AI widely. They cover the data, the use case, the tools, the rules, and the numbers you will watch. The page does not cite an adoption percentage. Stages are plain descriptions that you can argue with. A percentage that nobody counted is just a rumor with a percent sign.

Here is a plain comparison of who each tool describes and what you leave with.

• Cisco’s index describes companies with more than 500 employees, so you get a comparison with pacesetters and not a score for one small firm.
• The OECD tool describes the smaller firms in the Group of Seven test, so you get a short self-check that stays in the browser.
• The AWS checklist describes a small or midsize business, so you get five steps and four stage names.
• The Readiness Check describes the business taking it, so you get four scores, a band, and three tips for each section.

The Readiness Check is the one built for the firm in the room. It teaches and points the way, but it does not diagnose. We publish the method on our site.

We draw that line on purpose. A diagnosis says what is wrong with you. A route says where to look next. Owners sometimes want the diagnosis because it sounds more serious, but serious is not the same as useful. If we told a clinic that it had failed readiness, we would be borrowing a doctor’s tone for a business talk. The check describes where you stand and points to a next page. That is the promise we make.

What each band means

The check scores from 0 to 100. The bands are Early (0–39), Forming (40–59), Ready (60–79), and Running (80–100).⁵

• Early: The work that matters still lives in people’s heads, so the first useful step is to write one job down and name who owns it.
• Forming: Parts of the work are written down, so a named owner and one place for the facts can turn that into something a team can run.
• Ready: The business can put a new tool to work, so the open question is which job to start with and who owns it day to day.
• Running: A real job already uses the tools, so the next gain is the job that still waits on one person.

A band tells you where you stand today. It is not a grade. The four scores behind it are data, process, people, and the decision. The check takes five to seven minutes with one question on each screen. It gives three tips for each section.

Early means the know-how is still in people’s heads. So you write one job down and name the owner before you shop. Forming means the writing is only partial. Two pieces are missing. One is a named owner, and the other is one place where the facts live. Without them nobody else can run the job when the usual person is out.

Ready means a new tool can land. The argument left is which job goes first and who lives with it on an ordinary day. Running means a real job already uses the tools. The gain left is the job that still waits on one person. Everyone goes around that job because it is “easier to ask her.”

People treat Running as the compliment and Early as the scolding, but we do not. Early is a clear description, and it is often the honest one. A clinic can be excellent at care and still be Early at the paperwork around recalls. Those are different skills. The band is about whether a tool has anything solid to stand on. It is not a review of the owner’s character.

The next note looks at how surveys count AI in small businesses. That is a different question from whether one firm is ready. A survey can talk about firms in general. It cannot say whether your front desk and your clinician agree on the chart note. We try not to let the wide question answer the narrow one.

A worked example

Say we are in a small physiotherapy clinic after the last patient has gone home. One person runs the front desk. One clinician sees most of the repeat visits. This is an example we walk through, and it is not a clinic we measured.

The owner has been shown a tool that would draft recall messages and sum up visit notes. She wants to know if the practice is “ready” before she signs. We do not start with the tool. We start with the four scores, in plain words, the way the check will later turn them into a band.

Data for the recall job is a mess in the ordinary sense. The front desk lead has a list in the booking screen. The clinician keeps a notebook of people he meant to see again. The owner trusts a spreadsheet she built when the booking screen lost a column. Asked which list is right, they look at each other. That is not a failure of care. But it does mean there is no one place where the facts live. We would not call the data “in place” while three lists can disagree and nobody minds.

Process is only partial. The easy path is spoken aloud. The clinician finishes, the front desk lead books the next visit, and the patient leaves with a card. The odd cases are never spoken at all. What if the patient will not book today? What if the clinician wants a recall in six weeks, but the front desk lead only sees open slots this month? What if the note is not done before the patient reaches the desk? The owner says they “work it out.” That is Forming at best, because a tool cannot follow it without making up the rest.

People are the surprise, as in most rooms. The front desk lead already pastes awkward wording into a personal chat tool so the recall text sounds less cold. The clinician did not know, and neither did the owner. Nobody has said whether those words may go to a patient. The people score is not about how eager everyone is. It is really about who has permission to do what.

The decision is still a blank. The owner says success would be “fewer no-shows.” But she cannot say who would turn the tool off if a message went to the wrong person. Until that answer exists, we will not talk about agents, trials, or pacesetters. Cisco’s picture of large firms is interesting. It has nothing to say about whether the front desk lead and the clinician share a list.

If we scored this room in the check’s language, we would expect Early or Forming. We would say so plainly. The work that matters for this job still lives in heads and side lists. The useful step is to write the recall job down and name who owns it. That is just what the Early line says.

How we’d run it

First, we name one job, and at the clinic it is the recall. It runs from the clinician deciding someone should come back to the patient actually having a time. If the owner says “everything,” we shrink it until it has a clear start and a clear finish.

Second, we pull the public tools apart so nobody borrows the wrong group of people. Cisco describes companies with more than 500 employees. The OECD tool is a short self-check that publishes no rate. AWS offers stage names without an adoption percentage. None of those is a score for this clinic, and we say so before anyone feels behind.

Third, we score data, process, people, and the decision in words. We use absent, partial, in place, or running. We use the last two only if we can point to proof. The front desk lead, the clinician, and the owner each answer on their own. When they disagree, that is the finding. So we write it down and do not average it into a mood.

Fourth, we take the check if they want the band. It takes five to seven minutes, with one question on each screen and three tips for each section. We read Early, Forming, Ready, or Running as a place to stand today. We do not add a pep talk that turns Early into “almost Ready.”

Fifth, we pick one next page and not five projects. If the record is the blank, we fix the list that the front desk lead and the clinician will both believe. If the writing is the blank, we write down what happens for the patient who will not book today. The tool demo can wait until that page exists.

Common mistakes

The mistake we see most is treating Cisco’s pacesetters as a small-business grade. Near 13% of a large-firm sample were 4 times as likely to turn a first trial into real use. That is a fact about those firms. Quoting it at a clinic as “only the top shops are ready” makes up a ranking that the study never made.

Another mistake is averaging the four scores into a feeling. Data can be absent while people are eager, and eagerness is not a score. The band names where you stand. The useful part is the weak spot, and not the average comfort in the room.

A third mistake is reading the OECD tool’s silence as a zero. It does not publish an adoption rate, but that does not show that small firms are unready. It shows a tool that refused to pretend.

We also watch owners jump from a band to a purchase in the same sitting. Ready still asks which job to start with and who owns it day to day. The band did not pick the vendor, so you still have to.

What we’d do

We would score the clinic in front of us and leave Cisco in the footnote where it belongs. In our view a small practice does not need to know whether it is a pacesetter. It needs to know whether the front desk lead’s list and the clinician’s notebook are the same list. Until they are, a recall tool will write confident messages from a record nobody owns.

If we were the owner, we would spend the next few ordinary weeks on one written recall path. It would include the patient who will not book at the desk. We would also ask the front desk lead to stop sending patient wording from a personal account until the owner has said what may go out. Then we would take the check again and read the band as a place to stand and not a prize. If it still says Early, we would believe it.