The AI reality gap: did any of it land?

Start here

Guess before you read on.

Think of your own company for a moment. If somebody stopped a hundred of your people in the corridor today and asked them to explain, in their own words, what your company is trying to do with AI, how many could give an answer you would recognise? Commit to a number before you scroll, because the real one lands differently once you have.

Out of every hundred employees and managers, how many understand their company's AI vision?

50

The error runs one way in almost every room I have asked this in. We guess high, because the version we hear is the version we said, and we have said it more times than anybody else has heard it.

You have seen this

Which of these was your last leadership meeting?

Tick the ones you recognise. Nobody sees this but you, and the honest count is more useful than the flattering one.

Somebody presented an AI roadmap and nobody in the room disagreed with any of it
The rollout was described as going well, and the evidence was that licences had been issued
You have said the phrase “we have communicated this” about something you said once, five months ago
The only people in the meeting who use AI daily were the ones presenting
Nobody could name a single team that had changed how it works because of any of it

Three or more is normal, and it is also the whole finding. Every one of those is a room agreeing with itself, and none of them is evidence of anything happening two floors down.

The AI reality gap is a company where the leadership team and everybody else are describing different years. Fifty-nine in a hundred executives say they communicate a clear AI vision. Eight in a hundred employees say they receive one.1

Two pictures, one company

Betterworks surveyed 2,387 people this year across employees, managers, directors and HR. Fewer than 16% of employees and managers understand their company's AI vision.1
59% of executives believe they communicate that vision clearly. 8% of employees agree.1
92% of executives are comfortable using AI to get work done. Among employees it is 51%.1
58% of employees say AI has not improved their day-to-day work at all, against 8% of executives.1
Team goals that mention AI: 45% of executives say theirs do, and 6% of individual contributors.1 That last number is where the gap actually lives.

You will know this one if

You presented a plan that everybody nodded at and nothing moved
Your board thinks you are further along than your floor does
Somebody asked a question that showed they had missed the whole thing
You suspect the good news reaching you has been through three filters

What you take away

Ten numbers that show how wide the gap usually is
A three-person test you can run this week without telling anyone
The research, with the sample and what it cannot prove
Three fixes that need no new meeting on anybody's calendar

Defining it

You are being reported to, and every report is honest.

Everybody in the chain between you and the work does something reasonable and slightly generous. A team lead rounds a partial rollout up to a rollout. A director summarises four teams and keeps the one that went well, because that is the one worth your time. By the fourth summary the pilot is a programme.

Meanwhile you personally have the best possible experience of AI in the company. You have the licences, you have interesting problems to point at, and you have the freedom to spend forty minutes trying something. Ninety-two in a hundred executives say they are comfortable using AI to get work done. Among employees that is fifty-one.1 Your own week is genuine evidence, and it is evidence about the most favourable position in the building.

Your own experience of AI is real, and it is the best one in the company. That is exactly what makes it a bad sample. Paul Musters

The other half is what happens to a message on the way down. You said it in an all-hands and in a strategy deck. It arrived as a slide somebody scrolled past on a Thursday. Employees see AI in the strategy, and they do not see it in their goals, their workflows or their next conversation about their career. So it stays abstract, and abstract things get filed as leadership language rather than as instructions.

There is one number in the Betterworks data that shows this cleanly. Asked whether team goals frequently include expectations to use AI: 45% of executives said yes, 30% of HR, 18% of managers, and 6% of individual contributors.1 Read that ladder from top to bottom. It is the same question about the same companies, and the answer thins out at every step down.

The numbers

The same question, asked twice.

Betterworks put the same questions to executives and to the people who work for them, in its 2026 State of Performance Enablement research. Every pair below comes from that one survey of 2,387 respondents. Read each row across the middle.1

ExecutivesEmployees

59%

Say a clear AI vision is communicated

8%

64%

Say employees understand AI's strategic value

15%

92%

Comfortable using AI to get work done

51%

8%

Say AI has not improved their work at all

58%

Four questions from one survey of 2,387 respondents, split by role. The last row runs the other way, which is the one worth sitting with.1

The last row is the one that should stop a leadership team. Fifty-eight in a hundred employees say AI has not improved their day-to-day work at all. Among executives that is eight. Those two groups are describing the same rollout, the same tools and the same year.

7×

more likely executives are than employees to say team goals frequently include using AI

1

2.7×

more likely executives are to say AI makes work easier, and 2.1 times more likely to credit it for creativity

1

6×

more likely executives are to believe performance reviews and goal-setting have kept pace with AI-driven work

1

And the training is not the missing piece

This is the part that saves you a budget line. Eighty-five in a hundred HR leaders in the same survey believe employees get adequate AI training, and eighty-eight report moderate to high AI fluency across their organisation.1 The exposure is there. What is missing is time, context and a clear answer to where this applies in the work somebody actually has on Thursday.

Betterworks puts it plainly: exposure does not equal application.1 Another course will not close a gap that was never about knowing what a prompt is.

One more number, and it comes with a caveat, because Betterworks quotes it from a SHRM report rather than from its own survey. Twenty-six in a hundred HR leaders say they have been involved in managing the change and adoption side of AI, while two thirds believe they should be, and seventeen in a hundred say their organisation's AI implementation has been successful.2 Whether or not HR is the right owner in your company, somebody has to be, and in most companies the honest answer is that nobody is.

AI doesn't fail because people fear it, but because leaders stop at the vision deck. Jamie Aitken, VP of HR Transformation at Betterworks1

One caution about all of it. These are self-reported answers from one survey, weighted towards office roles, and executives and employees may partly be answering slightly different questions in their heads about what a clear vision means. The direction of the gap is consistent across ten separate items, which is what makes it worth acting on. The exact size of any one of them is not the point.

The mechanism

Both groups are honest. They are describing different companies.

Saying it once feels like saying it

You worked on that message for two weeks. By the time you delivered it you had heard it forty times, so it felt worn out on the way out of your mouth. The person hearing it heard it once, in a fifty-minute meeting, between two things that affected their week more. Then it never came back. What feels like repetition at the top is a single exposure at the bottom.

Good news travels up faster than bad news

Nobody in the chain is hiding anything. Each person is summarising honestly and choosing the parts worth your attention, and the parts worth your attention are the ones that are working. Four teams get compressed into the best team. Do that twice and a promising pilot has become a company-wide capability without anybody saying a false sentence.

Four summaries later, the pilot has become a programme, and nobody lied once. Paul Musters

It never reached the one place people actually look

People read their goals, their review and their next conversation with their manager. If AI is in the strategy and in none of those three, it is a topic rather than an expectation. Forty-five in a hundred executives say team goals include using AI. Six in a hundred individual contributors say the same.1 That distance is not a communication problem, it is where the message stopped being connected to anything.

Now the level underneath it

Six questions and no name attached, three minutes. Where the company sits, what that costs, and what the next level changes.

Do the Culture Level scan

The test

Three people, separately, this week.

This costs you twenty minutes and it is the fastest honest measurement available to you. Pick three people who do not report to each other and are not on the leadership team. Ask them the same three questions, one at a time, and write down what they say rather than what you think they meant.

1

What are we trying to do with AI?

In their words, not yours. You are listening for whether the sentence they give you resembles anything you have said. A blank, a guess or a joke all count as the same answer.

2

What does that mean for your work this quarter?

This is where most companies fail. If the first answer was fine and this one is empty, your message reached their ears and never reached their goals.

3

Who would you ask if you got stuck with it?

A name means there is a route. “I would probably just work it out” means there is not, and it also tells you why nothing is spreading between teams.

Then compare the three answers side by side. Most leadership teams expect variation and find something worse: three answers that are consistent with each other and have almost nothing to do with the plan. That is a much better problem to have found in an afternoon than in a year.

Do not do this in a survey. A survey gives you a number and takes six weeks, and the thing you need is the exact wording somebody uses when they are guessing.

In the five levels

Three answers show you the gap. They do not show you the shape of it.

The three conversations tell you that the distance exists. What they cannot tell you is where in the company it is widest, and that matters, because the fix in a team that is already ahead is the opposite of the fix in a team that has not started.

An Operating Profile in use. Personality type and AI level in one profile, with the agents that fit it.

So we look at one person at a time. How they think and work, and how far along they are with AI. Fifteen minutes each, everyone does their own, and what comes back is a map of where your company actually is rather than where the reporting line says it is. In almost every company that map surprises the leadership team twice: once at the top of the range and once at the bottom.

It also gives you something better than a plan to communicate. Instead of one message for everybody, each person gets a picture of their own position and a next step that fits it, which is the version people can act on the same week.

About emaho

emaho measures one Operating Profile per person: personality type and AI level in a single profile. On that we build a personal set of AI agents that fit how that person works, inside the tools they already use. Fifteen minutes to complete, first profile free, built for companies between 20 and 500 people.

See your own profile, free Find your AI Culture Level

Fifteen minutes per person. No credit card, no strings.

What actually helps

Nothing new on the calendar.

Every fix for this fits inside a meeting you already have. That matters, because a gap this wide usually gets answered with a programme, and a programme is another thing people hear about rather than something that changes their Thursday.

In the meeting where goals get set

Put one line about AI into a goal that already exists, per team, and make it about the work rather than the tool. Name the specific thing this team will try to do differently, and what would count as it having worked. Six in a hundred individual contributors currently have anything like this.1 Moving that number is the single most direct thing on this page.

In the one-to-one you already hold

Ask the three questions from the test, once a quarter, and let the manager keep the answers. That turns a message into a conversation with a name attached, and it puts the person who actually influences somebody's week in charge of it.

In the all-hands you were going to run anyway

Replace the update with fifteen minutes of one team showing what they changed and what broke. A real example from someone at their level does more than any restatement of the vision, because it answers the question people are actually asking, which is whether this is real here.

One thing to stop while you are at it. Stop treating “we have communicated this” as done. You have said it. The measurement of whether it arrived is the three answers you write down, and that measurement takes twenty minutes and is available to you today.

What did the three of them say?

If you run the three questions, send me the three answers to the first one. I have collected a lot of these and the pattern in the wording usually says more than the content. One message back, written by me.

Message me on WhatsApp

The cost of waiting

You keep reporting a transformation nobody is living in.

The cost you can see arrives as a decision made on the wrong picture. You commit budget, headcount or a deadline on the basis that the company is somewhere it is not, and you find out at the point where the plan meets the actual capability of the actual teams. That correction is expensive and it is public.

Underneath sits something slower. Fifty-eight in a hundred employees say AI has not improved their work at all.1 Every month that stays true, the word transformation loses a little more meaning inside your building, and the next initiative you announce starts from further back than this one did.

And there is the part that lands on your managers. They are the ones standing between a leadership team that believes the change has happened and a floor that has not felt it, and they are asked to explain both to each other without being given anything new to say. That is a quiet, grinding job and it wears out exactly the people you least want to lose.

Asked and answered

Questions about the gap between the story and the floor

What is the AI reality gap?

It is the distance between the AI transformation a leadership team describes and the one employees actually experience. In Betterworks' 2026 research among 2,387 respondents, 59% of executives said they communicate a clear AI vision while 8% of employees agreed they receive one. Both groups are answering honestly about the same companies.

Do executives overestimate their company's AI progress?

Consistently, across every item measured. Executives were 7 times more likely than employees to say team goals frequently include using AI, 2.7 times more likely to say AI makes work easier, 2.1 times more likely to credit it for creativity and 6 times more likely to believe performance reviews have kept pace with AI-driven work.

How many employees understand their company's AI strategy?

Fewer than one in six. In the 2026 Betterworks State of Performance Enablement report, under 16% of employees and managers said they understand their company's AI vision. Only a quarter of managers said the vision is clear, and a third said the strategic value is clear.

Why do employees say AI has not changed their work?

Because for most of them it has not. 58% of employees in the 2026 Betterworks survey said AI has not improved their day-to-day work at all, against 8% of executives. The tools exist and the expectation to use them has not reached their goals, their workflows or their reviews, so nothing about their week is different.

How do you find out where your company really is with AI?

Measure it rather than ask the room. emaho places companies on five AI Culture Levels, from everyone quietly doing their own thing to AI being part of how the place runs. Leadership teams almost always place themselves a level too high, which is the same error the 59-versus-8 gap shows. The scan is six questions, three minutes and anonymous, and it asks about behaviour rather than intention.

Is more AI training the answer to low adoption?

Usually not. In the same survey, 85% of HR leaders believed employees receive adequate AI training and 88% reported moderate to high AI fluency across their organisation. Betterworks summarises the problem as exposure not equalling application: people know what the tools are and lack the time, context and specific use cases to apply them to the work in front of them.

How do you find out whether your AI message actually reached people?

Ask three people who do not report to each other, separately, this week. What are we trying to do with AI, what does that mean for your work this quarter, and who would you ask if you got stuck. Write down their words rather than what you think they meant. Twenty minutes gives you a more honest reading than a six-week engagement survey.

Why does good news travel up and bad news stay down?

Because every person in the chain summarises honestly and keeps the part worth your attention, and the part worth your attention is the part that is working. Four teams become the best team, twice over, and a promising pilot arrives at the top as a company-wide capability without anybody saying a false sentence.

Why is a leader's own experience of AI a bad guide?

Because it is the most favourable one in the building. Executives have the licences, the interesting problems and the freedom to spend forty minutes trying something: 92% say they are comfortable using AI to get work done, against 51% of employees. The experience is real. It is a sample of one, taken from the best seat.

How do you see where inside the company the gap is widest?

Person by person, because the fix for a team that is already ahead is the opposite of the fix for a team that has not started. emaho measures one Operating Profile per person, personality type and AI level in a single profile, so the picture you get is the real distribution rather than an average. Most leadership teams are surprised at both ends of that range.

Where should AI expectations actually appear?

In goals, in one-to-ones and in reviews, because those are the three places people actually read. In the 2026 Betterworks data, 45% of executives said team goals frequently include using AI, against 6% of individual contributors. A message that lives only in the strategy deck stays a topic rather than an expectation.

Are employees still afraid AI will take their jobs?

Much less than a year ago. In Betterworks' 2026 survey, 9% of employees worried about being replaced by AI, down from 51% in 2025. The dominant feeling has shifted from fear to ambiguity: people are open to using it and unclear about what their company is actually trying to do with it.

Who owns AI adoption inside a company?

In most companies, nobody clearly. Betterworks cites SHRM data showing that only 26% of HR leaders say they have been involved in managing change and employee adoption of AI, while two thirds believe HR should be involved in things like assessing readiness and defining use cases, and only 17% say their organisation's AI implementation has been successful.

How do you close the gap without launching a programme?

Use meetings you already have. Put one AI line into an existing team goal and make it about the work rather than the tool. Ask three questions in the one-to-one you already hold, once a quarter. And give fifteen minutes of the next all-hands to one team showing what they changed and what broke. A new programme is another thing people hear about; these three change what somebody does on Thursday.

Close by

Widening the same gap

Each one is a different reason the picture at the top stops matching the floor.

06

Usage gap

How far off the leadership team's estimate of who uses what usually is, and the two-minute way to check.

Read this one 09

The human cost

What people do with the silence between the announcement and the next update. It is rarely nothing.

Read this one 04

Leaders are the bottleneck

What happens when two people on the same leadership team hold different pictures and never compare them.

Read this one

All 25 AI challenges, and what to do about each

Twenty-five things that break inside a company once people start using AI, each with the research behind it and the level where it starts to bite. This one starts to bite at Blueprint, and 7 of the others start there too.

Open the full set

Getting going

Get the real picture before the next decision.

Fifteen minutes per person, and you can see where your company actually stands rather than where the reporting line says it does. Most leadership teams are surprised at both ends of the range.

Ask a hundred people to explain the strategyNot a survey. One question in a corridor, and you count the answers that match.
Then measure the distance where it is widestOne profile each, and the gap turns up somewhere the leadership team was not looking.

Start with your own profile, free Ask me anything first Plan a 30-minute call

First profile free · no credit card · built for companies of 20 to 500 · you decide what your team gets to see

Not ready to put your team in anything yet? Start with the level of the company instead. The Culture Level scan is six questions, three minutes, and asks nothing of you.

Paul MustersFounder of emaho

Paul Musters

Fifteen years of leadership and team development in Dutch scale-ups. That practice now sits in software: one Operating Profile per person, with agents that actually fit. He writes these pages from what he runs into with clients, not from a research summary.

LinkedIn · paul@emaho.world · WhatsApp

Sources and numbers used on this page

Betterworks, 2026 State of Performance Enablement: The Real ROI of AI, published 2026. Global HR research report based on 2,387 respondents, of whom 36% individual contributors, 30% managers or supervisors of people, 16% directors or above and 18% HR managers or above. Source of: fewer than 16% of employees and managers understanding their company's AI vision; 9% worrying about being replaced by AI, against 51% in 2025; 59% of executives believing they communicate a clear AI vision against 8% of employees who agree; 64% of executives saying employees understand AI's strategic value against 15% of employees; only one quarter of managers saying the AI vision is clear and one third saying the strategic value is clear; 92% of executives being comfortable using AI to get work done against 51% of employees; 58% of employees saying AI has not improved their day-to-day work at all against 8% of executives; executives being seven times more likely to say team goals frequently include using AI, 1.6 times more likely to say AI improves productivity, 2.1 times more likely to credit it for creativity, 2.7 times more likely to say it makes work easier and six times more likely to believe performance reviews have kept pace; team goals including AI expectations at 45% of executives, 30% of HR management, 18% of managers and 6% of individual contributors; 85% of HR leaders believing employees get adequate AI training and 88% reporting moderate to high AI fluency; and the statement that exposure does not equal application. The quotation is from Jamie Aitken, VP of HR Transformation at Betterworks, in the same report. Betterworks sells performance management software, which is worth knowing when reading its conclusions about performance management. The figures are self-reported and weighted towards office-based roles.
SHRM, as reported by Betterworks in the same 2026 report: only 26% of HR leaders say they have been involved in managing change and employee adoption of AI, two thirds believe HR should be involved in activities such as assessing organisational readiness and defining AI objectives, and only 17% say their organisation's implementation of AI has been successful. This is a second-hand citation and the underlying SHRM report has not been read directly.

Numbers are quoted as published. The meeting described in the recognition block and the quotations attributed to Paul are drawn from client situations rather than transcripts.