Accorda Advisory.

Accorda Advisory LLC · Camas, Washington

Why AI adoption fails when the technology works

The research puts the cause in the human layer rather than the technology. BCG finds that 70% of AI value comes from workforce change, with only 10% from the algorithms and 20% from technology implementation. That is why a company can have the tools installed, licensed and working exactly as sold, and still see nothing change in the work.

In a founder-led company the pattern has its own shape, because what carried adoption at twenty people stops working once there are managers leading managers.

Why do AI initiatives stall, and how do you fix it?

Not for one reason. That is what makes it hard to fix. A stall is spread across all four PACE pillars at once, and each one is holding a different piece of it — which is why correcting the loudest problem so rarely changes anything.

How this gets found

A stall does not announce itself. From the inside it looks like ordinary friction — a slow quarter, a busy team, a tool nobody got round to. Nobody wakes up and calls it a stall.

So the work is to go looking in all four pillars rather than the one everybody is already arguing about, and to come back with counts instead of impressions. That is what the AI Adoption Gap Diagnostic™ is for: it names which pillar is carrying the strain, so the next twelve months of effort go where the return is not arriving instead of where it is easiest to spend.

The pattern underneath it is consistent. The budget goes where the value is not: licenses get bought, training gets scheduled, and the part that decides the outcome — whether people actually change how they work — gets left to happen on its own.

Three findings, from three different pillars:

  • It is human, not technical. 63% name human factors as the primary challenge. Prosci, 1,107 respondents. The obstacle was never the tool.
  • CEO accountability changes the result. Only 24% name the CEO or executive committee as ultimately accountable for AI-informed decisions. Where accountability does sit at CEO level, 57% report meaningful business value, against 21% where it does not — and are nearly four times likelier to report established ROI (14% against 4%). KPMG. Nearly triple the value, on ownership alone.
  • The perception gap. 75% of executives say their AI rollout succeeded. 45% of employees agree. EMARKETER and Writer. This is why a stall goes unnoticed. The people who would have to see it are the least likely to.

What are the four human barriers?

Resistance is rarely one thing, and the four causes need different responses. Diagnosing which one is in the room is most of the work.

  1. 01Identity threat“If the software does that part, what am I here for?” The role, not the job, is what feels at risk.
  2. 02Tech shameFear of looking inadequate in front of peers, which is often stronger than fear of losing the job.
  3. 03Cognitive overwhelmToo many tools, too fast, on top of the work that was already there.
  4. 04Lack of meaningNobody explained what this is for, so it reads as a cost exercise.

Tech shame is the one most rollouts miss entirely, because nobody raises their hand to report it. It shows up instead as a quiet preference for the old way, and it gets read as stubbornness.

How do you get employees to use AI?

The question skips a step, and it is the step that decides everything after it.

You cannot roll this down to employees while the leadership team is not aligned on it — and in practice they are almost never as aligned as they believe they are. Different executives are quietly running different definitions of what AI is for, how far it is meant to go, and what happens to the people whose work it touches.

Employees hear all of those at once. What they respond to is the disagreement, not the plan.

So the first measurement is not of employees at all. The Accorda Leadership Index scores each executive individually, then shows where the team sits together and where it does not — including the gap between what the owners believe and what the executives and managers report. That gap is the finding, and it is almost always there.

Once leadership is genuinely aligned and it starts rolling down, this is what works with the people doing the work.

Start with what the person is protecting rather than with the tool.

People do not sustainably sacrifice their values. When someone believes there is only one way to protect what matters to them, resistance wins every time, because they are not being stubborn, they are defending something real. Once they can see the value survives the new behavior, they choose the change themselves.

Protect the value.
Change the strategy.

The values people defend in a change like this are ordinary, and worth naming plainly.

  • contribution
  • competence
  • purpose
  • agency
  • belonging
  • recognition
  • security
  • autonomy
  • quality
  • predictability

ExampleA craftsperson resisting an AI draft is usually protecting quality. A long-tenured manager resisting a new workflow is often protecting competence. Neither of them is being difficult, and neither of them will say it out loud in those words.

You do not remove the fear.
You give the value
a new way to survive.

What follows from that is practical. Name what actually changes about the role rather than reassuring people nothing will. Give permission to experiment without a penalty for getting it wrong. Keep the first uses small enough to build confidence instead of confirming the fear. And make sure the people who try it early are visibly better off for having done so.

Automation does not equal job loss. Tasks get replaced and roles evolve, and the human stays essential.

Where this shows up, and how it gets done

This one is hard to locate because it does not sit in a single pillar. Value-protection surfaces in People, when someone is defending whether their expertise still counts. In Adaptive Culture, when it is not safe to try something and have it not work. And in Capability, when nobody was actually taught and saying so would cost them something.

The same behavior, three different causes, three different fixes. Guessing which one you are looking at is how companies spend a year on the wrong pillar — and it is why the assessment reads all four rather than the one being complained about.

Here is what makes this hard to run in-house. People do not tell the person who signs their review what they are actually protecting. Not out of dishonesty — naming it out loud feels like handing over leverage.

  1. Outside. The first conversation is had by somebody with no stake in the org chart, who will not be in the room at anyone’s next performance cycle. That is usually the only way the real answer surfaces.
  2. Inside. It moves deliberately back in-house. A workshop gives the leadership team shared language for what was found, so the argument stops being about the tool and starts being about the value underneath it.
  3. One to one. Each executive and manager works it in their own department, because a manager in operations is protecting something different from a manager in sales.
  4. Re-scored. The same instrument, run again a few months later, after the new behavior has had time to either hold or quietly revert.

Anyone can run a workshop. Anyone can offer coaching. Step four is the one almost nobody does, and it is the only one that can tell you whether the first three worked. Without it you are holding a good day and a set of intentions.

Most executives have never been trained to run the step-one conversation. That is not a gap in their ability. Nobody teaches it, because until recently nobody needed it.

Why do mid-level managers resist AI most?

Prosci’s research identifies mid-level managers as the most resistant group, and the reason is structural rather than personal.

They carry the change without having chosen it. They stay accountable for output while the method underneath is being rewritten. And a good deal of their authority was built on knowing the work better than the people they manage, so a tool that answers faster than they can lands very differently in that seat than it does in the executive suite.

Treating that as an attitude problem guarantees it stays one. Treating it as a role that needs redefining is what moves it.

Where it goes right, the difference shows up in daily use. BCG found 88% of managers at companies getting real AI value use it daily, against 25% at the laggards.

How this gets handled

Middle management is where communication stops being specific. The board hears a strategy. The floor hears a tool announcement. The layer in between is handed the job of turning one into the other, usually without being told what it means for the people they manage — or for their own role.

They are also the first to watch work get automated and often the last to be told what happens next. So they protect two things at once: their team, and their own position. Both are entirely rational responses to incomplete information, and most of us would do the same.

Which is why this reads as resistance and is usually something else. It is a communication gap wearing a behavior problem’s clothes — and the gap sits upstream of the manager, not in them. That is a People and Adaptive Culture finding, and it is one of the most common.

The Accorda Leadership Index scores each executive and manager individually, so the pattern shows up per person and per department instead of disappearing into a company-wide average. The work then happens where the vagueness actually starts, which is rarely where it is being felt.

Is AI adoption just change management?

Traditional change management assumes a defined end state with a finish line, which is exactly what makes a project plan work. You describe the destination, you move people to it, and you close the project.

AI does not offer a destination. Tools change monthly and use cases keep expanding, so there is no state to land in and declare finished. That makes adoption an ongoing transformation rather than a project that closes, and it needs a pace rather than a deadline.

Workplace paradigm shifts arrive roughly every fifteen to twenty years. The PC in the early eighties, the web in the mid-nineties, mobile in 2007. Almost nobody in the building has led an organization through one, which makes this a timing problem rather than a character problem.

Implementation has a deadline.
Adoption has a PACE.

See the four pillars and the twelve domains

What that looks like in practice

Change management runs a plan and reports whether the plan was completed. This measures which of the four pillars is under the most strain, works on that one, and then re-scores the same instrument a few months later.

The difference is what you are holding at the end. Not a record of activity, but a number that moved or did not, in your own reporting. Seeing where you actually stand is the first step.

The sources

  • 70% of AI value comes from workforce change

    10% from the algorithms, 20% from technology implementation.

    BCG, AI Transformation Is a Workforce Transformation, February 2026
  • 63% name human factors the primary challenge

    Roughly 38% of implementation difficulty is user proficiency against roughly 16% technical. Mid-level managers reported as most resistant.

    Prosci, n=1,107
  • 88% of managers at AI leaders use it daily

    Against 25% at laggard companies.

    BCG, Build for the Future x AI global study
  • 43% of U.S. employees never use AI

    One quarter strongly agree their organization has a clear AI plan.

    Gallup Panel, Workforce Study Q2 2026
  • 24% name the CEO or executive committee as accountable for AI-informed decisions

    Where accountability is clear at CEO level, 57% report meaningful value against 21% where it is not.

    KPMG Global AI Pulse Q2 2026, n=2,145
  • 75% of executives say the rollout succeeded

    45% of employees agree.

    EMARKETER and Writer, 2026

Frequently asked questions

Why do AI initiatives stall, and how do you fix it?

They stall in the human layer rather than the technology. BCG found that 70% of AI value comes from workforce change, with 10% from the algorithms and 20% from technology implementation. Prosci research with 1,107 respondents found 63% naming human factors as the primary challenge, and roughly 38% of implementation difficulty coming from user proficiency against roughly 16% from technical causes.

How do you get employees to use AI?

Start with what the person is protecting rather than with the tool. People do not sustainably sacrifice their values, so when someone believes the only way to protect what matters to them is to resist, they resist every time. Once they can see the value survives the new behavior, they choose the change themselves. In practice that means naming what changes about the role honestly, giving people permission to experiment without penalty, and making the first uses small enough to build confidence rather than overwhelm.

Why do mid-level managers resist AI most?

Prosci research with 1,107 respondents identifies mid-level managers as the most resistant group. They carry the change without having chosen it, they are accountable for output while the method underneath is being rewritten, and their authority has often been built on knowing the work better than their team. A tool that answers faster than they can lands differently in that seat than it does in the executive suite.

Is AI adoption just change management?

Traditional change management assumes a defined end state with a finish line, which is what makes a project plan work. AI is continuous. Tools change monthly and use cases keep expanding, so there is no state to land in. That makes adoption an ongoing transformation rather than a project that closes, and it needs a pace rather than a deadline.

Start here

Find out where your adoption is actually breaking down.

If any of the four barriers sounded like your building, the Snapshot will tell you which one is widest. Five minutes, no sales call attached.

Book a call Take the 5-minute Snapshot

Not ready to talk yet

Get the next one in your inbox

People who find this page are usually reading something that finally names a piece nobody had named for them. Every week or two I publish one more of those — where AI adoption is actually breaking down inside companies, built from the research and from what I see in real leadership teams.

No pitch, no sequence designed to wear you down. If it stops being useful, one click and you are out.

Would you rather start with a read on your own organization? The AI Adoption Gap Snapshot™ takes five minutes and covers all twelve domains.