Diagnose First: A Manager's Playbook for Stalled AI Adoption

Key Takeaways

• Same Stall, Different Causes: Don't treat a stall as one problem with one fix
• Skills Gap Myth: Look past the skills explanation first
• Readiness State Matters: Determine exactly where a team stands
• Three Paths, Not One Framework: Match the move to the actual gap
• Bounded Control: Focus on determining the gap, not fixing the whole system

A rollout lead pulls the AI adoption numbers for the quarter and finds no clear story. Some departments show strong usage. Others barely touched the tools they were trained on. A few land somewhere in between, using AI for a handful of tasks and ignoring the rest.

Leadership wants an explanation, so every manager gets the same instruction: find out why your team stalled.

One manager starts there, expecting to find a skills problem, since that's usually the answer. Instead, the team already completed the recommended training, and people say AI helps whenever they use it. But usage still hasn't taken hold, and there's no shared definition of what “adoption” is even supposed to look like.

Other managers are asking the same question, in departments with different workflows and different constraints. The label on the problem is the same. What's driving it isn't. Usage dashboards flatten these differences, since they measure whether people opened a tool, not why they stopped.

Look Past the Stall to Find the Cause

When adoption stalls, it's tempting to treat it as one problem with one fix. It rarely is. A stall is a shared symptom, and what's driving it looks different from department to department, sometimes even team to team within the same department.

Sorting out what's happening starts with keeping three things distinct. Confusing them is what leads a manager to fix a non-existent problem. Here's how they differ:

• Root cause: the underlying reason adoption stalled
• Readiness state: where the team stands right now, based on what people do
• Capability path: the specific move that closes the gap for the team, whether that's alignment, training, or practice

Diagnosing this doesn't require deep AI expertise. It requires paying attention to what a team does day to day, not the technology itself.

Leaders often default to a skills explanation first, since it's familiar and easy to act on. A completed training session and an embedded capability aren't the same thing, though, and the space between them is where the real opportunity sits. Most stalls trace back to something a manager can see and address once they know where to look.

A few signs point to a genuine skills gap. Others point somewhere else entirely.

• Avoidance: someone still does a task by hand that AI could handle faster
• Misapplication: someone pastes a full report into an AI tool with a vague request instead of a specific task
• Permission, not skill: people know how to use the tool but aren't sure they have permission
• No shared standard: no one has defined what “good” looks like
• Invisible bottleneck: output doesn't move forward because it needs approval

These signs show up in different ways: a solid draft sits unused because no one has confirmed that submitting AI-assisted work to a client is fine, two people produce equally strong output and a reviewer flags one and not the other, or approval sits with someone who doesn't know the output exists.

There is a cost to getting the diagnosis wrong. Another workshop won't fix confusion about what people can do, another course won't fix a habit that never formed, and leaving the rules unclear means managers keep inventing their own.

Spotting the right sign in the pattern of daily work, not in a report, avoids all of that. A single accurate observation does more than assumptions about skill level ever could, because it points to something specific enough to act on.

Diagnose the Real Adoption Barrier

Understanding where a team stands starts with determining its readiness state, the pattern of adoption it shows in practice right now.

Self-perception gets in the way here, since managers tend to assume their team is further along than it is. A training session happened, and someone tracked attendance, but what people do the next week differently is a better measure than who showed up.

Most teams fall into one of five common states, and each one points to a different starting move.

• Aware, not active: people know the tools exist and have heard the pitch, but nothing has moved into daily use
• Tried once, stalled: a tool got tested once, and no one went back to it
• Inconsistent use: a handful of people rely on it regularly while most don't touch it at all
• Narrow use: the tool stayed in rotation, but only for one task, without expansion to others
• Embedded: usage is steady and consistent, and no longer needs a champion to keep it alive

A team sat through a company-wide AI demo months ago and never opened the tool since. Another ran one pilot project, called it a success in the wrap-up meeting, and never ran a second one.

On a six-person team, two people use AI daily while the other four have never opened it. One team only uses AI to draft routine emails because no one has shown them what else the tool can do. Another reaches for AI without anyone reminding them to.

Seeing these five states laid out is one thing. Recognizing a team in one of them is another.

A team's readiness state is rarely subtle once a manager knows to look for the pattern in daily work. A team stuck at aware-not-active needs something different from a team stuck at narrow use, even though both look like the same flat line on a usage report. Determining the state turns the fix from a guess into a specific move, built for what's happening.

Build the Capability Your Team Needs Most

Once a manager knows which problem they're solving, the conversation changes. A readiness state points to one of three moves: alignment, role-based training, or applied practice. These aren't a second framework to learn. They're the action that follows from the diagnosis already made.

Each path solves a different kind of gap, and picking the wrong one means solving a problem the team doesn't have.

• Alignment: the barrier is confusion about what people can and can't do
• Role-based training: the skill gap is real, but generic training never covered what roles demand
• Applied practice: the team attended training, but without repetition in daily workflows, habits never formed

When permission confusion is the barrier, more training doesn't help. Someone on the team might already know AI could speed up a vendor report but hesitates because no one has said whether pulling company data into an outside tool is fine. The fix is a clear, written policy the whole team gets once, instead of individual managers guessing their own answers.

A different path applies when a team never got instruction tailored to their role. A generic AI overview course covers the same ground for a finance analyst and a customer support rep, even though the two barely overlap in what they'd use the tool for day-to-day. Role-based training means building or requesting instruction around the specific tasks a team already performs.

Applied practice skips scheduling another workshop, since the team already has the skill. What's missing is repetition built into their actual workflow, so the manager builds one AI-assisted task into a recurring weekly process, something the team already does, rather than adding a new tool to learn on top of everything else.

None of these three moves stays contained to one team, since no single manager controls every lever that makes it possible. Tool approval, data security, and training budgets all sit with different people, which means closing a capability gap in one department usually means coordinating with several others.

When the right capability takes hold, the signs are concrete. Employees reach for AI without a reminder. Teams land on the same workflow instead of a dozen individual habits, and approved tools become the default option rather than the exception. Managers spend less time enforcing use, and employees spend less time working around the tool instead of with it, freeing everyone for higher-value work.

Adoption gaps this wide aren't rare. Workforce data compiled by Worklytics in 2025 points to AI usage varying sharply by department within the same organization, with some teams reaching 80 percent usage or higher while others stay under 20 percent.

This pattern holds across industries, which means no department is immune to a stall, and none is destined for one either. What a manager can't control is the organization-wide picture. What they can control is choosing which of these three moves most benefits their team.

Act on What You Control

A manager doesn't need authority over budget or company strategy to make progress here. Diagnosing their team is an action within their reach. The manager from earlier can act: they name the team's readiness state, then pick the move it points to, rather than settling for another vague skills problem.

Four managers hand back four distinct diagnoses, each pointing to a different fix, rather than one flat report saying adoption stalled across the board. The picture stops being a single blur and starts being something leadership can act on, department by department.

This kind of diagnosis doesn't close every gap by itself. Still, it stops an organization from missing issues hiding behind a flat report. Once managers know what to look for, gaps that used to blend into "adoption is stalled" surface sooner, before they've had time to compound into something harder to fix.

New Horizons partners with organizations to close gaps like these, starting with whichever capability a team's readiness state points to.

If your team's adoption numbers look stalled right now, the question is which of the three moves closes the gap.

Explore New Horizons' AI training and start closing that gap on your own terms.