AI Adoption: How to Make It Stick

By Johannes Sundlo, AI & Future of Work Advisor. I help leaders and HR teams turn AI into adoption that sticks, through keynotes, workshops and change programs.

Most organisations don’t have an AI problem. They have an adoption problem. The tools work, the pilots look promising, and then nothing changes in how people do their jobs. This is the practical guide to closing that gap: what real adoption means, why it stalls, and the conditions that make AI stick.

What “AI adoption” means

Adoption isn’t licences purchased or pilots launched. It is a measurable change in how everyday work gets done: people reaching for AI on their own because it makes the work better. Access is easy to buy. Adoption has to be earned, and it’s the only part that creates value.

Why most AI initiatives stall

  • It’s bolted on, not built in. AI lives in a separate tool people forget to open instead of inside the workflow they already use.
  • No early, visible win. A 12-month transformation with nothing to show in month one loses the room.
  • Fear goes unaddressed. People quietly worry AI replaces them, so they don’t lean in.
  • It’s treated as IT’s job. Adoption is a change-leadership challenge wearing a technology costume.

The “Adoption that Sticks” model

Across keynotes and change programs with organisations of every size, the same five conditions show up wherever AI takes hold. Miss one and adoption stalls.

The gap is not unique to any one company. Gallup finds that over half of US employees now use AI at work, while Randstad reports that only 13% of workers have been offered any AI training. Tools spread faster than the ability to use them, and that is exactly the gap adoption work closes.

  1. A painful, specific starting point. Begin with one repeated, low-risk task people already dislike, not a strategy.
  2. Built into the workflow. The AI step lives in the template, checklist, or doc people already use.
  3. Human in the loop. AI drafts, a person decides, so people trust the output and own the result.
  4. Visible local proof. “This used to take 40 minutes, now 8.” Concrete wins spread faster than any mandate.
  5. Leaders who go first. When leaders use AI out loud, everyone else gets permission to.

Where to start, by role

Adoption looks different depending on who you are. Two practical guides go deep on each:

How to measure AI adoption

Track three signals, not vanity metrics: breadth (how many teams use AI in real work without being told to), depth (is it touching meaningful tasks, not toys), and momentum (are new use cases appearing on their own?). When all three trend up, you’ve moved from pilots to a culture.

AI change management: a five-phase framework

What is AI change management? AI change management is the work of getting people to use AI in their daily tasks after the tools are bought. The framework I use has five phases. Readiness: map where AI already shows up, which tasks people dislike, and what they fear. Leaders first: executives use AI visibly on their own work for four weeks before any rollout. Role-based enablement: one hour of foundations plus one real task per team, instead of generic training. Guardrails: a one-page policy that says which data never goes in, who checks the output, and which tools are approved. Measurement: breadth, depth and momentum, reviewed every quarter. The framework overlaps with Prosci’s ADKAR model but starts with leadership behaviour rather than awareness campaigns. In the 2025 State of AI in HR survey, 81% of Swedish HR professionals had access to generative AI and 33% had training, so the plan starts where the gap is.

PhaseWhat happensOwnerSignal it worked
1. ReadinessInventory of current AI use, three painful tasks per team, a short fear surveyHR with one executive sponsorA shortlist of tasks people want help with
2. Leaders firstExecutives use AI on their own week and show it, for four weeksCEO and leadership teamEvery leader can name what they used AI for last week
3. Role-based enablementOne hour of foundations, one real task per team, prompts saved and namedTeam leads with HREach team has three named prompts in use
4. GuardrailsOne-page policy: data that never goes in, who signs off, approved toolsHR, IT and legalThe “can I paste this?” questions stop
5. MeasurementBreadth, depth and momentum reviewed every quarterExecutive sponsorNew use cases appear without being asked for
The five phases, who owns each one, and how you know it worked.

How this differs from ADKAR

ADKAR (Awareness, Desire, Knowledge, Ability, Reinforcement) is Prosci’s general change model, and it works for AI with one adjustment. Awareness is no longer the problem, since most employees already know AI exists and many use it quietly. Desire and Ability are the gaps, and both move fastest when leaders use the tools in the open and teams practise on a real task. That is why the framework above puts leaders first and training third. If you want the phases run with your leadership team, that is what the executive AI workshops and change programs are for.

Common questions about AI adoption

Why do most AI pilots fail?

Most AI pilots fail because they run as isolated side-projects, disconnected from the work people do every day. A pilot proves a tool works in a demo; adoption means changing habits, which is a harder job. Three patterns kill them: no early, visible win in the first month; no owner with the time and mandate to carry it; and no behaviour change from leaders. In a 2025 survey of 461 Swedish HR professionals, 81% had access to generative AI, yet only 25% used AI agents and 33% had any formal training, so the bottleneck was enablement. Pilots that succeed start with one painful, repeated task, build the AI step into an existing workflow, keep a human in the loop, and make the result visible so it spreads on its own. See the full State of AI in HR data.

How long does AI adoption take?

Expect meaningful change in a quarter, not a week. Lasting AI adoption is a behaviour shift, and behaviour shifts take a few months of deliberate effort, not a single workshop or a big-bang launch. A practical 90-day cadence works best: in the first month, leaders experiment openly and set simple guardrails; in the second, back two or three real pilots with protected time; in the third, build AI into onboarding, templates and how work gets reviewed so it becomes the default. Speed matters on learning, patience matters on scale. The organisations stuck after a year are usually the ones that either rolled out everything at once or analysed endlessly without shipping anything. Momentum compounds: each visible win makes the next team more willing to try. For year-over-year benchmarks on where Swedish HR teams are, see the State of AI in HR data.

Who owns AI adoption in an organisation?

AI adoption is shared, but it has to be led. Leaders set the direction, remove friction, and model the behaviour by using AI in their own week; teams own the specific workflows where it lives. It is not “IT’s project”. The moment AI becomes one department’s side-task, it stays in a corner and never reaches daily work. Ownership is still unsettled. In 2025 only 3% of HR professionals saw AI primarily as an HR question, down from 6% the year before, while 39% called it an IT question. So the function best placed to drive people-and-change adoption is the least likely to be leading it, which is both the gap and the opportunity for HR. The fix is a leader who goes first, a clear owner per workflow, and visible local proof. More in the State of AI in HR data.

What is the best way to implement AI change management in an organisation?

Start with leaders. Run a four-week period where the leadership team uses AI on their own work and talks about it. In parallel, pick three painful, repeated tasks per team and give each team one hour of foundations and one hands-on session on a task. Publish a one-page policy the same month. Measure breadth, depth and momentum after 90 days. Organisations that start with a company-wide training program or a tool rollout usually see high initial logins and low use after two months, because nothing changed in the daily workflow.

Make AI adoption happen

Keynotes that make AI click for the room and workshops that turn it into habits your team uses the same week. Let’s find the right format.

Pilot vs. adoption: what’s the difference?

A pilotReal adoption
GoalProve the tool worksChange how work gets done
ScopeOne team, time-boxedSpreads on its own
After it endsOften nobody owns itBuilt into daily workflows
Key signal“It worked in the demo”Used without being told
Typical outcomeQuietly diesBecomes the default
The gap between a successful AI pilot and real adoption.

The five conditions at a glance

The five conditions for AI adoption that sticksThe 5 conditions for AI adoption that sticks1Painful, specific startOne repeated, low-risk task people already dislike2Built into the workflowThe AI step lives in the tools people already use3Human in the loopAI drafts, a person decides and owns the result4Visible local proofConcrete wins spread faster than any mandate5Leaders go firstWhen leaders use AI out loud, teams follow
The five conditions of the “Adoption that Sticks” model.

The data: see State of AI in HR: Sweden 2024 to 2025, original survey data on the AI adoption gap.

Related guides

All guides on this site

Courses and learning

Speakers and workshops

AI in HR practice

Leaders and adoption

På svenska