Resistance to AI Adoption: What Changed and What Works

Resistance to AI adoption looks different now. Two years ago it meant staff refusing to touch the new tool. Today it usually means the opposite problem: your people are already using AI on their own terms, quietly, while the business has not decided what it considers acceptable.

The gap is measurable. Office for National Statistics analysis published in July 2026 found that 55% of employees report using AI for work or education, against 35% of businesses reporting use of at least one AI technology. Twenty percentage points sit between what staff do and what their employers think is happening. That gap is where most stalled rollouts now live.

This guide covers what actually drives resistance to AI adoption in UK and Irish SMEs today: selective non-use, accuracy fears, job security worries, and the policy vacuum feeding all three. Each one gets a practical answer you can apply this quarter.

What UK Adoption Data Says About Resistance to AI Adoption

Adoption is climbing, but it is thin. The ONS reported that 35% of UK businesses with 10 or more employees used at least one AI technology in June 2026, up from around 12% in late 2023. Depth has barely moved. The average adopting business runs 1.6 AI technologies, up from 1.4 over the same period, and only 10% of adopters describe their use as extensive.

Size changes the picture. Among businesses with 0 to 9 employees, 28% report using at least one AI technology, rising to 49% among firms with 250 or more staff. Sector matters more still: 58% in information and communication against 13% in construction.

For an SME owner, the useful read is this. Most of your competitors have started. Almost none have finished. And resistance to AI adoption is rarely the thing separating the two groups, because 41% of businesses told the ONS they had hit no barriers at all in the previous three months. What separates them is whether anybody in the business owns the change.

The Gap Between What Staff Do and What the Business Knows

The employee figure and the business figure come from two different ONS surveys asking two different questions, so treat the 20-point gap as directional rather than exact. The direction is clear enough. Only 15% of businesses report that more than half their staff use AI daily, which means a large share of AI use in UK workplaces is happening below the level anyone is tracking.

This reframes the problem. If you’re an SME owner worried that your team will reject a new tool, the more likely scenario is that several of them are already using one you haven’t approved. Giving that informal habit a shared standard is what practical AI training for your team is for, and it works better than a rollout announcement. Pairing it with a clear digital strategy for the business gives people a reason to change rather than an instruction.

Table 1: UK AI adoption at a glance

Source: ONS, Artificial intelligence in UK businesses, June 2026 data.

Resistance to AI Adoption Has Changed Shape

Henley Business School’s World of Work Institute surveyed 2,900 full-time UK workers across 29 sectors for its twelve-month AI pulse check, published in June 2026. Optimism edged up to 58% from 56% a year earlier. The share feeling overwhelmed by the pace of change didn’t shift at all, holding at 61%. Asked to pick one word for how AI feels at work, more workers chose cautious than anything else.

One finding should interest any SME owner more than the rest. Nearly two-thirds of workers, 63%, sometimes choose not to use AI tools in their role even when those tools are sitting there available to them.

That’s the modern face of resistance to AI adoption. Not refusal, and not sabotage. Selective opting out, task by task, by people who are willing in principle and unsure in practice. It is much harder to spot than the old version, because nobody says no in a meeting.

Why This Bites Harder in a Small Team

A large employer can absorb a pocket of quiet non-use. In a business of ten or twenty people, two reluctant team members stall the whole project, and there’s no transformation office to notice. Smaller firms carry less slack, which makes early, honest handling of the problem more important rather than less. The upside is that leaders in small teams can speak to people directly instead of through three layers of management.

Table 2: Four resistance patterns and what answers each

Accuracy Fears and the Checking Tax

Accuracy worry is real, and it’s rational. It is also the form of resistance to AI adoption most likely to be mistaken for diligence. In the Henley survey, 28% of workers said they worry about being able to spot errors and bias in AI output. Above that sat two related concerns: 42% worry about becoming too dependent on the tools, and 35% worry about losing critical skills.

The cost shows up as double-handling. Staff run the task through the tool, then redo enough of it by hand to be confident, and the promised time saving quietly disappears. When an SME owner says the pilot delivered nothing, this is usually why.

The Fix: Tier the Work Before You Roll Out

Sort your tasks into three bands and write them down. Band one is draft only, where AI produces a first version nobody sends without editing: internal notes, first-pass copy, meeting summaries. Band two is check before sending, where a named person signs off: customer emails, quotes, anything with a number in it. Band three is off limits, where the tool isn’t used at all: legal wording, medical or financial advice, anything involving special category data.

Two things happen when you do this. People stop checking everything, because you’ve told them which things need checking. And the person who worried about being blamed for a machine’s mistake now knows exactly who signs off what.

A second tactic is worth the ten minutes it takes. Demonstrate the tool failing before you demonstrate it succeeding. When a team watches the system get something wrong and a colleague correct it, the threat level drops. The tool stops being a mysterious authority and becomes an assistant whose work a person owns.

Job Security: What the Headcount Data Actually Shows

Fear of redundancy remains the strongest single driver of resistance to AI adoption, and the sentiment data is stark. Research from ADP reported in April 2026 found only a quarter of UK workers felt confident their job would not be eliminated by AI. The split by seniority tells its own story: 35% of C-suite executives felt safe, against 23% of middle managers and 18% of people below management level.

Workers in repetitive roles were least confident of all, at 16%, compared with 30% of knowledge workers. The same research found people who felt secure in their job were six times more likely to be fully engaged and three times more likely to describe themselves as highly productive. Insecurity is not just uncomfortable. It is expensive.

The Honest Answer Beats the Reassuring One

Blanket reassurance fails because staff can read the news. What works is specificity, and the specifics are less alarming than the fear. The ONS found around half of businesses using AI reported no change at all to workforce headcount, with just under 7% of medium-sized businesses reporting a decrease. Among firms using AI to improve operations, 63% reported no change, 6% a decrease and 1% an increase.

CIPD research published last November put the employer view at one in six businesses expecting AI to reduce their workforce over the following twelve months. That’s a real number, and it deserves to be said out loud rather than smoothed over. Most UK businesses using AI have not cut jobs; some will, and the roles most affected so far are administrative, clerical and creative tasks rather than whole jobs.

Say which of those applies to you. Where roles will genuinely change, say so and pair it with retraining, because a person who can see a route forward stops scanning for signs that their role is at risk.

The Policy Vacuum Is Now the Biggest Single Driver

Here’s the finding that ties everything together. In the Henley research, 60% of UK workers said their employer either has no AI guidelines or they aren’t sure whether it does. That figure improved from 68% the year before, which still leaves most of the UK workforce making daily decisions about AI with nothing written down to guide them.

Put that next to the training picture. The ONS found only 11% of businesses with 10 or more employees had given AI-related training to more than half their workforce. Staff is using the tools, the rules do not exist, and nobody has been shown how. Resistance to AI adoption in that environment isn’t stubbornness. It’s a reasonable response to being asked to take a risk without cover.

The Fix: One Page, Three Lists, No Punishment

An SME does not need a governance framework. It needs one page that anyone can read in two minutes:

• Approved tools, named. Which accounts, on which plan, and who pays for them.
• Red-line data. What never gets pasted into an AI tool: customer records, payroll, anything covered by a confidentiality clause.
• Sign-off rules. Who checks what before it leaves the building, matching the three bands above.

Start with an amnesty. Ask the team, anonymously, which AI tools they already use and what they’d lose if those disappeared tomorrow. Say plainly that nobody is in trouble. You’ll get an honest map of what’s already happening and a ready-made list of what your approved tools need to do.

The ONS data backs the training response too: 62% of businesses citing a lack of expertise as a barrier said they were training or retraining existing staff, against 26% of businesses reporting no barriers. Firms that feel the gap go and close it. Structured digital training for your staff turns that from an intention into a date in the diary.

Change Management That Works in a Small Team

Strong change management turns a tense rollout into a calm one, and none of it is complicated. Resistance to AI adoption drops when people understand why a change is happening, what it means for their week, and where they can raise a concern without it counting against them.

Involve People Before You Decide

The fastest way to create pushback is to hand staff a finished decision. Before committing to a tool, ask the people who’ll use it where the pain sits. Finance staff know which invoice tasks waste their week. Sales reps know which admin steps slow them down. When people shape the choice, they own the outcome instead of enduring it.

Pilot, Learn, Then Widen

Company-wide rollouts overnight invite failure. Pilot in one team, fix the rough edges, gather real numbers, then go wider. A colleague vouching for a tool moves opinion far more than a directive from the top, and a successful pilot gives sceptics evidence rather than promises.

Train for Roles, Not in the Abstract

A generic AI overview rarely calms a worried team. Role-specific sessions, where finance, marketing and admin staff each see how the tool handles their own tasks, land far better. Hands-on practice in a safe sandbox lets people make mistakes without consequence, and confidence built that way sticks. The ONS found training and retraining existing staff is by far the most common way UK businesses bring AI skills in, reported by around 40% of medium and large firms. Recruiting new people with AI skills stays rare, rising from about 2% of the smallest businesses to 10% of the largest.

Here is what Ciaran Connolly, Director of ProfileTree, has observed working with Northern Ireland SMEs:

“The resistance we meet in Northern Ireland businesses has flipped in the past year. It used to be staff refusing to touch the tool. Now it’s owners discovering half their team already use AI at home and nobody has told them what’s allowed at work. The turning point in our training is almost always the same moment: a sceptical team member watches the tool get something wrong and a human fix it. That single demonstration does more for trust than any slide deck.”

Appoint People Who Want the Job

Pick a willing person in each team as the go-to for questions. Peer support spreads adoption in a way top-down instruction cannot, and it gives you early warning of where reluctance is building. Champions also keep momentum after the initial training, which is exactly when most rollouts quietly fade.

Building that capability internally, rather than buying a tool and hoping, is what ongoing AI transformation support is designed to do.

What that looks like from the other side, in a client’s own words:

“Throughout the 8 sessions i was able to gain new skills and knowledge to strengthen my confidence performing my job role in marketing. We have been provided with lots of new tools and apps to use going forward and lots of branding advice and best practice for many areas of our digital marketing such as SEO, AI, Canva and Social Media Management.”

Rachel Adams, Google review, March 2025

The Legal and Ethical Ground for UK Businesses

Governance isn’t a soft extra in the UK, and getting it right reduces resistance to AI adoption because people feel protected rather than exposed.

UK GDPR and the Information Commissioner

Any AI tool processing personal data falls under UK GDPR, overseen by the Information Commissioner’s Office. Staff are right to ask where customer data goes and how it’s stored. Answering those questions before anyone has to push removes a common source of pushback and keeps you on the right side of the regulator. The UK’s approach still leans on existing regulators rather than a single AI statute, so the obligations you already have are the ones that apply.

Keep a Person in the Loop

For decisions affecting people, such as hiring or lending, meaningful human review is both an ethical expectation and increasingly a regulatory one. Building that oversight into the process protects staff from the fear of carrying the blame for a machine’s mistake, which is a direct driver of reluctance. Well-built AI chatbot solutions make the handover to a person straightforward rather than an afterthought.

Bias and Fairness

A model trained on skewed data produces skewed results. Checking outputs for bias, and being honest with staff and customers about what the tool can’t do well costs little and buys a lot. Teams that watch their employer take fairness seriously push back far less, because the tool stops looking like a risk imposed on them.

Where to Start This Quarter

If you take one thing from the data, take this: your team is probably further ahead than your policy. Resistance to AI adoption in a UK SME today is less about people refusing technology and more about people waiting for permission, clarity, and a bit of training.

Three moves, in order. Run the amnesty and find out what’s already in use. Write the one-page policy naming approved tools, red-line data, and sign-off rules. Then train by role, in a sandbox, with a pilot team small enough to fix things quickly.

Ready to bring your team with you? ProfileTree’s hands-on AI training programmes are built around role-specific learning for SMEs across Northern Ireland, Ireland and the UK.

FAQs