Subscribe to our newsletter to receive business tips, learn about new funding programmes, join upcoming events, take e-learning courses, and more.
Take the first step to successfully starting and growing your business.
Join for free)
Posted: Tue 7th Jul 2026
Last updated: Tue 7th Jul 2026
11 min read
In 2024, the fintech company Klarna announced that its AI assistant was performing the work of 700 customer service agents, portraying the company as a leading example of an AI-first business.
By May 2025, however, Klarna's CEO informed Bloomberg that the company was rehiring human agents, acknowledging:
"Cost unfortunately seems to have been a too predominant evaluation factor when organising this. What you end up having is lower quality."
Klarna maintains its AI-first approach, with its assistant continuing to manage most routine queries. But this situation illustrates that prioritising cost-cutting over human considerations can undermine results.
If an organisation with extensive engineering resources and a substantial technology budget can misjudge the human aspects of AI, smaller firms are unlikely to succeed without deliberate effort.
The central argument is that three human fault lines – trust; skills and confidence; and leadership and culture – frequently impede AI adoption in small businesses.
Small business leaders are under persistent pressure to implement AI.
Numerous tools claim to save businesses time, lower their costs and make their operations more efficient.
But actual adoption rates remain low. UK government research, based on interviews with 3,500 businesses, indicates that only one in six UK firms currently uses AI, and among these, only approximately 30% of staff use it.
While some employees are experimenting with AI tools, many disregard them, which means regular workflows don't actually change all that much. This difference is what ultimately stalls AI adoption.
In the US, Pew Research Center surveyed American workers and found that around half voice concern about how AI will be applied to their workplaces.
At the same time, about one-third see AI in a positive light. Many experience both sentiments at once.
Employees appreciate the potential to remove some of the most repetitive tasks, but remain apprehensive about job security, evolving expectations and the degree of control they'll retain.
Without dealing with these human factors, businesses are unlikely to be able to adopt AI tools, no matter what capabilities they can demonstrate.
The rest of this blog looks at the three primary fault lines that commonly hinder AI adoption in small businesses and outlines workable strategies for each.
The first fault line is trust. Not just "Do people trust AI?" but how they trust it.
In smaller teams, that question usually appears in two extremes side by side:
People who don't trust AI at all and quietly avoid it
People who trust it too much and accept outputs without checking them
The psychological foundations are well established.
Lee and See's seminal research on trust in automation demonstrates that the aim isn't simply to increase trust, but to achieve calibrated trust. That is, trust that corresponds with the technology's actual capabilities.
Insufficient trust prevents organisations from realising benefits, whereas excessive trust discourages critical evaluation. That brings significant risks when AI influences decisions related to customers, finances or safety.
Over-trust has real consequences.
Air Canada's chatbot invented a refund policy in conversation with a customer, and a tribunal forced the airline to honour it, rejecting the argument that the company wasn't responsible for what its bot said.
The AI made the error, but the business carried the liability.
Under-trust is quieter but just as costly. Slack's Workforce Index found that only 7% of US desk workers see AI answers as completely trustworthy for work-related tasks. That's a lot of potential benefit left on the table.
Warning signs you have a trust fault line:
People say "I just don't buy it; I'd rather do it myself."
Others copy and paste AI outputs into emails or documents with little or no checking.
Staff aren't sure who's responsible if the AI gets something wrong.
You hear whispers about "being watched" when AI is used for monitoring or reporting.
In summary, people want to know what the AI is good at, where it's weak, how it works in broad terms and who's accountable when it's used.
Recent research on AI literacy finds that when people receive brief training on how AI works, its limits and its risks, their attitudes improve and they become more willing to use the tools.
This is a clear sign that more understanding around AI reduces anxiety.
The appetite is there. An EY survey found that employees overwhelmingly want more guidance on AI, and 73% worry there won't be sufficient training and upskilling opportunities.
In many small and medium-sized enterprises (SMEs), a major obstacle is employees not wanting to appear uninformed.
Observing colleagues who confidently discuss prompts and models, some people choose to continue with familiar techniques rather than risk embarrassment.
Warning signs you have a skills and confidence fault line:
The same two or three people are always the "AI champions", and everyone else opts out.
Team members say "I don't have time to learn this" but can't explain what they'd need to learn.
People worry about "breaking something" or "saying the wrong thing" to the AI.
You don't have any shared guidelines on safe, ethical use, so everyone is guessing.
In summary, promoting general AI literacy, rather than just training on a specific tool, is vital.
People need simple concepts, clear do's and don'ts (especially around data and privacy) and room to practise without feeling judged.
Many SMEs approach implementing AI as they do with other IT procurement – choosing a tool, deploying it and telling staff that it's available.
However, because AI significantly changes processes, individual responsibilities and performance standards, a comprehensive change management process is needed, not a conventional software rollout.
Leaders also consistently misread the room.
A 2025 survey of 1,400 US employees published in Harvard Business Review found that 76% of executives believed their employees were enthusiastic about AI adoption.
Only 31% of those employees were. Leaders were more than twice as optimistic as their own people.
Picture a 15-person agency that announces an AI copywriting tool purely to "do more with less". Within weeks:
the strongest writers quietly fear they're being replaced
junior staff avoid the tool entirely
no-one tells the founder that half the team has reverted to the old workflow
The tool works fine. The rollout doesn't.
This issue is closely related to the concept of psychological safety, as defined by Harvard's Amy Edmondson – individuals must feel secure in asking questions, identifying mistakes and expressing discomfort.
Edmondson's research indicates that teams only adopt new working methods when they feel psychologically safe.
When leaders disregard concerns, emphasise only the benefits and downplay any risks, employees often refrain from voicing their objections.
Some may experiment with AI privately without disclosing what they've found, while others avoid engagement altogether, citing caution.
Warning signs you have a leadership and culture fault line:
Announcements about AI are mainly about saving money, not about helping people do better work.
No-one in leadership is visibly using AI themselves.
There's no clear way for staff to give feedback or report issues.
AI decisions (like using tools in hiring or performance) are made without input from the people affected.
In summary, your people watch what you do with AI, not just what you say about it.
To work out where you're stuck, start with three simple questions:
Trust: If you asked your team, on a scale of 1 to 10, how much they trust your AI tools to support good decisions, what would they say, and why?
Skills: Do most people feel they've had enough time and support to learn how to use AI in their role, or just a general overview?
Leadership: When you talk about AI, do you emphasise helping people and customers, or mainly speed and cost?
As you think about recent projects and conversations, one of these areas will usually feel heavier than the others. That's your first fault line to tackle.
You can even share these questions with your leadership team or the whole company and compare answers.
Remember the Harvard Business Review finding above – the gap between what leaders assume and what staff feel is often the most revealing data point you have.
Once you've identified your main fault line, you can choose targeted, realistic actions to take.
If trust is your main fault line:
Be explicit about what each AI tool is for, and what it's not for.
Share examples of good and bad outputs so people can see the range.
Make clear that humans remain accountable. The AI is an assistant, not a decision-maker. Air Canada learned this in court, but you don't have to.
Create a simple route for people to report concerns or instances when AI has behaved unusually.
If skills and confidence are your main fault line:
Run short, role-specific sessions where people practise AI on real tasks from your business. The research is detailed in that brief, well-designed training, which shifts attitudes.
Create a one-page AI "code of practice" covering privacy, checking outputs and basic ethics.
Pair more confident users with those who are curious but hesitant.
Appreciate small wins where someone used AI to improve their work, not just to go faster.
If leadership and culture are your main fault line:
Use AI yourself in visible, appropriate ways, and talk honestly about what you're learning, including your mistakes.
Involve staff early in decisions about where the business will use AI, especially where it affects people processes.
Build AI into existing rhythms (team meetings, one-to-ones, retrospectives) rather than treating it as a side project.
Show that speaking up about AI, whether positive or critical, is valued, not risky.
AI adoption is fundamentally a human challenge presented within a technological context.
Small businesses that build trust and confidence and demonstrate effective leadership are more likely to realise the benefits that implementing AI can bring about.
So, if you recognise one of these fault lines in your own business, here's what to do. Pick a single AI use case, involve the people closest to the work and deliberately design for trust, confidence and psychological safety.
The return on AI rarely comes from the tool alone. It comes from the way your team learns to work with it.
