Why managers could make or break AI adoption
AI adoption has moved fast and the harder part now is making it work in practice.
Our latest AI research found that nearly half (49%) of respondents now use AI daily for work, up from 34% last year. At an organisational level, informal experimentation has fallen from 48% to 15%, while the proportion with AI implemented in workflows has risen from 9% to 29%.
It’s obvious that AI has moved out of the experimentation phase and found its place in our everyday work.
However, giving people access to AI and successfully adopting it are two different things. As AI becomes part of more workflows, organisations need people to make decisions about when to use it, when to question it, what information is safe to share, and where human judgement still needs to take the lead.
And increasingly, those decisions are landing with managers.
Go-live isn’t the finish line
Technology rollouts often put a lot of attention on the moment of launch. The tool goes live, the announcement goes out, initial training is delivered, and adoption is expected to begin.
But that’s exactly when the people side of change gets real.
True adoption happens when new behaviours become part of everyday work, and that takes ongoing communication, reinforcement, feedback and opportunities for people to learn from each other.
AI makes this particularly important because there’s rarely one universally correct way to use it. Employees are constantly making judgement calls about where AI can help, which outputs they can trust and how much human oversight a task requires.
As Cameron Robinson, Head of Enterprise Solutions here at Solve, puts it:
“Giving people access to AI is the easy part. The harder work starts afterwards: building the confidence, habits and judgement people need to use it effectively in their day-to-day work.”
A clear AI policy doesn’t guarantee clear behaviour
Most organisations aren’t starting from zero.
Our research found 79% of respondents describe their organisation’s AI rules as very or somewhat clear. But ask a more practical question and confidence drops considerably: just 16% are very confident people understand what data can and can’t be entered into AI tools.
That’s a significant gap, telling us that a policy can state which tools are approved or what information shouldn’t be shared, but it can’t make every decision for them in the moment.
The risks identified by respondents show what that looks like in practice. 57% nominated entering confidential or client data as a leading AI risk, while the same proportion identified relying on AI outputs without checking them. Almost half (47%) pointed to using AI to make decisions without human oversight.
There problems won’t disappear once a policy document is uploaded to a company’s intranet.
Sarah Blanchard, Head of Talent Advisory, explains:
“An AI policy can tell people the rules. It can’t make every judgement call for them. The real work is turning those rules into behaviours people can confidently apply in the moment, then reinforcing them until responsible AI use becomes part of how work gets done.”
And guess where that ambiguity lands?
For many organisations, with the manager.
Managers are sitting at the intersection of AI adoption, quality control, productivity expectations, policy enforcement and team behaviour. They’re being asked to translate organisation-wide AI strategies into decisions their teams make every day.
And yet, only one in four managers feel very prepared to manage AI use within their team according to our latest survey.
Their biggest challenges are revealing. Data security and compliance tops the list at 36%, followed by managing overuse or poor use of AI (29%) and knowing when human judgement is required (29%). Measuring productivity gains (23%) and setting expectations for quality (21%) aren't far behind.
Interestingly, training the team comes last, at just 13%.
As Cameron explains:
“It's surprising that training the team ranks lowest because better training would likely help ease many of the challenges above it, from data security and compliance to overuse, poor use and knowing when judgement is required. We're seeing more AI enablement roles emerge, but relying on one person or team to carry that responsibility is risky; managers are still the critical link between company objectives and day-to-day execution.”
While dedicated AI enablement teams can create frameworks, deliver training and provide specialist support, they can't be present every time an employee needs to decide whether to trust an output, share a piece of information or use AI for a particular task.
You can't centralise your way out of a decentralised behaviour change problem.
AI adoption is becoming a management capability
The role of the manager is changing alongside the technology.
In our survey, 38% expect AI to change middle-management roles without reducing demand, while another 14% believe it will increase pressure on managers without enough support. Just 8% expect little impact.
That means organisations need to think beyond just training managers on how to use individual AI tools.
Managers increasingly need the capability to coach responsible use, recognise poor outputs, understand organisational guardrails, identify where human judgement matters and help teams redesign work around the technology.
They also need permission to have the conversations that sit underneath adoption.
There’s a common misconception that usage equals buy-in, but just because people are using a new technology doesn’t mean they’re convinced why the change is happening, what it means for their role, or how they’re expected to work differently.
That’s why manager enablement can’t simply mean sending leaders a different slide deck before everyone else gets theirs. It means equipping them to translate change to their teams’ everyday work.
Training isn’t an event
Another warning sign our data revealed was that almost one in four respondents (23%) say their organisation has never provided AI training or a mandatory policy refresh. Only 24% have received one within the past month.
For a technology that’s evolving so fast, a one-off training session isn’t going to cut it.
Sustainable AI adoption requires repetition. People need opportunities to practise, ask questions, share what is and isn’t working and adapt as the tools and expectations around them change.
New behaviours are more likely to stick when they’re reinforced through everyday routines, peer learning, feedback loops and visible leadership.
As Sarah describes:
“AI enablement can’t be a one-and-done training exercise. The organisations that make adoption stick will build it into the rhythm of work: giving people opportunities to practise, share what's working, ask questions and keep adapting as the technology changes.”
The ultimate goal is to build a culture of continuous and consistent learning across your organisation.
Making AI adoption stick
AI adoption is an ongoing behaviour change and, while organisations have made significant progress putting AI into people’s hands, now comes the harder part: making responsible, effective use part of how work actually gets done.
That means practical guidance instead of policy alone. Capability building instead of one-off training. Feedback and reinforcement instead of assuming adoption after launch. And managers who have the skills, clarity and backing to coach their teams through all of it.
Get in touch.
We'll Get back in touch soon.