Your people are already using AI
By Sascha Cutura

Artificial intelligence is already becoming part of everyday work. Employees are using tools such as ChatGPT, Microsoft Copilot, Gemini and AI features built into the software they already open every morning. They draft emails, summarise documents, analyse data, support customers and tidy up workflows, often without waiting for a formal programme.
That is useful. It is also a responsibility. If people are using AI at work, they need to understand how to use it safely and well.
The gap opens quietly
Some organisations have a formal adoption programme. Many are still working out their approach. In plenty of teams, people are already experimenting before policies, training or governance have been agreed.
The government's AI adoption survey for 2026 underlines how quickly this is spreading. Of UK businesses using or planning to use AI, 31% expect more than half of their staff to be using it within one to two years, compared with 22% who said this was already the case. Among businesses that had adopted AI, 67% said limited skills, expertise or knowledge was a barrier. Among those planning to adopt, the figure was 78%.
People will try the tools. Trouble starts when they do so without a shared view of the limits. Common problems include bias, poor data handling, over-reliance on AI outputs, inaccurate information, a lack of transparency, and confidential material being entered into tools that have not been approved for business use.
Organisations can keep moving with AI. They also need to equip their people properly.
Awareness first, then judgement
Early training has a modest aim: enough awareness and judgement for people to use these tools well in their actual jobs, rather than a new specialist qualification for every employee.
That usually means a clear introduction to what these systems can and cannot do, how generative tools differ from predictive ones, and what to watch for in a workplace setting: privacy, bias, over-confidence in the output, and when to check a result with a person or a source you can stand behind.
A simple set of principles helps. Fairness, accountability, sustainability and transparency is one practical frame. People remember it, and it gives managers a shared language when something looks off.
Train for the role, not for a generic audience
A single slide deck for the whole company rarely changes behaviour. The questions a finance lead asks are different from those of a customer service adviser or a department head.
Start with the people closest to the first use case, and with the leader who owns it. They need enough confidence to use the tool, enough caution to check it, and a clear route for escalation. A wider programme can follow once that first group is settled.
Keep the sessions practical. People should leave knowing what they may use, what must stay out of a prompt, and what good looks like in their own work. Labs and real examples beat a long theory module.
Training belongs in the adoption plan from the start
It is tempting to roll out a tool and "do the training afterwards". That is usually how unofficial use takes hold, and how weak checks become normal.
A more reliable sequence starts earlier. Inventory what is already in use, including free and embedded tools. Agree a short policy people can follow. Train the first cohort. Then widen. That is also why skills belong in a 90-day adoption plan alongside governance and a first delivery, rather than in a separate workstream that never quite starts.
The organisations that get value from AI will give people access to good tools. They will also help those people use the tools with care, confidence and judgement. That work can begin now, even if the wider programme is still being designed.



