Access to AI tools can be arranged in days. Using them well on the organisation's own work, within its policies, takes practice on that work, and practice takes working time. The time people get decides whether skill grows across the organisation or stays with the few who have time to spare. How often the time is given, and whether it is spent on real work, often matters more than the total.
On their own timeLink to this section
People learn to use AI in their evenings or between tasks. There is no direct cost, and skill varies widely: those with the least spare time, such as parents and carers, tend to learn the least. The organisation has no record of what people know, so the policies on information have to work with no AI learning at all.
Easy to miss. Home use of AI tends to be personal tasks on personal accounts, such as planning a trip or writing a letter, where little is confidential. It builds confidence, but judgement about what can go into a work tool comes from practice on the work itself.
When to choose it. It makes sense as a start for keen people, if what they find is gathered and shared.
Occasional sessionsLink to this section
People attend a session now and then, such as a demonstration or a talk from a tool's supplier. The cost is small and predictable, a few hours a few times a year, and sessions are easy to arrange, easy to report as done and a good place to explain the policies. Between sessions, people need somewhere to take questions about what is allowed.
Easy to miss. The examples in a session tend to be general, and people only work out how they apply to their jobs by trying them on real tasks. If they do not in the days after, other work takes priority, and when the need arises, they have forgotten the detail of what to ask and check.
When to choose it. Choose it as a starting point, with each session followed by a specific task people try on real work that same week.
A set programmeLink to this section
Everyone works through the same programme over a few weeks, so people share a vocabulary, a set of basics and the same policies in the same words. The cost is known, and completion is recorded, which helps when an insurer, auditor or client asks how people were taught. Quicker learners wait, some roles get content that doesn't fit their work, and content needs updating as the tools change.
Easy to miss. Completion records that people were taught, which says little about whether they now work differently. Without use on their own tasks, week after week, people forget what a programme teaches, as they do after a single session, only more slowly.
When to choose it. It is best when the programme ends with each person using AI on a real task, and a check a few months later on what is still in use.
Regular protected timeLink to this section
People get a set amount of working time each week or fortnight to try AI on their own work. The hours are the main cost, and they show in every team's week: an hour a week is a little under 3% of a 38-hour week, across everyone. In return, skill builds steadily, and the time is a natural place to learn new policies as they change.
Easy to miss. Protected time is often the first thing given up when a deadline appears. People follow their manager's example, and when the manager cancels it, the team stops taking it too.
When to choose it. It suits organisations where managers take the time themselves in view of their teams, someone checks it is taken, and there is a specific task to do in it, such as trying AI on this week's work.
Part of every roleLink to this section
AI learning is an expected part of every job, written into role descriptions and discussed in one-to-ones and reviews. The cost is ongoing time, hard to separate from the work, and skill stays current as the tools and the work change. Policies on information are learnt through daily work, which needs managers who know them well. People who find AI hard feel the expectation more sharply.
Easy to miss. The risk is one standard applied everywhere. A finance analyst and a site supervisor get very different things from AI, and a single expectation becomes a formality where it helps least.
When to choose it. It is worth choosing where each manager sets what good use looks like in their area, and reviews it with their team.
How to chooseLink to this section
- Which roles gain most from AI, and which least?
- How much working time can each team give in a normal week, and who decides when it is given up?
- Do managers use AI well enough to show their teams, and will they take the time themselves?
- How will people learn the policies on information, and how will the organisation know they have?
Tensions with other decisionsLink to this section
These apply when people learn On their own time or through Occasional sessions.
- A Quick pace or faster. Teams start before most people have had much time to learn, so skill varies widely and the same mistakes happen in team after team. Give people regular time to learn during the change, or slow the pace.
- A risk appetite of Balanced or higher. Each person decides what is safe to put into a tool, so each decision depends on whatever they have learnt. Add regular time to practise the policies on real work, or move the appetite towards clearer limits.
- Whatever it takes on investment. The money buys tools quickly, and renewals are due before most people know what the tools can do. Put part of the money towards time to learn, or delay purchases until people have learnt to use the tools they already have.
When to revisit itLink to this section
- When the pace or risk appetite changes, a new tool is introduced, or the policies on information change.
- When protected time is given up most weeks, a sign that what happens in practice no longer matches the agreed position.
If the leadership team would like support with this decision, see Getting outside help.
Something missing or out of date here? Tell us.