The challenge is turning an investment in AI into a way of working that people find useful enough to keep.
Many South African businesses have moved beyond debating whether to use AI. They have bought licences, launched pilots and spent time developing strategies, yet everyday use often remains concentrated among a small group of enthusiasts while the rest of the organisation carries on much as before.
There are understandable reasons for this. People have deadlines to meet, established ways of working and little spare time to experiment with something whose value they have yet to see. Giving them access to a tool does not automatically give them a reason to change.
This is where the work of adoption begins, and where leaders need to pay closer attention if their investment is going to deliver.
Bringing the strategy into everyday work
My approach to AI follows a three-phase model: analytics builds understanding, AI adds intelligence and automation scales impact. It gives an organisation a clear direction and helps connect individual initiatives to a broader business purpose.
Putting that model into practice means understanding how people spend their time and where they encounter difficulties. For someone preparing a monthly report or working through a backlog of customer queries, the immediate concern is whether AI can help them complete the task well, with less effort and within the time available.
Adoption often begins with someone trying AI on a task they know well enough to judge the result. They might find a quicker way to prepare a first draft, make sense of a set of documents or deal with a recurring administrative frustration. If the approach proves useful, they use it again and may share it with a colleague.
Over time, those experiences give the strategy substance. People begin to understand where AI fits into their work because they have seen it help.
What helps people keep using AI
Running AI enablement across our business has taught us that the most useful starting point is usually a problem people already want to solve. A session built around “what is slowing you down this week?” gives people a reason to participate and a practical way to assess what they have learned.
Feature demonstrations have their place, but people still need help applying those features to their own responsibilities. That takes time at the keyboard, working with familiar tasks and asking questions as they go. It also means encountering answers that are incomplete or wrong, then learning how to improve them and when to question them.
That experience matters because using AI well requires judgement. People need to know what a good result looks like, what needs checking and whether the effort involved is justified by the outcome.
Sharing work helps others develop that judgement too. A colleague explaining how they approached a task, what worked and what they had to correct gives people something they can relate to and try for themselves. Showing the rough edges can be as useful as showing the finished result, particularly for someone who assumes everyone else already knows what they are doing.
Clear guidance also makes a difference. If people are unsure which tools are approved or what information they are allowed to use, hesitation is a reasonable response. Practical rules, supported by someone they can turn to for advice, give them the confidence to begin.
All of this needs continuity. Regular sessions, opportunities to bring a difficult task and time to exchange ideas help people keep learning after the initial excitement has passed. Managers play an important part by making room for that learning within the working week.
Where adoption loses momentum
Once-off workshops can generate enthusiasm, but people often return to their desks without the time or support to apply what they have seen. As other demands take over, the intention to experiment slips down the list.
A mandate to “use AI” creates a similar problem when it comes without a clear purpose. People may comply by trying a tool occasionally, but that tells us little about whether it has become useful to them or improved their work.
The measures we choose can obscure this distinction. Login counts and usage figures can help show where interest is growing, yet they cannot tell us whether a report is better, a process is quicker or a team has reduced the amount of rework it does.
Adoption also becomes harder when it is treated primarily as a technical responsibility. IT teams provide essential expertise on platforms, access and security, while the people doing the work understand its demands and where improvements would matter. Both need to be involved in deciding how AI is used.
The influence of a colleague
One of the strongest influences on adoption is seeing someone you work with use AI successfully on a task you recognise. It makes the benefit easier to assess and creates an opportunity to ask the practical questions that a formal demonstration may leave unanswered.
These early adopters can emerge anywhere in the business. Often, their value comes from knowing their work well and being curious enough to try a different approach. They understand the frustrations, the exceptions and the standards a useful result has to meet.
Finding and supporting these people gives an organisation a valuable source of learning. They need time to help colleagues, opportunities to share their work and recognition for the contribution they make. Expecting them to take on that role entirely around the edges of their existing workload makes it difficult to sustain.
Their experience can also help leaders see where adoption is encountering obstacles, whether those involve access, uncertainty, skills or a tool that simply does not suit the task.
Looking for evidence of improvement
The question leaders should keep returning to is whether the work is improving. That might mean a recurring task takes less time, a report needs fewer corrections or a team can handle a process with fewer handovers.
The answer will vary across functions, so each team needs measures that reflect its own work. A useful assessment should also account for the time spent checking and correcting AI outputs; a task completed more quickly at the first attempt may still create additional work later.
Listening to the people using the tools helps make sense of those results. They can explain where AI has become useful, where it remains awkward and what would help them use it more effectively.
Making AI stick requires this kind of ongoing attention. Leaders need to connect it to business priorities, managers need to make space for practice, and teams need support as they work out where it earns a place in their day.
Buying AI is the easy part. Making it stick is the real work.
Kylan Pather is an AI and Automation Engineer at Ignition Group.
