Across Africa, there is no shortage of AI pilots. PwC’s 2026 research found that 82% of organisations are running them, while 64% of workers are already using AI. However, few organisations have managed to scale that activity across the enterprise.
At the start, people will try all sorts of things simply because the tool is there. What I pay more attention to is what they keep coming back to on their own. That tells you far more about whether it has found a useful place in the working day.
I would not invent exercises for an AI pilot. There is already enough to do in a normal working week. If someone comes back after two days out, opens a crowded inbox, and uses Copilot to get their bearings, that gives us something concrete to watch.
That kind of task is useful because the person already knows what doing it normally involves. They can tell fairly quickly whether Copilot has saved them time or simply given them something else to check.
The signal I pay most attention to is whether somebody reaches for the same capability again without being reminded. A clever prompt can create a good first impression. Repeated use tells you the technology has started to fit into the workday.
Selecting the right group
An evaluation made up entirely of enthusiastic AI users starts with a built-in bias. They will usually find something interesting to do with the technology. I would rather involve people who already depend on one another to get work done. Their work passes back and forth during the day, giving the trial a more realistic test of how the technology fits into how the team actually operates.
People will settle into using AI at different speeds, and that variation is useful. It gives the business a better view of where people may need support instead of allowing the most confident users to define what successful adoption looks like for everyone else.
A pilot should make uncomfortable things visible
Workplace AI can make old information housekeeping harder to ignore. A user may have access to documents they no longer need, or a SharePoint site may have accumulated broader permissions over time. That access may have existed for years without attracting much attention. AI can make the information much easier to surface during everyday work.
Security also needs attention before people start experimenting. Microsoft advises companies to check MFA and Conditional Access and also take a proper look at who can see what in SharePoint. Old permissions and overshared content are much easier to overlook until an AI tool starts surfacing information people already technically had access to.
If that work uncovers weak permissions or poor information practices, the pilot has already taught the organisation something useful. There may be housekeeping to do before wider adoption makes sense.
Evidence should decide what happens next
At the end of an evaluation, I would want a clearer picture of who kept using the technology, which pieces of work drew them back, how much checking the output still needed and what controls required attention.
McKinsey made a similar observation in its 2026 work on AI transformation. Its research argues that individual productivity gains rarely translate into lasting enterprise value when the organisation around them stays unchanged. More people using AI can be encouraging, but usage becomes much more useful when it connects to how work is actually being done.
This thinking is shaping how we approach Microsoft 365 Copilot Business evaluations at IPT. Microsoft currently offers eligible businesses a fixed 25-user, 30-day Microsoft 365 Copilot Business trial, and its customer guidance centres the experience on real work rather than practice exercises.
A month gives people enough time to get beyond their first experiments and see what they actually return to. By then, the pattern is likely to be uneven. A few uses may have found their way into the working day, while things that seemed impressive at the start have quietly fallen away. That is useful information in itself.
The same exercise can also expose work still needed around permissions, support, or user confidence. If that means a broader rollout should wait, the trial has still done its job. It has given the business a better basis for deciding what comes next.
A worthwhile AI pilot leaves the business with a clearer picture of where the technology earns its place in the working day. That is a better basis for the next investment than another adoption dashboard full of logins.
For more information, go to: https://ipt.za.com/lp-work-
