Welcome to The People Insider — a weekly brief for anyone who employs people in South Africa.

Every Wednesday: one headline, one tool, one leadership story, one stat and one question. Under five minutes. No fluff. Let's get into it.

Created with Tess Gouws, who helps teams get unstuck at Goya XD.

Back in 2023, 97% of business leaders told KPMG they expected generative AI to free their people from manual work and make them more thoughtful and creative.

It was a lovely idea. But, unfortunately, the evidence since has pointed the other way, and it's not because the tools are bad.

The ideas were good, they were just all the same

Wharton researchers ran a set of experiments asking people to invent a toy using a brick and a fan. The ideas that came back from people using tools like ChatGPT were good, often very good. They were just strikingly similar.

Nine participants, working independently, named their toy the Build-a-Breeze Castle. Across the pool, only 6% of the AI-assisted ideas were genuinely distinct from one another, while in the human-only group, 100% were.

The business version of the same problem shows up in the returns. A widely cited yet equally criticised (because of its small scope and methodology) MIT study of enterprise AI claimed that 95% of generative AI pilots delivered no measurable impact on profit.

The reason was rarely about which AI model they used, but rather how the organisation used it.

What's really happening here

AI is an extremely good executor. Give it a vision, and it'll map the most beautiful, intricate and efficient pathway to making that vision happen. What it isn't good at is origination.

Why? Think of how it's trained: enormous data sets that are, most likely, of varying quality. AI companies use what they call Preference Optimisation to align their models' output with human values, safety guidelines and general consensus — it's what helps them filter outliers from acceptable consensus data.

Positive as that sounds, the most creative ideas are often in the outlier category. Cut them out and all you're left with is the generalised middle, which is great for finding tried-and-tested processes, but not great sparks of inspiration.

It's why most advise thinking of AI as a tool: just like the paintbrush doesn't create the painting, real creativity can only come from people living out in the world, together, bumping into real problems and feeling the friction that the internet has already smoothed flat.

Which is actually good news for you

Everyone is now building from the same body of knowledge, with tools that increasingly think alike. Your competitors can buy every tool you have bought, on the same day, at the same price. What they cannot buy is your particular group of humans.

Spend a little less time thinking about the paintbrush and more about the hands you put it in.

The faster your business moves, the less you get to inspect. Decisions that used to come to you get made three levels down while you are in another meeting, and you find out afterwards. That shift changes what you should be hiring and promoting for, and most bosses have not adjusted.

Speaking at Executive Day in Johannesburg last week, MTN Group chief human resources officer Paul Norman made the point directly. In highly adaptable, fast-moving organisations, he argued, decision-making has to be distributed, and that changes the calculus entirely.

Competence is the easy part to assess. It shows up in a CV, a work sample, a reference call. Character only shows up in the moments you are not in the room, which is precisely why it matters more the bigger you get. You can supervise for competence. You cannot supervise for character.

The question worth sitting with: In your last promotion, did you weigh how the person behaves when nobody is checking as hard as you weighed whether they could do the job?

41%

is the share of South African employees who say they are thriving at work, in the SA findings of Mercer's Global Talent Trends 2026. The global average is 44%, but the share of people reporting they are thriving has fallen everywhere, from 66% in 2024 to 44% now — the lowest since Mercer started tracking it. So South Africa’s employees don’t just feel below average fulfilment at work; they feel below an average that has collapsed by 22 points in two years. Roughly six in ten of your people are getting through the day rather than doing their best work.

Source: Mercer, Global Talent Trends 2026.

Got an event SA employers should know about? Reply and tell us.

Know someone who employs people?
Forward this to one person who should be reading this.