Group update

Dive into our worlds - Donné Stevenson

AI for the real world

I was working at OLX, one of the companies in the Prosus ecosystem, when I first saw the Prosus team testing early GPT-3.5 inside a Slack assistant called Toqan. It could read files, search the web, and transcribe - all in one place, not over 5 interfaces. It wasn’t a demo either; people were actually using it. I moved from the OLX portfolio to the Prosus AI team because I wanted to build more of that: AI that earns its place by being useful.

Our core AI team at Prosus has grown to around 35 specialists – a mix of engineers, researchers, and product experts. Challenges are always evolving, but the collaboration within the team stays constant. You’re trusted to try, learn fast, and try again - with enough users and data to make every experiment count and the right backing to scale what works.

How we build for real users

We identify the problems we want to solve and the experiences we want to improve, then find the best way to do that. If GenAI is the right tool, we use it - but the goal is always impact, not novelty.

We prototype fast, wire in the unglamorous parts that make it solid (speed, quality checks, fallbacks), and then we test. That validation step is what turns “we used an LLM” into “we built something people rely on.”

“We want to avoid AI being the hammer so everything’s a nail.”

Our rhythm is simple: start with the job to be done and what could go wrong, spin up the smallest version that works, wrap it with measurement - quality, latency, fallbacks - then put it in front of real users. Feedback is fast, honest, and at scale.

The power of the portfolio

Prosus doesn’t ship a single product - it’s an ecosystem. Our AI team plugs straight into portfolio companies’ data, product and engineering teams. On good days it feels like one team - shared stand-ups, shared code, shared wins.

At iFood, colleagues ran experiments on personalised push notifications (timing and content) with LLMs in the loop. Rather than overhauling entire systems, the focus is on thoughtful personalisation and proper evaluation. The result was a real uplift in completed orders in a place where moving the needle is hard. Equally important were the ideas that didn’t land the first time - we cut them, kept the learning, and moved on. That’s real adoption - when AI is part of the daily rhythm, not a side project.

Across the wider Prosus portfolio, we have more than a thousand AI specialists sharing learnings, approaches, and practical applications that make the whole ecosystem smarter.

Culture that keeps speed safe

You’re expected to own your work and ask for help when you need it. We move quickly, and when something breaks we talk about it honestly and fix the process so the next person moves faster. And the best part is: you’re never doing any of it alone.

There’s a real team bond here,  the kind where people jump in without hesitation, whether it’s a late-night bug or a half-baked idea that needs pressure-testing.

I love the spontaneous “we need a decision now” calls that unlock a problem. Those moments only work because the trust is real. That’s building with backing — trust, tools, autonomy — and platforms to test with real users across markets. When an idea misses in one product, it often lands in another. Because we can test at scale, we don’t spend months guessing. We move fast together, and that’s what makes the speed sustainable.

Right place, right moment

What excites me is being in the middle of what AI becomes - how it reshapes everyday products - with people who are pushing that change. Not watching from the sidelines. At Prosus you don’t just talk about AI - you build it, measure it, and make it useful. That’s how we’re unlocking an AI world for billions.

In the middle of what AI becomes - not watching from the sidelines.