Prosus has data from a billion customers. Now it wants to turn that into an AI advantage

Prosus is reorganising its technology leadership around a bet that the most valuable AI models in commerce will not be trained only on text.

The group has appointed Euro Beinat as chief AI officer, Paul van der Boor as chief technology officer and WeTransfer co-founder Ronald "Nalden" Hans as chief product officer. Their job is to turn Prosus's existing AI research and portfolio data into products that can operate across businesses such as iFood, OLX and PayU.

Prosus says its intelligence layer learns from the commerce behaviour of around one billion customers. At its Prosus Forward technology event earlier this year, the company said its wider ecosystem generated about 500 million daily interactions and served five million partners.

That scale makes the strategy more interesting than another corporate AI reorganisation. Prosus is trying to turn transactions, searches, reorders and other commercial behaviour into a proprietary advantage.

Commerce data is different from internet text

Large language models are mostly associated with enormous collections of text, code and images. Prosus's pitch is that commerce produces another kind of training signal: what people actually do.

A user can say they want something and still behave differently when money is involved. Searches, clicks, orders, cancellations, delivery choices and repeat purchases create data that can reveal intent in ways a text corpus cannot.

Prosus calls the resulting technology its Large Commerce Model, or LCM. It is meant to sit underneath agents and assistants across the group's ecosystem rather than exist as a standalone chatbot.

The ambition is that an AI system learns enough about commerce to help customers buy, help merchants operate and help portfolio companies make decisions with less manual intervention.

The billion-user claim needs careful reading

Prosus uses different customer counts in different corporate material. Its current group description refers to an AI-first world for more than two billion customers, while the technology-event material describes an ecosystem serving around one billion customers and five million partners.

Those numbers should not be treated as one unified database of identifiable people.

The important technical question is how data moves between portfolio companies. Does a model learn centrally from OLX, iFood and PayU information? Is training federated? Are datasets anonymised or aggregated? Can one business use signals collected by another?

Those questions matter commercially and from a privacy perspective.

South Africa sits outside the main operating footprint — but not the strategy

Prosus is based in the Netherlands and its largest consumer internet businesses are concentrated in Latin America, India and Europe. The company is nevertheless rooted in Naspers and remains one of the most important technology companies connected to the JSE.

For South Africa, the significance is less about whether a local user gets a Prosus AI assistant tomorrow and more about where the group's research, investment and product decisions flow next.

Prosus Ventures is actively backing AI, robotics, drones and other frontier technologies. If the group can prove that proprietary commerce data creates better AI products, that thesis is likely to shape what it buys and funds.

The next phase is product, not demos

The new leadership structure is a signal that Prosus wants to move from scattered portfolio AI projects into a more coordinated product strategy.

That creates a measurable test. Can the company show that its AI layer increases conversion, improves merchant margins, reduces support costs or creates entirely new revenue?

If the answer is yes, the group's one-billion-customer claim becomes a genuine competitive moat. If not, it is just another very large dataset looking for a product.

Prosus now has a dedicated AI chief, technology chief and product chief to answer that question.