Tech investor Prosus is stress‑testing artificial intelligence at massive scale. The Dutch company now runs roughly 60,000 AI agents—autonomous systems that execute tasks, run analyses, and automate processes. Yet only a tiny fraction actually pays off.
That’s according to Euro Beinat, Prosus’s head of AI, in an interview with the Financieele Dagblad. He says about 2% of the agents deliver meaningful cost savings or new revenue. Still, that small set of winners more than offsets the thousands of less effective experiments, Beinat argues.
Prosus’s approach shows how big tech is increasingly treating AI as a large‑scale innovation bet. Rather than preselecting “winners,” the company runs thousands of agents in parallel and then doubles down on the few that create real economic value.
Unlike traditional chatbots, AI agents can operate independently without constant human oversight. They can automate customer service, trigger commercial actions, or produce internal analytics. Prosus deploys these systems across multiple companies in its global portfolio, including e-commerce, fintech, and education platforms.
Beinat says the biggest gains don’t come from a single breakthrough model but from a broad spread of trials where a small share performs exceptionally well. That mirrors the wider playbook in big tech: scale and rapid iteration drive innovation.
At the same time, Prosus concedes the current scale is hard to sustain. The company plans to cut the number of AI agents in the coming years as infrastructure and compute costs surge—especially for agents that run continuously or handle complex tasks, which demand heavy cloud resources and specialized AI chips.
The remarks stand out because few large companies openly share how many AI projects fail. Today’s AI narrative skews toward wins and growth stories, while many experiments never turn profitable.
Prosus’s numbers suggest the AI sector is shifting from pure experimentation to return on investment. Companies are scrutinizing which applications truly boost productivity or drive new revenue. At the same time, Prosus’s strategy underscores that scale itself may become a key competitive edge in AI.
For companies and investors, Prosus offers a more grounded view of the AI market: most experiments flop, but a handful of hits can still generate outsized economic value.
Loading