LONDON: As global executives gathered in Davos to discuss how to scale artificial intelligence beyond pilot projects, one message stood out: the next phase of AI will not simply assist human decisions, but act on them, including spending money.

Speaking on Tuesday at a panel discussion titled “Scaling AI: Now Comes the Hard Part,” Visa CEO Ryan McInerney said AI is moving rapidly toward what he described as “agentic commerce,” where autonomous systems are empowered to search, select and purchase goods on behalf of consumers.

“In 2026, most of us will continue to shop on our AI platform of choice,” McInerney said. “But now we’ll be able to buy natively on the platform. The buy button will be there.”

McInerney said the bigger shift would come when consumers allow agents to transact independently, a change that would require new levels of trust across the payments ecosystem.

“For that to work, you need to trust your agent, merchants need to trust that the agent is really acting on your behalf, and your bank needs to trust that when it authorizes a transaction, you really wanted that to happen,” he said.

While McInerney outlined how Visa is preparing for AI to act autonomously, other panellists pointed out that letting systems operate without discipline could actually undermine trust rather than build it.

Aramco CEO Amin Nasser spoke about how disciplined deployment can preserve trust while generating real value.

He said scaling AI requires moving beyond experimentation and embedding the technology into core operations, with clear accountability for results.

“Everybody talks about AI and the impact of AI, but where is the value?” Nasser said.

He told the Davos audience that more than “100 AI use cases” at Aramco had moved from pilot to full deployment, contributing billions of dollars in verified technology value.

In 2023 and 2024, the company achieved $6 billion in technology-realized value, with more than half attributed to AI, and it expects to report $3 to $5 billion for 2025 once third-party verification is complete.

“Each use case is treated like a project, with a timeline, deliverables and impact,” he said, adding that third-party verification was used to validate results.

 FASTFACT

Nasser said that data quality, governance and subject-matter expertise, rather than algorithms alone, were the decisive factors in scaling AI.

 “If you don’t have the data quality, it’s garbage in, garbage out,” he said.

The contrast between AI’s future-facing promise and the operational discipline required to deploy it safely was echoed by Roy Jakobs, CEO of Philips, who spoke about the challenges of scaling AI in healthcare.

 “The real breakthrough will come when we rethink how humans and agents work together,” Jakobs said, adding that AI is already reducing administrative burdens and supporting clinical decision-making.

Julie Sweet, CEO of Accenture, said many companies remain stuck in pilot mode because they lack the organizational discipline to scale AI across the enterprise.

 “One of the biggest barriers to scale has been the lack of willingness to put value on this, to see it in the P&L (profit and loss statement) and embed it in leadership objectives,” she said.

The discussion showed a shift in how executives are thinking about AI.

As AI systems move closer to autonomous action, whether in payments, industrial operations or healthcare, the challenge is no longer technical capability, but trust. Who controls AI systems, how they are governed, and how their impact is measured, the audience heard.