Enterprise AI Gets Real as Davos 2026 Focuses on Agents 

That shift is showing up not only on panels, but in how Davos itself is being run. In one of the most concrete “AI-in-production” proofs at the event the Forum and Salesforce rolled out an agentic concierge called EVA for the event, positioned as a step beyond traditional chatbots into “doers” that can prioritize and act on a participant’s behalf (from managing agendas to generating briefing documents).

Salesforce CEO Marc Benioff emphasized that EVA is “far more than a chatbot,” aligning the deployment with his broader thesis that the “agentic enterprise” is a new enterprise architecture rather than a feature add-on.

On stage, the message from the largest enterprise platforms has been similar: AI’s next phase hinges on measurable outcomes and broad adoption. In a Davos conversation highlighted by the Forum’s live coverage, Microsoft CEO Satya Nadella argued the global priority is putting AI to work in ways that “change the outcomes of people and communities.”

That focus on “production value” over prototypes also runs through the Forum’s own research cadence this week. A WEF release on scaling AI cites examples of organizations moving beyond pilots and includes Foxconn and Boston Consulting Group scaling an “AI agent ecosystem” to automate 80% of decision workflows and unlock an estimated $800 million in value.

For financial services leaders, the most consequential Davos-adjacent AI storyline may be the emerging rails for agent-led shopping and payments. Mastercard is explicitly positioning itself as the infrastructure and rules layer for agentic commerce, arguing that the competitive battle isn’t only about model performance but about trust, identity, and secure authorization when software agents spend money.

“Agentic commerce will only scale at the speed of trust,” Mastercard executive Sherri Haymond told Axios.

Mastercard’s roadmap matters for banks and fintechs because it points to where the industry may standardize. The company says its Agent Pay is being integrated into Microsoft’s Copilot Checkout and OpenAI’s Instant Checkout program in ChatGPT, with an emphasis on intent-verified payments embedded in AI shopping flows.

If “trust” is the headline, security and governance are the subtext in nearly every conversation about autonomy. In a Davos dispatch, EY’s Raj Sharma captured the core enterprise concern with agentic systems: “It has no name,” he said, describing the identity and lifecycle-management gap for agents with access to sensitive data.

KPMG US CEO Tim Walsh, in the same report, pointed to why CISOs are dragging timelines: firms are pausing to harden environments and, in some cases, keeping data “onprem” longer. And looming behind that is the next security discontinuity: “Quantum breaks everything,” Walsh warned.

WEF’s own content track has echoed the identity-and-fraud risk. A Forum article on AI agents argues that any agent economy worth having must sit on “bulletproof KYC,” because agent identity is only as trustworthy as the human or organization behind it and without strong identity frameworks, “good bots” and fraud bots become indistinguishable at scale.

Even as the week’s AI talk gets more operational, Davos speakers keep returning to human agency and shared guardrails. In her first interview as Meta’s new president and vice chairman, Dina Powell McCormick called AI a “group sport” and urged cooperation so “humanity” stays at the center. And in a separate Davos session recap, Accenture CEO Julie Sweet distilled the governance stance in a single line: “It’s human in the lead, not human in the loop.”

For banks, payment networks, fintechs and enterprise software providers, Davos 2026 is converging on a practical thesis: agentic AI is inevitable. But the winners will be those who can prove identity, authorization, auditability and resilience as these systems begin to initiate real-world financial actions.