What World Economic Forum leaders revealed about trust, value and the future of enterprise AI
7 mins
At the World Economic Forum Annual Meeting in Davos, artificial intelligence was framed less as a possibility and more as a system already shaping enterprise reality.
Leaders spoke about AI as already being present in enterprise environments, shaping how work is coordinated and how decisions are made. Much of the discussion focused on how these systems are deployed and governed once they move beyond controlled settings.
That shift matters. As AI becomes more embedded in day-to-day enterprise use, expectations around reliability and governance increase alongside it. Across multiple sessions, leaders returned to a shared concern: AI capabilities are advancing faster than governance models and operating practices are maturing.
Chris Lehane, Chief Global Affairs Officer at OpenAI, described this imbalance directly, noting that institutions, regulation and governance frameworks are struggling to keep pace with the speed of AI development. Salesforce CEO, Marc Benioff, raised a similar warning, arguing that insufficient oversight risks undermining trust and could lead to real societal harm.
Enterprise AI has reached a turning point. The question now is whether it can be governed and trusted at scale. At Davos, leaders returned to seven connected signals shaping how the next phase of AI adoption will unfold.
1. AI has moved beyond pilots, increasing exposure
For much of the past decade, enterprise AI efforts focused on pilots and proofs of concept. These initiatives allowed organizations to explore potential use cases while keeping risk contained.
That containment is narrowing.
At Davos, leaders from enterprise software and technology companies described AI being used inside core systems that support finance, workforce management and enterprise data platforms. Carl Eschenbach, CEO of Workday, spoke about AI agents operating within environments that already underpin how organizations manage people, money and performance. Judson Althoff, CEO of Microsoft’s commercial business, described similar momentum as large enterprises adopt platforms that allow AI agents to work across data stored in multiple systems.
When AI operates at this level, errors are no longer isolated to experimental teams. They can affect broader systems and business outcomes.
As a result, AI risk now extends beyond technical performance. Failures can carry operational consequences and reputational impact, with regulatory scrutiny following close behind. Reliability and governance are becoming baseline expectations for enterprise deployment.
2. Capability is outpacing governance, shifting responsibility inward
A recurring theme across Davos discussions was the growing gap between technical capability and governance.
Leaders acknowledged that public policy and regulation are evolving, but not at the same speed as AI development. Cristiano Amon, President and CEO of Qualcomm, and Dina Powell McCormick, President and Vice Chairman of Meta, pointed to the rise of edge AI as a complicating factor. As AI systems are built and deployed closer to users and devices, traditional control structures become harder to apply.
In this environment, enterprises cannot depend on regulation alone to manage risk. Responsibility is moving inward.
AI is also being applied in more sensitive contexts. Systems increasingly handle proprietary data, support business decisions and generate outputs with legal or financial implications. Demis Hassabis, CEO of Google DeepMind, noted that enterprise customers are pressing AI providers to offer clearer assurances around safety, security and predictable system behavior, particularly when sensitive data is involved.
The implication from these discussions was consistent. Governance needs to be designed into AI deployment from the outset, with clear ownership and oversight embedded into how systems are used.
3. AI orchestration reflects a growing need for control
One of the more concrete developments discussed at the Forum was the growing focus on AI orchestration.
Enterprise software leaders described increasing demand for platforms that coordinate how AI agents access data and operate across systems. Eschenbach positioned Workday as a potential “front door to work,” drawing on its existing role in managing HR and financial data, permissions and performance for human workers, and extending that approach to AI agents.
Althoff described similar momentum at Microsoft, where large enterprises are adopting data fabric and agent platforms that provide unified access to information spread across multiple systems, reducing the need for wholesale data migration.
These conversations pointed to a broader challenge. As AI agents become more common, organizations need ways to maintain visibility and accountability. Orchestration is not only about integration. It is about understanding how decisions are made and where responsibility sits when systems act on behalf of the business.
Without clear ownership and governance, scale introduces complexity rather than control.
4. From ambition to proof, AI value remains uneven
Despite sustained investment, leaders at the Forum were open about the uneven returns many organizations are seeing from AI.
Mohamed Kande, Global Chair of PwC, highlighted a widening gap between AI ambition and execution, noting that foundational work is often underestimated. Expectations around impact frequently run ahead of process maturity.
Several leaders also addressed inflated assumptions about near-term breakthroughs. Demis Hassabis emphasized that artificial general intelligence remains years away, reinforcing that near-term value must come from practical applications. Yann LeCun, former Chief AI Scientist at Meta, echoed this caution, arguing that today’s large language models will not achieve human-level intelligence without a fundamentally different approach.
Dara Khosrowshahi, CEO of Uber, grounded the discussion in execution. He said companies often overestimate the gains from applying AI to existing processes. In Uber’s experience, stronger results came only after moving away from rigid, rule-based systems and redesigning how work was done.
Across these perspectives, the message was consistent. Incremental change produces limited results. More significant gains depend on rethinking how work is structured and governed.
5. Enterprise demand is reshaping the AI market
Another signal from Davos was the continued shift toward enterprise-driven AI adoption.
Sarah Friar, Chief Financial Officer at OpenAI, said enterprise customers already account for a substantial share of the company’s business, with that proportion continuing to grow. Dario Amodei, CEO of Anthropic, described a similar dynamic at his company, attributing roughly 80% of Anthropic’s business to enterprise customers, with the remainder coming from consumer use.
Enterprise buyers are looking for systems they can rely on in regulated environments. Questions around data handling, validation and predictable behavior feature prominently in buying decisions. Hassabis noted that enterprise customers are increasingly pressing providers for stronger guarantees in these areas.
Christoph Schweizer, CEO of Boston Consulting Group, observed that this shift is also changing leadership dynamics. He noted that 72% of CEOs now say they must personally lead AI transformation, up from 36% a year earlier, with around half believing their job depends on whether they successfully bring AI into the business. As AI outcomes move onto board agendas, governance becomes a differentiator rather than an obstacle.
6. Jobs and skills remain a central concern
The impact of AI on jobs and skills featured heavily in Davos discussions.
Leaders emphasized that entry-level and early-career roles are particularly exposed. Kristalina Georgieva, Managing Director of the International Monetary Fund, warned that AI could affect around 60% of jobs in advanced economies and 40% globally, describing its impact on the labor market as a “tsunami.”
There was also broad agreement that productivity gains depend on changes to how work is organized. Julie Teigland, Global Vice Chair at EY, stressed that deploying tools alone delivers little return if organizations are unwilling to redesign roles and invest in training. Ravi Kumar, CEO of Cognizant, reinforced that reskilling needs to be treated as core infrastructure, integrating human and digital labor.
Abhijit Dubey, CEO of NTT Data, highlighted the challenge of transition. Workers are likely to feel disruption before new roles emerge, and neither companies nor governments are fully prepared to support people through that period. Without planning, organizations risk eroding trust at a time when confidence is already fragile.
7. Economic uncertainty is sharpening expectations
Economic uncertainty formed the backdrop to many of these conversations.
Christine Lagarde, President of the European Central Bank, described an environment shaped by weakened trust and ongoing volatility, complicating planning and investment decisions.
Georgieva noted that while growth has held up, it remains fragile, increasing pressure on leaders to show clearer returns. Ngozi Okonjo-Iweala, Director-General of the World Trade Organization, warned that trade disruption is likely to persist.
In this context, AI is being evaluated on its ability to improve day-to-day performance under real-world conditions. Enterprises are prioritizing systems that can be relied on as uncertainty becomes a constant feature of the operating environment.
What enterprise leaders should take away
Taken together, the signals from the World Economic Forum point to a decisive phase for enterprise AI.
Adoption is accelerating. Governance models continue to evolve. Value creation remains uneven, while accountability is rising.
Organizations that navigate this phase successfully are likely to be those that apply human judgment where it adds value, define ownership early and embed context into how systems are used.
Across the Forum, one practical implication stood out. Organizations need ways to test AI in real operating conditions before committing to scale. Structured prototyping helps teams move from ambition to evidence, reducing risk while clarifying where AI can deliver durable value.
Across all seven takeaways, the message from Davos is consistent: enterprise AI is now judged by how well it is governed, proven and sustained in real operating conditions.
This is where enterprise AI starts to earn trust in the real world.
Need help turning AI ambition into enterprise-ready execution?
Talk to our experts about building AI systems you can trust – across markets, languages and real-world complexity.