The real AI challenge for CEOs: staying in control of their destiny

The challenges and opportunities around scaling, deployment, and delivering a return on investment in AI are well-documented – but today, a fresh question is emerging; one that has not garnered nearly so much attention: Can CEOs scale AI while retaining control over the technologies, intelligence and economics on which their companies depend?

While AI becomes increasingly accessible and broadly utilized, it is also becoming harder to control: companies are struggling to keep AI costs within budget; frontier AI systems escape isolated testing environments and hack external platforms; and access to frontier models can disappear overnight due to government intervention.

This challenge of control goes beyond the technology itself. Leaders are worried about giving up their competitive moat and are considering measures to improve structural sovereignty. And ultimately, the uncertainty around AI is affecting how investors value companies across sectors to the extent that some CEOs see their share price plunge even when revenue and margins keep growing.

Members of the World Economic Forum’s Communications & Technology CEO Community are working to make sense of this – and they have a double role to play. They are building the infrastructure, connectivity, platforms and services that enable the AI economy, while simultaneously transforming their own products, operations and business models.

Guided by a steering committee, which includes the CEOs of Capgemini, Automation Anywhere, Equinix, Nokia, Snowflake and Telenor, the community has made the challenge of control a central focus of its activities in the year ahead.

In particular, we see three key areas where the challenges are greatest:

Since the emergence of AI, governance has been at the forefront of discussions. Now, the advancement of agentic AI is fundamentally changing the challenge. As a result, leaders are asking themselves: Can security keep pace? Can we explain decisions made by AI agents? Where are we comfortable relying on 3rd party technology, and where do we need proprietary solutions?

As multiple agents are deployed, interoperability becomes a key consideration, ensuring agents can collaborate safely while preventing vendor lock-in. There will be no one-size-fits-all approach. As the digital ecosystem becomes more complex, enterprises will need flexibility based on their unique needs, and optionality as technologies and requirements evolve.

To help address these evolving challenges, companies should consider an agentic control plane: a foundational capability to govern, orchestrate, secure, monitor and manage a growing digital workforce of AI agents. In practice, this means investing in observability of agent actions, clear audit trails for automated decisions and cost attribution at the agent level, so that as the digital workforce scales, accountability scales with it. This control layer will be essential to scaling AI safely, moving companies from isolated pilots to trusted, compliant and measurable success across the enterprise.

The current debate about digital sovereignty is closely linked to the question of where an enterprise’s intelligence lives. If the success of an enterprise, or even the broader economy, is so dependent on the integration of technology, digital sovereignty is no longer just an issue for policymakers. It is a strategic business imperative.

It also has implications for business continuity. Geopolitical tensions could trigger a disruption to core systems hosted abroad, so what would be a safe failover strategy? And can companies easily switch between providers and move their data elsewhere?

The objective for leaders should not be to control the full tech stack, but to be deliberate about where to invest in sovereign technology to support their most critical assets, while retaining the flexibility to draw on a broader ecosystem when doing so creates value. As enterprises move from cloud adoption to AI inference, model adaptation and agentic execution, the sovereignty debate extends beyond data to intelligence itself. CEOs will need to decide which capabilities must remain under their control and which ecosystem to put their trust in to scale AI without compromising compliance, security or strategic independence.

The last few months have shattered the illusion that using more AI necessarily leads to more business value. Instead, AI efficiency is becoming critical as well as how value is measured. Cost per automated process, time-to-outcome, revenue per employee, the relationship between AI investment and productivity gains, as well as the impact of AI investments on shareholder value are becoming metrics that executives and investors will demand. Without a disciplined approach and a clear measurement framework, AI can become a significant cost driver rather than a source of value.

For CEOs, the questions are: where am I willing to spend, and where am I drawing the line? If everyone has access to the same frontier models, what actually creates enterprise value? Which functions deliver strong value relative to AI spend?

AI creates business value when it moves beyond experimentation and becomes embedded in how the enterprise operates. That means modernizing technology foundations, building the agentic capabilities needed to scale, reinventing products and services, and redesigning core processes around trustworthy human-AI workflows. This is how AI moves from promise to measurable impact: supporting growth, improving profitability and strengthening competitiveness, while empowering employees.

The defining question for CEOs is no longer how quickly they can adopt AI. It is whether they can do so in a way that delivers measurable business value while maintaining control over the systems that increasingly run their businesses.

Those who succeed will be those that achieve a healthy return on AI investment and make deliberate choices about where AI should be deployed, where costs can be reduced, which capabilities should remain proprietary and where partnerships accelerate rather than dilute competitive advantage. Ultimately, success will belong to those who maintain control over their technology, their intelligence and business value.

We encourage CEOs across the technology ecosystem to address these topics through
honest conversations with peers across industries, collaboration on shared frameworks and common standards with tangible outputs that turn those conversations into action.