AI Is Racing Ahead. The Brakes Aren’t Ready

Daryl Plummer reached for a racing metaphor. “We put brakes on cars so they can go fast safely,” the Gartner analyst told chief information officers, as IT Pro reported from the firm’s Symposium keynote.

Gartner has now taken that message to Orlando and Barcelona, where its IT Symposium/Xpo drew more than 6,500 CIOs and IT executives, according to CDO Magazine.

The numbers behind the metaphor reveal a widening gap between AI ambitions and oversight. A Gartner survey of more than 2,000 CIOs and technology executives, conducted in the second quarter of 2026, found that 71 per cent of organisations expect to deploy agentic AI within the next 12 months. Only 21 per cent currently have the governance capabilities to oversee it.

That leaves a 50-percentage-point gap between deployment plans and governance readiness. Gartner published the findings in a release dated November 9. The same figures appeared in coverage of the firm’s Orlando keynote on October 19, according to Techstrong. Barcelona therefore brought an existing warning to a second audience rather than unveiling new findings.

The consequences of poorly governed AI are already visible in workplace productivity. Gartner estimates that 40 per cent of workers received AI-generated “work slop” in a single month, with each spending about two hours correcting it. For an average S&P 500 company, the annual cost reaches $52 million, IT Pro reports.

Vendors, meanwhile, promote productivity gains of 50 per cent or more. Those figures are vendor claims, while Gartner’s own data points to gains of about 16 per cent.

Plummer also said only one in five AI projects succeeds, unchanged from last year. The successful projects now cost 18 times more than they did two years ago.

AI is nevertheless delivering some value. A separate 2026 Gartner C-Suite AI Survey of more than 1,000 functional leaders found that 58 per cent were seeing some positive value from AI initiatives. However, organisations are struggling to replicate successful projects and scale their benefits across the enterprise, CDO Magazine reports.

That distinction matters. Positive value does not necessarily translate into financial returns, Techstrong notes.

Gartner also found that 86 per cent of CIOs believe AI risk is growing faster than AI value, while only 25 per cent have aligned technology risk with business risk, according to National CIO Review’s account of the Orlando keynote.

Gartner’s proposed response is organisational before it is technical. The firm recommends establishing a central decision-making authority called the “Flywheel Governor”. Described as a virtual organisation, it would oversee the company’s AI strategy and provide a consistent framework for weighing value against risk without requiring every business function to apply identical standards.

National CIO Review reports that Gartner proposed a leadership role that could rotate between the CIO and chief financial officer, with the chief information security officer and legal counsel permanently involved.

The framework divides organisational forces into “momentum”, which accelerates outcomes, and “drag”, which slows them down. Gartner argues that not all drag is harmful. Data quality checks, access restrictions, audit trails and human review are examples of what CDO Magazine describes as “good drag”.

Kristin Moyer, Gartner’s co-presenter, explained the importance of anticipating risks before they materialise. “You can’t see the harm until after it happens,” she said, according to National CIO Review.

Plummer framed the challenge as a shift in how organisations measure AI adoption and performance. “Your task is to stop compounding volume and start compounding value,” he said, as IT Brew reported.

His message to CIOs was to judge AI by the business value it produces, rather than the volume of tokens it consumes.