The more AI an organization uses, the harder it becomes to control the technology or generate a return on investment, according to analyst firm Gartner, which delivered that sobering assessment at its annual IT Symposium in Australia this week. Distinguished VP analysts Daryl Plummer and Kristin Moyer argued that artificial intelligence and its leading proponents haven't reached maturity for enterprise deployment. The firm's research paints a picture of technology racing ahead of organizations' ability to manage it safely or profitably.

Gartner's research found that 86 percent of chief information officers see risks from AI growing faster than the value it creates, in part because early wins with the technology generate more demand than IT departments can safely meet. The firm also discovered that 40 percent of workers have run into AI-generated nonsense, and deciphering such content typically consumes two hours of time, which translates to $9 million worth of work across a year at a 1,000-person organization. AI model lifecycles now run about six months, yet vendors refuse to support legacy versions, leaving enterprises trapped in what Plummer called "an out-of-control pace of innovation" they can't afford to ignore.

Plummer said trust in leading AI labs isn't warranted yet, noting they're not enterprise grade and don't understand enterprise terms and conditions, liability, consistency, or continuity. He pointed to AI companies' practice of frequently changing their models with apparent disregard for whether those updates might break applications that rely on their output. Moyer described the problem of "careless consumption," where workers use AI frivolously or inappropriately, and warned that tracking AI use has become difficult because "AI is invading your enterprise inside products you already own." Plummer also criticized vendors for being "desperate to monetise AI" by selling products that include it or simply pushing token purchases, adding that AIOps tools claiming to offer one-click fixes are dishonest: "Their motivations are to market to you and get you to buy their stuff, not to get you to put in the right solution."

The report recommends several measures to tame AI deployment. Organizations should establish an "AI central bank" with responsibility to oversee use across the enterprise and ensure accountability, which conventional ERP or CRM applications handle easily through audit trails but AI often doesn't. Gartner also advocates for "guardian agents," AI tasked solely with monitoring other agents to ensure their behavior stays within intended bounds, with authority to terminate rogue systems. The firm suggests creating disaster recovery teams dedicated to unwinding messes made by AI, warning that "if you are called to account and your answer is 'AI did it', you are in trouble" because CIOs will be held accountable for AI failures. Plummer urged enterprises to remember proven automation technologies like function calls for database interactions rather than building agents unnecessarily, while Moyer suggested organizations should have some experience controlling careless consumption after spending a decade reining in developers' preference for the most powerful and costly cloud instances. The fragmented market for AI governance tools Gartner observed signals that the industry's maturity gap extends beyond just the models themselves to the entire ecosystem meant to make them safe. Enterprises now face a choice between sitting out a transformative technology wave or riding it without the safety infrastructure typically required for mission-critical systems.