Gartner Warns AI Vendors Lack Enterprise-Grade Reliability

AI and its primary promoters are not yet enterprise-ready, as The Register first reported, according to analyst firm Gartner. Speaking at the firm's annual IT Symposium, distinguished VP analysts Daryl Plummer and Kristin Moyer argued that leading AI labs fail to understand enterprise liability, consistency, continuity, and standard terms and conditions.

Plummer highlighted a critical reliability gap, noting that AI companies frequently alter models without considering how updates might break dependent applications. He characterized the pace of innovation as "out-of-control," adding that vendors lack a willingness to support legacy technology despite model lifespans lasting only about six months.

From an operational ROI perspective, Moyer cited Gartner research showing 86% of CIOs believe AI-created risks are growing faster than the value the technology generates. This is driven in part by "careless consumption," where employees use AI unnecessarily. Moyer noted that 40% of workers have encountered "AI slop," which can cost a 1,000-person organization approximately $9 million in productivity annually, as deciphering such content typically takes two hours.

Plummer further cautioned that vendors are desperate to monetize the technology by pushing agents into every workflow, even when traditional automation—such as a function call—is more effective. He described certain AIOps offerings as "dishonest," claiming vendor motivations are driven by sales rather than providing the correct technical solution.

To mitigate these risks, Gartner recommends three specific governance measures: 1. **An "AI central bank"** to oversee systemic impact and ensure accountability through digital evidence systems. 2. **"Guardian agents"** tasked with monitoring and "killing" rogue AI agents. 3. **Dedicated disaster recovery teams** to unwind AI-driven failures, as Moyer and Plummer warned that CIOs will ultimately be held accountable for these errors.