Gartner: 70% of vendor-built AI agents abandoned by 2028

Gartner's forecast sits on a shift in how software vendors package AI work. The article explains that FDE originally meant vendor engineers embedding at a customer site to jointly develop services tailored to that customer's problems. But Mukul Saha, a senior director analyst at Gartner, points out the term is now widely used as a label for implementation, professional services, solution engineering, or AI consulting.

Gartner therefore frames the risk as structural rather than technical. It says many projects will fail structurally before they fail technically, because they do not meet its three FDE principles: using FDE only for problems that require deep product expertise, rapid adaptation, and tight integration of the vendor's technology and operating environment; embedding FDE into internal business experts, engineers, and end users so important knowledge is shared and reflected in the work; and clarifying the exit strategy from the start.

The prediction that under 20% of vendors will reflect ongoing customer needs into their core products suggests the benefit of FDE may stay one-off rather than compound. For enterprises commissioning agent projects, the stakes appear to hinge on whether the engagement is genuine FDE or relabeled consulting, and on whether a clear exit strategy exists from day one.