40% of AI Agent Projects Could Be Dead by 2028
AI agents are scaling 8x faster than governance, while Gartner expects over 40% of projects to die before 2028 unless enterprises tighten control right now.
Summary
Over 40% of today’s agentic AI projects will die before 2028, Gartner forecasts, and the models are not the killer. Costs, fuzzy value, integration pain and weak controls are. McKinsey’s 2026 AI Trust Maturity Survey scores average responsible-AI maturity at 2.3 out of 4, with only about 30% reaching level three or higher. Nearly two-thirds call security and risk their biggest scaling challenge.
Agent deployments are growing roughly 8x faster than governance maturity. Of thousands of “agentic AI” products, Gartner says only around 130 are truly autonomous. Winners use narrow agents, human checkpoints, full decision trails and controlled data. The EU AI Act’s human oversight deadline is December 2027. After 2024 and 2025’s autonomy race, 2026 to 2027 is about trust.
Positives
- Scoped agents make failures smaller, easier to audit and safer to contain.
- Human checkpoints can stop high-stakes mistakes before data moves or transactions post.
- Controlled environments limit the blast radius and simplify audit trails.
Risks & concerns
- Gartner expects more than 40% of current agentic AI projects to miss 2028.
- Governance maturity trails agent deployment growth by roughly 8x.
- Too many human approvals can turn AI agents into expensive manual automation.