AI Forecast Tracker
← Back to forecasts

Companies Are Betting Billions on AI Agents. Most Will Fail.

Will more than 60% of enterprise AI agent projects be abandoned or scaled back by end of 2027?

If your company is betting on AI agents, there's a 60% chance the project gets killed or downsized — but the 40% that succeed will be transformative.

Target: Dec 2027(664 days until resolution)
Assessed Probability
55%
More likely than not
Based on 4 expert predictions, 4 evidence items
Community Forecast
Cast your vote
Be the first to weigh in below

Your Prediction

Where do you think this lands?

Join others who've weighed in

5%95%
50% — More likely than not
MIT found that 95% of enterprise AI pilots deliver zero measurable ROI. S&P Global reports 42% of companies already scrapped most AI initiatives in 2025. But the narrative is shifting faster than expected. Enterprise agentic adoption hit 57% — blowing past Gartner's 40% prediction for end 2026. Salesforce Agentforce reached $800M ARR with 18,500 customers averaging 12 agents per organization. Claude agent teams are in production. The acceleration in framework quality since late 2025 is real. Most enterprise AI projects will still fail — that's the base rate for any enterprise technology (60-80%). But the success rate is improving faster than prior technology waves, and the successful deployments are delivering outsized returns. The AI-native startups (Harvey, Distyl, Basis) prove that the technology works — the enterprise adoption gap is about governance and integration, not capability.

Scenarios

Current value: 88-95% AI pilot failure rate (various surveys, 2025); but 57% enterprise agentic adoption (exceeding Gartner's 40% prediction)

S-curve position: Entering slope of enlightenment — early adopters proving value, late majority still struggling

Bear Case

75% failure (complexity overwhelming, runaway costs, but lower than before due to framework maturation)

Base Case

55-65% failure (improving from historical base rate, concentrated in companies without AI expertise)

Bull Case

40% failure (frameworks mature fast, governance commoditized, AI-native companies set the template)

How We'll Know

What we measure
Percentage of enterprise AI agent deployments that are abandoned, significantly scaled back, or fail to meet stated objectives per analyst surveys
Confirmed if
Analyst firms report >60% failure/abandonment rate for enterprise agentic AI projects by end 2027
Refuted if
Failure rate stays below 40%, OR successful agent deployments clearly outnumber failures
Data sources
  • Gartner Hype Cycle reports
  • McKinsey AI adoption surveys
  • S&P Global enterprise AI tracker
  • Industry earnings calls mentioning agent project outcomes

Evidence Trail

Evidence For

  • Mar 7, 2026

    MIT: 95% of AI pilots deliver zero ROI. S&P Global: 42% of companies scrapped most AI initiatives (up from 17%). Gartner: >40% of agentic AI projects expected to fail. 90% of orgs lack internal capability to scale agents. Historical enterprise AI project failure rate 60-80%.→ Probability: 70%

  • Mar 7, 2026

    Enterprise agentic adoption at 57% (exceeded Gartner's 40% end-2026 target). Salesforce Agentforce $800M ARR, 18,500 customers, avg 12 agents per org. Claude agent teams in production. AI-native startups (Harvey $195M ARR, Distyl $1.8B, Basis $1.15B) proving the technology works when built from scratch. Framework quality improving rapidly.→ Probability: 60%

  • Mar 9, 2026

    Goldman Sachs developing autonomous agents for trade accounting and onboarding. Lloyds Banking Group expects enterprise-wide agentic AI in 2026 projecting £100M added value. JPMorgan saving $1.5B through AI. Gartner projects 40% of enterprise apps embed agents by end 2026 (up from <5% in 2025). Systematic evaluation tools increase agent production success rate by 600%.→ Probability: 55%

Evidence Against

  • Mar 7, 2026

    Salesforce Agentforce $800M ARR with 18,500 customers suggests real value delivery. 57% of enterprises running multi-step agent workflows. Agent frameworks improving rapidly (LangChain, CrewAI, Claude agent teams). Enterprise AI spend still growing.

How Our View Evolved

  • Mar 9, 202660%55%

    Financial services showing real agent value: Goldman/Lloyds/JPMorgan deploying agents with measurable ROI. Gartner projects 40% of enterprise apps embed agents by end 2026. High-end success stories temper the failure thesis.

  • Mar 8, 2026Initial assessment: 60%

    Baseline — initial published assessment

What Experts Say

Andrej Karpathy

AI Researcher, former Tesla AI Director, educator

Track record: 8/10
It will take 10 years to build all the agents that can do meaningful work
Jan 2026 | interview
We assess this claim as 70% likely

Gartner

Technology Research and Advisory

Track record: 7/10
40% of enterprise applications will feature task-specific AI agents by end of 2026, up from <5% in 2025
Aug 2025 | report
We assess this claim as 70% likely

Gartner

Technology Research and Advisory

Track record: 7/10
Over 40% of agentic AI projects will fail by 2027 due to unintended decisions, runaway costs, and governance challenges
Oct 2025 | report
We assess this claim as 75% likely

Gartner

Technology Research and Advisory

Track record: 7/10
Agentic AI could drive ~30% of enterprise application software revenue (~$450B) by 2035
Aug 2025 | report
We assess this claim as 30% unlikely

What Could Go Wrong

Agent tooling matures faster than expected. A killer framework emerges that makes agent governance easy. 'Failure' definition is too broad — many scaled-back projects still deliver partial value. The 60% threshold is actually the historical base rate, making this almost a tautology.

What should we track about this topic?