95% of enterprise AI pilots produce zero P&L impact. The models aren't the problem.

MIT's NANDA initiative studied 300 AI deployments, interviewed 150 executives, and surveyed 350 employees.

The finding: 95% of enterprise generative AI pilots delivered no measurable impact on profit and loss.

Not "limited impact." Not "slow to ramp." Zero.

Against $30-40 billion in enterprise AI spending.

Here's what makes this devastating for the C-suite:

The models are fine. Frontier models clear 80% on SWE-bench, ship with 1M-token context windows, and pass every benchmark that matters.

The failure is organizational.

MIT found a "learning gap" — tools that don't learn from corrections, budgets misallocated to sales and marketing when back-office automation delivers the real ROI, and internal builds succeeding only33% of the time versus67% for external partnerships.

90% of employees already use personal AI tools for work. Only 40% of companies have official LLM subscriptions. Your workforce is running shadow AI right now. The question isn't whether they're using AI — it's whether you're capturing any value from it.

BCG's data confirms the pattern: only 26% of companies have moved beyond proof-of-concept. Gartner projects 30% of GenAI projects abandoned after PoC by end of 2025.

Stop running pilots the way 2024 ran them.

Audit your AI portfolio today. Kill any pilot without a pre-deployment baseline. Route your next AI investment through accounts payable, not sales enablement. The300 pilots MIT studied failed because nobody measured what mattered before the model touched production.

SOURCE: https://fortune.com/2025/08/18/mit-report-95-percent-generative-ai-pilots-at-companies-failing-cfo/
VERIFIED: MIT NANDA initiative (August 2025), Fortune, BCG "Where's the Value in AI?" report
SIGNAL: Enterprise AI spending is hitting $2.5 trillion in 2026 — and 95% of pilots generate nothing. The ROI reckoning is here.