Case · 2025
Multiple Fortune 500
Enterprise GenAI pilot deployments across organizations
Maturity stage
Pilot/POC
Use-case type
Multiple
Function
Multiple
Company size
Enterprise
Evidence
95% of companies report GenAI pilots falling short
ROI / outcome figure
5% achieve rapid revenue acceleration
Deep dive
The setup
MIT NANDA's State of AI in Business survey (synthesised by Furr & Shipilov in HBR) sampled GenAI pilot programmes at Fortune 500 companies. The headline finding has become the most-cited statistic in enterprise AI - that 95% of pilots fall short of their stated objectives.
What happened
Across hundreds of pilots, the failure pattern was consistent. Pilots launched without crisp success criteria, ran on enthusiasm rather than instrumentation, and quietly stalled when leadership attention rotated to the next priority.
Root cause
Measurement, not model capability. Teams could not say what success looked like before they started, so they could not tell whether they had achieved it. This is the textbook IS-success failure mode - per DeLone & McLean, net benefits cannot be evaluated without information- and system-quality baselines.
Takeaway for teams considering similar work
Define the KPI before the pilot, not at the readout. RAPID's Impact dimension scores exactly this: do you have baselines, KPIs, and a review cadence in place before you commit budget?
What RAPID would have flagged
Failure mode: Measurement — Inability to track AI outcomes, unclear attribution, or missing baseline metrics that prevent learning and justification
Dimensions a pre-deployment RAPID assessment would have surfaced
- Measurement Maturity (low score < 50%)
Mitigations the framework recommends
- Define success metrics and baselines before deployment (DeLone & McLean IS Success Model)
- Build real-time measurement dashboards tracking AI-specific KPIs
- Isolate AI contribution through A/B testing or controlled rollouts
- Establish quarterly ROI review cadence with executive stakeholders
Dimensions this case illuminates
RData & Technical Readiness
Is the data infrastructure and technical capability adequate?
More from Multiple
Multiple Fortune 500 · CASE-003 · Failure
30% of GenAI projects abandoned after POC by end of 2025
UW academic study (Mistral AI, Salesforce, Contextual AI open-source LLMs) · CASE-012 · Failure
White-associated names preferred 85% vs Black-associated names 9%; demonstrates that off-the-shelf LLMs carry resume-screening bias
Multiple US Enterprises · CASE-015 · Success
23% of employed respondents used GenAI for work weekly; 9% daily
Apply this to your team
Take the RAPID assessment to see whether your organisation is exposed to the same failure modes as this case - or already has the discipline that made it work.