A new MIT research report highlights a stark difference in success rates between companies that purchase AI tools and those that build them internally, revealing high failure rates for proprietary experiments.
📉 The 95% Failure Rate
The report suggests that nearly 95% of initial generative AI initiatives fail to deliver value. However, the path to implementation matters significantly:
- Buy vs. Build: Companies that purchase AI tools from specialized vendors succeed about 67% of the time.
- Internal Struggles: Internal builds succeed only one-third as often as external solutions.
This finding is particularly critical for highly regulated sectors like financial services, where many firms are attempting to build proprietary solutions in 2025 but facing challenges.
🔑 Keys to Success
Besides choosing the right vendors, the report identifies other success factors:
- Empowering Line Managers: Adoption should be driven by managers closer to the work, not just central AI labs.
- Deep Integration: Tools must integrate deeply into existing workflows to be effective.
🚧 Workforce Disruption
Disruption is already underway, particularly in customer support and administrative roles.
- Silent Reduction: Rather than mass layoffs, companies are increasingly choosing not to backfill vacant positions.
- Outsourced Jobs: Changes are concentrated in jobs previously outsourced due to perceived low value.
👥 Shadow AI & The Future
There is widespread use of "shadow AI"—unsanctioned tools like ChatGPT used by employees without efficient oversight. Looking ahead, advanced organizations are experimenting with Agentic AI systems that can learn, remember, and act independently.