The Honest Math on Enterprise AI ROI

Scott Galloway and Goldman Sachs' Eric Sheridan just had one of the more candid conversations about the AI market I've heard this year, at Section's AI ROI event.

Their core finding: consumer AI has hit escape velocity. Enterprise AI still has a ROI problem.

The Numbers Behind the Discomfort

An MIT study of 300 AI deployments found that 95% of pilots delivered zero measurable P&L impact, despite $35 to 40 billion in enterprise investment. The models work fine. Poor integration and misaligned priorities cause the failures.

Galloway's math is uncomfortable. Justifying current valuations requires roughly $1T in new revenue or efficiencies. He sees two paths there: massive job destruction or a major correction.

Where the Value Actually Lands

Here's the part I keep thinking about. Sheridan said the outsized ROI has never been made on the infrastructure layer. It shows up at the application layer. Galloway added that the LLMs themselves are converging to parity.

The data backs this up. Companies that buy and partner succeed 67% of the time. Internal builds succeed a third as often. Winners pick one pain point and execute well.

What This Means for Your Firm

This is the thesis we're building Experio on: knowledge infrastructure that turns your firm's contracts, proposals, and institutional expertise into intelligence that shows up in the P&L.

The coming shakeout clears out AI without ROI. AI that connects to real outcomes survives it.

Worth the watch: https://vimeo.com/1229233776