ENTERPRISE AI · 9 MIN READ

Enterprise AI Investment Is Entering a Value-Discipline Phase.

Initial enthusiasm is giving way to harder questions about measurable returns, portfolio choices and whether organisations can separate strategically important investment from expensive experimentation.

INTAI Research · Published 2 June 2026

Read the Analysis

The Spending Curve and the Value Curve Have Diverged

Worldwide AI spending is forecast to total $2.52 trillion in 2026, a 44% increase year over year, a rate of growth that would be unremarkable if the value being generated were scaling at anything like the same pace. It is not.

MIT Media Lab’s widely cited “GenAI Divide” research, based on a review of more than 300 publicly disclosed enterprise AI initiatives, 52 organisational interviews and 153 executive surveys, found that despite an estimated $30–40 billion in enterprise generative AI investment, 95% of corporate AI initiatives were showing zero measurable return. Only around 5% of pilots had reached production with demonstrable value attached. The gap between capital committed and value realised is, at this point, the defining fact of enterprise AI investment, not a temporary lag that will close on its own as the technology matures.

Why the Gap Persists: A Leadership Alignment Problem, Not a Technology One

MIT’s research is explicit that the divide is not explained by model quality or by regulatory friction. It is explained by approach, and a specific approach failure shows up consistently across separate research on enterprise AI leadership.

Grant Thornton’s 2026 technology survey found CIOs and CTOs rank strategic alignment as the dominant driver of their AI investment only 34% of the time, while CFOs and COOs, funding the investment and living with its operational consequences, rank it at 61% and 58% respectively. That disconnect helps explain why so much enterprise AI spending produces pilots rather than production systems: the people closest to the purchasing decision are frequently the least focused on the business case that would make the investment durable.

Central Insight

A value-discipline phase does not mean less AI investment. It means a harder filter applied before capital is committed: does this initiative address a named business constraint with a measurable metric attached, or is it an exploration that should be funded and evaluated as exploration, with a clear decision point for whether it graduates to production or is retired. Treating every pilot as though it is already a strategic investment is precisely the pattern behind the 95% figure.

What Portfolio Discipline Looks Like in Practice

The organisations in MIT’s 5% share specific habits: they design for friction rather than avoiding it, embedding AI into genuinely high-value workflows with deep system integration and memory or learning loops, rather than deploying generic tools that look impressive in a demo and prove brittle against a real process’s exceptions.

Applied at the portfolio level, that translates into treating AI investment less like a single technology budget line and more like a genuine investment portfolio, with a small number of scaled, accountable, measured deployments generating the bulk of realised value, and a clearly bounded exploration budget for the much larger number of pilots that may or may not graduate, evaluated honestly against the same value-discipline standard rather than allowed to persist indefinitely on the strength of early enthusiasm alone.