AI readiness is the missing layer in enterprise AI ROI.

If you want to understand why enterprise AI has disappointed so many companies, don’t start with the technology. Start with the number that has quietly become the most important statistic in enterprise software.

The number every CFO is now afraid of

In its widely cited study The GenAI Divide: State of AI in Business, MIT’s Project NANDA found that 95% of organisations achieved zero measurable return on their AI investment — despite tens of billions of dollars in enterprise spending. Later analyses through 2026 confirmed the plateau rather than breaking it. The same wall shows up across independent studies:

Source	What they found
MIT Project NANDA	95% of organisations saw zero measurable return on AI investment.
S&P Global	A large share of companies abandoned most of their AI projects.
Morgan Stanley	Only about a fifth of large companies could cite a measurable AI benefit at all.

Read that again. Not 95% saw small returns. Ninety-five percent saw zero measurable return. Meanwhile spending keeps climbing — average enterprise AI budgets are projected to rise sharply year over year, even as most firms cannot prove a return on what they’ve already spent.

This is the central tension of the current AI cycle: the gap between AI spending and AI proof has never been wider.

The comforting explanation is wrong

The comfortable story is that the models aren’t good enough yet. The evidence says otherwise. MIT’s own researchers were explicit that the failure wasn’t the technology — it was implementation. The two most common mistakes they identified:

1. Building instead of buying — companies constructing bespoke AI systems instead of deploying proven ones.
2. Deploying AI where it doesn’t move the needle — pouring effort into flashy front-office use cases while ignoring where returns are actually higher.

But underneath both of those sits a deeper, more human failure that most analyses skate past: the last mile between the tool and the person was never built.

The last-mile gap

Domino Data Lab’s 2026 research named it precisely: a last-mile gap between AI models in production and the business users who are supposed to unlock their value. As their COO put it, getting a model into production used to be the milestone that mattered — and it no longer is. The real milestone is the moment a business user can actually act on what the AI enables. For too many enterprises, that moment simply isn’t happening at the pace or scale of business.

This is the missing layer. A company can buy the best AI tools in the world, deploy them flawlessly, and still see zero return — because the employees who were supposed to use them never became ready to. And crucially, nobody is measuring that readiness.

What companies measure today	What it actually tells you
Licences purchased	That you spent money. Nothing about whether anyone uses the tool.
Training courses completed	That someone clicked through. Course completion is not capability.
AI readiness, per person	Whether each employee actually changed how they work — the thing the whole investment depends on. Almost nobody measures this.

What the missing layer actually is

The missing layer is workforce AI readiness — measured, specific, and per-person. It answers the question the whole AI investment depends on and that almost no company can currently answer:

For each employee, how ready are they to actually use the AI we’ve deployed — and what will it take to close the gap?

McKinsey’s 2026 research pointed in exactly this direction: organisations seeing significant AI returns were roughly twice as likely to have redesigned how work actually happens before selecting models. The transformation of the people and the work comes first. The technology follows. Companies that invert that order — buy the tool, hope for adoption — join the 95%.

Why this is where the ROI hides

Think about what the missing layer unlocks. If you can measure each employee’s readiness, you can:

Capability	What it changes
Target upskilling	Invest where it actually changes behaviour, instead of spraying generic training across everyone.
Prove ROI	A before-and-after readiness score — the measurement CFOs are demanding and can’t currently get.
Prioritise	Decide which tools and teams to invest in next based on data, not vendor enthusiasm.

This is precisely the layer Forsight was built to be. We measure every employee’s AI readiness against what their role actually requires, pinpoint their specific gaps, and generate a personalised 90-day plan to close them — tied directly to the AI tools the company already pays for. The re-assessment at 90 days produces the one thing the entire AI investment has been missing: proof.

Forsight is an AI workforce readiness platform from EpicureAI Labs. To see how it works, book a demo. Sources: MIT Project NANDA, “The GenAI Divide: State of AI in Business”; Domino Data Lab 2026 State of AI; S&P Global; Morgan Stanley; McKinsey (2026).