This summer, the MIT Media Lab published some stark findings about the gap between AI adoption and transformation within enterprises.

According to the report, businesses indicate that just 5% of enterprise AI projects drive measurable productivity gains. In other words, 95% of projects are extracting no value for the business.

This is despite the enormous investment in AI initiatives, say the report authors, who cite a $30-$40 billion dollar enterprise investment in GenerativeAI.

[Source: The Gen AI Divide by MIT NANDA, July 2025]

So, what is going on here? And, is this the beginning of the AI bubble burst that so many are talking about?

Here are two points from my perspective:

1. Early AI adopters were far too quick to assume that AI can replace people

Since the mass adoption of GenAI in late 2022, I have seen marketing budgets slashed across the board in technology. Anecdotally, almost all of my BtoB technology clients have reduced the size of their marketing departments. Some of them have gotten rid of entire departments.

Forrester reports:

“We are seeing a noticeable decline in the number of CMOs in the Fortune 500. Only 58% of companies have a marketing executive reporting directly to the CEO or sitting in the C-suite, down from 63% last year. B2B companies are leading the retreat, with executive-level CMO presence dropping from 48% to just 42%.” [Source: Forrester, CMO Fortunes Falter Amid Economic And Role Uncertainty]

Why this doesn’t work: AI can make people more efficient, but it cannot replace professionals with years of career experience

• Nothing takes the place of real business experience. While it might be the case in the future, with true agentic AI, that some tasks will be tackled by AI without human intervention, at the moment humans are in the driving seat and should remain so. Even when AI can undertake tasks on our behalf, do we really want to be running a business on auto-pilot? At what stage does an automated marketing campaign look bad? At the stage when it starts to look robotic and un-humanlike. In other words: when AI takes over.

2. AI adoption has been driven by the employee, not the employer

Plenty of businesses have been dancing around the topic of AI adoption rules and regulations for employees since ChatGPT became widely adopted. Employees were tinkering and learning how to use the tool long before their bosses were producing guidelines for how AI can be used in their workplaces.

Why this hasn’t worked: Businesses haven’t been systemic enough in AI adoption

In addition to the disjointed way AI has been adopted in many offices, training programmes have failed in showing people how AI can truly help them. Why? Most training programmes are too generic, simply listing out AI tools that are available, showing briefly what they can do, without linking the tool back to how a specific person would actually use it in the business.

AI is a very personal tool — it should be viewed like you are training an intern, a very different approach to interacting with predictable, computational software. AI training programmes also require a level of intimacy with the person, their job function, and how the tool can be used in their role, helping with their day-to-day tasks.

How do you feel AI adoption has failed within enterprises?

Please comment below if you agree/disagree with my perspective on this bold report.