AI ROI: Why 85% of Users Create No Business Value

AI ROI Doesn't Come From Access, It Comes From Real Use Cases

Your team has access to the best AI tools on the market. And they're using them to rewrite emails.

That's not a guess. The Section AI Proficiency Report analysed 4,500 AI use cases inside companies. The result: only 15% are likely to generate ROI. 85% don't. Not because the tools are bad. Because nobody asked beforehand what they were actually supposed to be used for.

AI ROI isn't a technology problem. It's a structural problem.

What the Numbers Actually Show

A few numbers from the report that are worth knowing:

• 59% of use cases are simple task help, disconnected from any real workflow
• Only 2% of reported use cases involve real automation
• 26% of employees have no work-related AI use case at all
• 54% of engineers don't use AI for coding
• 87% of product managers don't use AI for prototyping
• 56% of marketers don't use AI for content drafts

That's the unsettling part: even in the roles where AI use should be most obvious, people skip the most obvious applications. That kills the argument that "our tech people will figure it out themselves."

They won't. Not without structure, anyway.

The Gap Between Management and Staff

The C-suite believes it's encouraging its teams to experiment with AI. Only 10% of individual contributors agree.

That's not a perception gap. It's a fundamental misunderstanding of what "encouragement" means. An email titled "We're all using AI now!" isn't direction. A ChatGPT licence without context isn't enablement.

I hear the same lines in companies again and again:

"We've rolled out AI for everyone." Meaning: we bought licences. Nobody knows what people actually do with them.

"Our people are trained." Meaning: they sat through a two-hour webinar. Nobody checked whether anything changed afterwards.

"We encourage experimentation." Meaning: there's a Slack message. It doesn't say what to experiment with.

Access gets mistaken for strategy. Adoption gets mistaken for value creation. And when nobody's using AI for anything useful after six months, the blame lands on the staff.

Why AI Use Cases Don't Emerge Without Structure

95% of enterprise GenAI projects showed no measurable financial return within six months. That sounds like bad tools. It's the same story again: no clear use case, no direction, no accountability.

There's a simple reason for this. Developing AI use cases is a skill, and skills have to be built. It doesn't happen just by handing out access and hoping for the best. The question "which of my daily tasks could AI actually help with?" isn't trivial. It takes knowledge of what AI can do, knowledge of your own day-to-day work, and the space to bring the two together.

Most employees are missing at least one of these three. The space is almost always the missing piece.

Buying everyone a gym membership and hoping the team writes its own training plan isn't a workplace fitness programme. It's wasted budget.

What Actually Works

The examples in the report that show how it's done all share one thing: structure and accountability.

Klaviyo, 2,400 employees: A CEO memo, a dedicated Enablement Month, and every employee had to build and demonstrate an AI use case within three months. 1,800 did. One junior finance hire taught herself HTML and CSS to build her own tool.

Horizon Media: From "AI-nascent" to 85% AI-certified in under a year. The CEO posted his own certification publicly. Legal runs watch parties for AI content.

What connects these examples: none of it was voluntary or self-directed. It was structured, measured, and modelled by leadership.

AI ROI Comes From Accountability, Not Enthusiasm

"When something becomes a measured responsibility, it gets done. When it's left to organic curiosity, it gets skipped." That's the core message of the Section report, and it lands.

As long as use-case development is a voluntary activity squeezed in around the day job, it barely happens. As long as leaders ask "who's already using AI?" instead of "who demonstrated their use case this week?", AI stays one option among many.

In practice, that means:

1. Define role-specific use cases, not generic lists. Sales needs different use cases than accounting.
2. Make use-case development a manager's responsibility, not an HR initiative.
3. Make good use cases visible. Monthly spotlights, shared channels, quarterly reviews.

The AI Design Sprint is a focused way in: teams work together on real problems from their everyday work. At the end, there's a working AI prototype and a team that understands what AI can actually do for their specific tasks. That's a use case with a real shot at being one of the 15%.

If you want to know more about how real AI competence builds inside a team, there's more in my article on how AI competence comes from doing, not knowing.

Your Next Step

If you want to know how your team can develop concrete, role-specific AI use cases and try them out in five days, book a free Discovery Call.

Section AI Proficiency Report 2026: 4,500 AI use cases analysed, 15% with a likely ROI
MIT 2025/2026: 95% of enterprise GenAI projects with no measurable return within 6 months
Klaviyo / Horizon Media: case studies from the Section AI Proficiency Report 2026

Back to overview