Yannick H.,
Too Long; Didn't Read
95% of GenAI pilots deliver no measurable ROI. 42% of companies will have abandoned their AI initiatives by 2025—a jump from 17% the previous year. Meanwhile, 69% of employees use AI tools without the IT department's knowledge. The result: chaos instead of business value. The inconvenient truth? Most are investing in the wrong use cases, measuring nothing, and then wonder why. The 5% that succeed do three things differently.
The uncomfortable truth about AI in the mid-market
Everyone is talking about AI. ChatGPT here, Copilot there, AI strategy at every conference.
The numbers tell a different story.
95% of enterprise GenAI projects deliver no measurable ROI.
That’s not 50%. Not 70%. 95 percent.
Companies worldwide have invested $30-40 billion in GenAI. The overwhelming majority of those investments? Burned, with no measurable impact on the P&L.
42% of companies in 2025 have canceled most of their AI initiatives. In 2024, the figure was still 17%.
(S&P Global Market Intelligence)
This is not a slow trend. This is a pullback.
The SME paradox: rapid adoption, little transformation
SMEs are among the fastest-growing adopters. Usage rose from 26% to 39% in just 12 months.
But here’s the problem: Only 8% have achieved true transformative integration.
The rest? Tools in use. Accounts created. Licenses paid. But no measurable business outcomes.
That’s the difference between "we use AI" and "AI creates value for us".
Shadow AI: The problem you can’t see
While you’re thinking about AI strategy, your employees have already decided.
69% of organizations suspect or have evidence that employees are using prohibited GenAI tools.
(Gartner Cybersecurity Survey 2025)
60% of employees use AI tools at work - but only 18.5% know the company policy on them.
The result is predictable: 46% of companies report internal data leaks caused by GenAI.
Samsung engineers entered chip design code into ChatGPT. Other companies found confidential customer data in public AI models. This is not theory - it happens every day.
What it costs: Shadow-AI breaches cost an average of $4.63 million - over $600K more than standard breaches.
(IBM Cost of Data Breach 2025)
Why 95% fail (and it’s not because of the technology)
The MIT study identified a pattern. The reasons are not technical - they are organizational.
Reason 1: No ROI framework
66% of companies struggle to establish AI ROI metrics at all.
Without defining success, there is no way to measure success. Without measurement, there is no proof the investment works. Without proof, there is no budget for the next phase.
The pilot dies quietly.
Reason 2: Wrong use cases
More than half of GenAI budgets go into sales and marketing tools.
The biggest proven ROI? Back-office automation. Eliminate BPO. Reduce agency costs. Streamline processes.
Companies invest where it sounds sexy. Not where it has measurable impact.
Reason 3: Build instead of buy
The desire for customized AI solutions is understandable. The success rates are not.
Vendor solutions: 67% success rate. In-house development: 33%.
With GenAI, buy clearly beats build. The technology evolves too quickly for custom development.
Reason 4: Change management ignored
Gartner 2025: Change management, not technology, is the most common reason GenAI programs fail.
Introducing tools is easy. Bringing people along is hard. Most skip the hard part.
The 5% that succeed
The MIT study also shows the other side: companies that use GenAI successfully achieve massive returns.
Top performers achieve $10.30 in return for every dollar invested.
(Wharton AI Adoption Report 2025)
What do they do differently?
1. They measure from the start. ROI framework before the first pilot. Baseline metrics documented. Success criteria defined.
2. They prioritize boring use cases. Not the demo-friendly AI chatbots. Instead, back-office processes that nobody sees but everyone pays for.
3. They buy instead of build. Specialized vendor solutions instead of custom development. The 67% vs. 33% success rate speaks for itself.
4. They have governance before tools. AI policy before the first account is created. Clear rules for what can go where. Shadow AI under control.
The three insights
95% failure is not a law of nature. It is the result of missing ROI metrics, wrong use cases, and governance chaos. The 5% prove it can be done differently.
Shadow AI is more expensive than no AI. $4.63M per breach, uncontrolled data flows, and 60% of employees who do not know what is allowed. Governance is not optional.
The ROI is not where you think. Not sales chatbots and marketing automation. Rather, back-office processes that nobody shows at conferences.
If you want to be among the 5%
We help SMEs approach GenAI the right way. Not tools-first. Strategy-first.
Use case prioritization: Focus on measurable results, not demo material
ROI framework: Metrics before the first franc is spent
AI governance: Shadow AI under control before it becomes a breach
Vendor evaluation: The right tools instead of expensive in-house development
Get in touch →
Related articles:
Sources:
