Your organisation bought the AI tools. Why aren't they delivering ROI?
88% of HR leaders say their organisations have not realised significant business value from AI tools. The problem isn't the tools. It's that nobody is measuring whether people can actually use them.
Your organisation deployed Copilot. Or ChatGPT. Or both. You ran training sessions. You announced the rollout. People said they were using it.
And yet, when you ask whether it's actually delivering value — the honest answer is: you're not sure.
You are not alone.
The gap is real and widespread
A Gartner survey of 2,986 employees conducted in July 2025 found that 88% of HR leaders say their organisations have not realised significant business value from AI tools. This is not a fringe finding.
EY's 2025 Work Reimagined Survey of 15,000 employees and 1,500 employers across 29 countries found that 88% of employees use AI at work — but only 5% use it in advanced or transformative ways. The gap between adoption and impact has a name: the AI productivity paradox.
Gartner Press Release, October 28, 2025. N=2,986.
EY Work Reimagined Survey 2025. N=15,000 employees, 1,500 employers, 29 countries.
Investment is rising but measurement is almost absent
McKinsey's 2025 Superagency in the Workplace report surveyed 3,613 employees and 238 C-suite executives. It found that over 80% of leaders say their organisations are not seeing tangible enterprise-level results from generative AI, and that less than one in five organisations track KPIs for their AI solutions at all.
BCG's 2025 Closing the AI Impact Gap report of 1,400+ C-suite executives found that only a third of companies have upskilled even one quarter of their workforce, and that most organisations do not track financial performance metrics for AI initiatives.
This is the core problem: organisations are measuring AI adoption — who has access, who completed the training module, who logged into the tool. They are not measuring AI capability — whether people can actually use AI effectively, evaluate its outputs critically, and make responsible decisions about when to use it.
McKinsey, "Superagency in the Workplace," January 2025. N=3,613 employees + 238 C-suite.
BCG, "From Potential to Profit: Closing the AI Impact Gap," 2025. N=1,400+ C-suite.
Behaviour data confirms the gap between claimed and actual usage
The most telling data comes not from surveys but from actual behaviour. ActivTrak's 2026 State of the Workplace Report analysed 443 million hours of digital workplace activity across 163,638 employees and 1,111 organisations. It found that while 80% of employees report using AI tools, 57% spend less than 1% of their total work hours in those tools. Only 3% fall in what ActivTrak identifies as the optimal productivity range of 7-10% of work time.
Access is not the same as adoption. Adoption is not the same as capability.
ActivTrak Productivity Lab, "2026 State of the Workplace Report," March 2026. N=163,638 employees, 443M hours, 1,111 organisations.
HR leaders know the skills gap exists but lack tools to measure it
A separate Gartner survey of 114 HR leaders published in October 2025 found that only 8% believe their managers have the skills to effectively use AI. Only 14% say their organisation provides support to managers on integrating AI into daily tasks.
The SHRM State of AI in HR 2026 Report, surveying 1,908 HR professionals, found that organisations average a maturity score of 3.85 out of 6.00 on AI readiness, and that about one third believe they are well behind other organisations.
The problem is not awareness — HR leaders know the gap exists. The problem is measurement. Without a way to assess actual AI capability rather than self-reported confidence, it is impossible to know where to invest, who needs development, and whether training is working.
Gartner Press Release, October 8, 2025. N=114 HR leaders.
SHRM, "The State of AI in HR 2026 Report." N=1,908 HR professionals.
The EU AI Act adds a compliance dimension
For organisations operating in or selling into Europe, there is now a legal dimension. EU AI Act Article 4, which entered into force on August 2, 2025, requires that providers and deployers of AI systems take measures to ensure a sufficient level of AI literacy of their staff. It applies to all organisations using AI tools — not just those deploying high-risk systems. The European Commission's own 2024 survey found that fewer than 25% of EU organisations using AI had a formal AI literacy programme in place.
A survey that asks employees how confident they feel does not satisfy this requirement. A performance-based assessment that measures actual capability does.
EU AI Act, Regulation (EU) 2024/1689, Article 4. Entered into force August 2, 2025.
What measurement actually looks like
Measuring AI capability means watching what people do with AI tools in realistic workplace scenarios — not asking how confident they feel. It means seeing whether a marketing manager's prompts produce usable outputs or whether they accept whatever the AI generates. Whether a finance analyst checks AI-produced calculations or publishes them unchecked. Whether a customer support agent knows when AI output needs human judgment before it reaches a client.
The gap between what people say they can do with AI and what they actually do is measurable. In our first cohorts at Probe Learning, it has consistently been the most clarifying data point for HR teams who have spent months wondering whether their AI investment was working.
If you want to know whether your organisation's AI tools are actually delivering value — start by measuring whether your people can use them. Book a demo at probelearning.com.