New report reveals business value is not keeping pace with widespread AI adoption
Most large organisations remain stuck in a "performative AI" stage, rolling out tools without embedding them into meaningful work, according to a new report from AI training and enablement firm Section.
The company's AI Proficiency Report, based on data from more than 5,000 knowledge workers in the US, UK, and Canada, finds that while AI access is spreading across large companies, business value is not keeping pace.
Section reports that 55% of knowledge workers now use AI at work at least once a week, yet 85% do not have an AI use case that delivers measurable business value.
More than a quarter of employees (26%) do not use AI for work at all, and fewer than three per cent qualify as AI "practitioners" or "experts" capable of embedding the technology into workflows to achieve substantive productivity gains.
"Too many AI initiatives still focus on access and basic prompting," said Section CEO Greg Shove in a statement. "But the new standard requires employees to use AI in ways that actually change how work gets done - and almost no one is there."
Still in 'performative AI'
The research assesses AI knowledge, usage, prompting skill, attitudes, and organisational readiness, including hands-on evaluations of how employees construct prompts and the quality of their work-related use cases.
It concludes that the gap between AI hype and results is being driven less by tools or training budgets and more by shallow, low-impact applications that rarely touch core processes.
"Most organisations are still doing 'performative AI' - roll out some tools, run a lunch and learn, and call it transformation," Shove said.
"But they haven't done the actual work of rethinking how work is done with AI. The result is exactly what we're seeing: everyone has access, almost no one is generating value."
The report comes as research forecasts that 2026 will herald a more advanced era for AI at work, where organisations move past the adoption stage.
However, Section's report indicates that most employees struggle even at the earlier stage.
According to the findings, 70% of workers are classified as "AI experimenters" who write weak prompts and rely on AI for low-value, one-off tasks rather than re-engineering how they execute their jobs.
The data suggests that expert-level practice remains rare inside large enterprises.
Less than 0.1% of workers have reached a stage where they are automating substantial portions of their work with AI and achieving significant time savings or impact, the report finds.
On average, knowledge workers score just 2.3 out of 10 on prompting ability.
The productivity effects are modest as a result. Nearly a quarter of employees, 24%, say they save no time at all with AI.
Another 44% report saving fewer than four hours per week, which Section characterises as well below what most organisations would require to see a clear return on investment from AI initiatives.
Even companies that are investing in enablement are not necessarily seeing stronger outcomes.
Among workers who receive AI training, average proficiency scores reach only 40 out of 100, indicating that access to learning resources alone is not closing the gap between experimentation and meaningful value creation, according to the report.
The problem with use cases
While low skills are part of the story, the report identifies use case quality as the more fundamental bottleneck.
Based on an analysis of more than 4,500 work-related AI use cases submitted by respondents, Section finds that 85% of knowledge workers have only beginner-level AI use cases or none at all. Over 25% of reported uses have no clear connection to broader processes or workflows inside the organisation.
Only 15% of use cases are judged likely to generate any real business value. The most frequent way employees say they use AI is as a replacement for search, rather than as a mechanism to redesign tasks or automate complex work.
Just two per cent of reported use cases are scored as advanced, underscoring how few teams have moved beyond surface-level experimentation into integrated, high-impact applications.
What can employers do?
The report warned that the proficiency gap in workplaces will grow wider the longer employers wait.
To address the problem, it urged employers to stop measuring AI success by access and adoption rates.
"Start tracking time saved per employee, use case quality, and business outcomes instead," the report read.
It added that leaders should start identifying and tracking at least three AI use cases for each direct report.
"Implement regular skip-levels focused specifically on AI adoption barriers, and require executives to shadow employees as they use AI in their daily work," the report stated.
It further highlighted that organisations need to build use case libraries, stressing that employees are stuck with low-value AI adoption because they don't know what problems AI can solve in their role.
"Build function-specific use case libraries, create role-based playbooks, and assign use case development as a measured responsibility for team leads," it suggested.
