Leaders see AI working far better than their workforce does, and the divide nearly doubles inside large organisations
Executives believe artificial intelligence (AI) is working far better inside their organisations than the employees who use it every day do, and the AI sentiment gap nearly doubles inside large businesses, according to new research.
The AI Execution Index 2026 from Asana found executives scored their organisations 75.2 out of 100 for AI execution – how well AI is put to work across the business and turned into measurable results. Knowledge workers gave a score of 58.2, a gap of 17.1 points.
The divide was 11.6 points in organisations with 10 to 999 employees, widening to 22.5 points in those with 1,000 or more. The survey of 1,000 respondents – 300 executive decision-makers and 700 knowledge workers – was conducted through Dynata’s panel between 31 July and 12 August 2026.
The two groups also disagreed on what leaders can see. Sixty-one per cent of executives said leadership has full visibility into day-to-day AI use. Just 36 per cent of workers agreed.
“Leaders can see the AI investment and the strategy. Workers can see whether AI actually helps them get their jobs done,” said Jo Gaines, general manager, Asia-Pacific (APAC) at Asana. “That gap matters because making individual tasks faster doesn’t automatically make the organisation more effective. When leaders’ and workers’ perspectives of AI progress are this different, it’s a signal to find out what is happening on the ground before declaring success.”
Why leaders and workers measure AI success differently
James Tidey, AI go-to-market lead at Asana, told HRD the disconnect often comes down to how work is measured. When he asks individual contributors what their workflows achieve, many struggle to answer.
“Many of these individual contributors, if you ask them, hey, what quantifiable outcome does this workflow actually drive, they can’t give you an answer in many cases. So it shows a fundamental disconnect,” he said.
Executives and employees are typically measured against different goals, Tidey said, and those incentives shape behaviour. “If that’s not aligned with organisational outcomes at the top level, then I think it’s really clear to see where the disconnect does actually come from.”
He distinguishes between “blue money” outcomes – time saved, such as using a chat assistant to draft an email in seconds – and “green money” outcomes that grow revenue, cut costs or reduce risk. Many organisations focus on single-agent use cases that lift individual productivity, he said, rather than redesigning whole workflows with measurable results.
“What I’m saying is employees at all levels should be metriced in a way where they all meet at the same outcome. And that’s how you get everyone singing from the same hymnbook, so to speak,” Tidey said.
The data supports the point. Ninety-one per cent of respondents reported at least one measurable improvement from AI in the past 12 months, but the average was just 3.1 of 10 outcome areas. Executives were 19 percentage points more likely than workers to report revenue growth from AI.
Bigger businesses report more AI adoption, weaker execution
Half of enterprise respondents described AI as organisation-wide or fully embedded, compared with 42 per cent in small and mid-sized organisations. Yet enterprises scored 61.7 on the index against 64.9 for smaller organisations, and lower across all 10 dimensions measured. The widest gaps were in workflow integration (58.8 versus 63.5) and cross-functional coordination (61.3 versus 65.7).
“Organisations are throwing more and more tools at problems, and then the problem is they’re not actually able to bring all of that together and have people working from the same place meaningfully,” Tidey said. “Smaller organisations, people know what their workflows look like. They’re probably sat next to the person that they’re running that workflow with in the office. In these bigger organisations, that’s just not the case.”
Shadow AI – the use of tools an employer has not formally approved – is also widespread. Seventy-two per cent of executives and around half of workers said employees use unapproved AI tools. Training is a clear differentiator: 59 per cent of workers in organisations classified as “AI Executors” reported comprehensive training in practical and responsible AI use, compared with 29 per cent in “AI Adopter” organisations. The findings echo warnings that AI adoption without applied training is scaling mistakes rather than results.
Only 22 per cent of respondents qualified as AI Executors, defined as combining advanced adoption or readiness with measurable improvements in at least four outcome areas.
What should HR leaders do about AI anxiety?
Tidey said rapid change is wearing on employees who never chose to work with AI. “I’m an AI leader and I’m experiencing AI fatigue myself,” he said.
Fear of job loss needs to be met with empathy, he added, and leaders risk mistaking compliance for genuine uptake – a pattern explored in our analysis of why some employees fake AI adoption at work.
“In a market that lacks clarity, organisations should be trying to bring clarity to their workforce. I think that’s something that isn’t just a priority, but I think it’s a responsibility as well,” he said.
Tidey said employers should be open about their intent. “Why are we investing [in] AI? What are our goals with AI? What does that mean for you, the people, and how are you going to use it? I feel that that is something that all organisations should almost be adopting as a set of values: AI transparency.”
The report recommends asking leaders and employees the same questions on readiness, visibility, workflow fit, training, ownership and value, then examining where answers diverge – a step that grows more urgent as employee guidance lags behind companies scaling AI.
“To deliver measurable impact, leaders must ask workers the same questions they ask themselves about AI progress, then investigate where the answers differ,” Gaines said.
