Category: Technology | Published: 2026-06-09
The promise of artificial intelligence in the workplace has always been the same: fewer repetitive tasks, less administrative overhead, more time for the work that actually matters. It is a compelling pitch. It is also turning out to be more complicated than most organisations anticipated.
New research from multiple sources is telling a consistent story: AI tools are being adopted at pace, but the productivity gains many businesses expected have not arrived in the way or at the scale they hoped for. In some cases, AI appears to be generating as much work as it removes.
The Numbers That Should Give Businesses Pause
Start with the scale of adoption. McKinsey estimates that around 89 per cent of companies had deployed AI in at least one business function by the end of 2025. That is near-universal uptake among larger organisations. Yet the same research found that only 6 per cent of those businesses reported significant measurable value from their investment.
A separate study by the National Bureau of Economic Research surveyed around 6,000 chief executives and chief financial officers and found that 90 per cent said AI had delivered no meaningful productivity impact to their organisations. MIT's NANDA research group found that only 5 per cent of enterprise AI pilot programmes had a measurable effect on profit and loss.
These figures do not suggest AI does not work. They suggest that most organisations are not yet deploying it in a way that works.
The Copy and Paste Economy
One of the clearest explanations for this gap comes from Workday, which surveyed 2,400 UK professionals about how AI is affecting their day-to-day work. The findings are striking.
One in four employees reported spending more than seven hours each week simply moving information between disconnected systems. More than eight in ten said they spend significant portions of their working day coordinating between teams, transferring data between platforms, or reconciling conflicting information from different tools.
Workday describes this pattern as the copy and paste economy: a workplace where people spend a substantial share of their time acting as the connection between systems that should, in theory, be handling that themselves. AI tools have been added to organisations, but the underlying infrastructure those tools sit within has not changed to match.
The result is that employees are faster at certain tasks but busier overall, because the number of tasks requiring their attention has grown alongside the tools designed to help them.
Employees Want AI to Work. The Problem Is How It Is Being Deployed
One of the more counterintuitive findings in the Workday research is how positively employees feel about AI. Ninety-seven per cent of those surveyed rated their day-to-day work positively. Eighty-one per cent said AI had improved their experience at work. More than half reported that AI had reduced the time it takes to complete individual tasks.
This is not a story about workers resisting technology. People are broadly enthusiastic about AI tools and want them to succeed. The frustration comes when faster individual tasks do not translate into a less pressured working day, because the surrounding processes have not kept pace.
As one IT director quoted in the Workday report put it: dealing with system glitches, chasing approvals and constantly fixing or redoing work because of inconsistent data keeps people busy, but does not feel like real progress. That distinction between activity and progress is at the heart of the problem.
The Human Middleware Trap
Workday's researchers use a phrase that captures the situation well: many employees have become the glue holding disconnected systems together. Rather than AI and software handling the flow of information automatically, people are doing it manually, stepping in wherever systems cannot or do not communicate with each other.
A director in the construction industry described how a typical day feels genuinely busy, yet not productive, when most of it is spent on coordination tasks and system-related interruptions rather than the substantive work that actually moves things forward.
This is the hidden cost of piecemeal AI adoption. Businesses buy AI tools. Those tools sit alongside existing software. Nobody rearchitects the processes connecting everything together. Employees absorb the friction.
More AI Tools Can Actually Mean More Noise
Research by workforce analytics firm ActivTrak adds another layer to this picture. After tracking how AI adoption changed working patterns, the company found that email time increased by 104 per cent and time spent on chat and messaging platforms increased by 145 per cent following AI tool rollouts.
The explanation is not hard to find. AI tools generate outputs: summaries, drafts, suggestions, alerts. Those outputs need to be reviewed, discussed, approved and acted upon. Each one creates a small communication overhead. Multiply that across an organisation and the result is more back-and-forth, not less.
ManpowerGroup's 2026 Global Workforce Barometer found a related tension: worker use of AI increased by 13 per cent over the course of 2025, but confidence in the usefulness of those tools fell by 18 per cent over the same period. People are using AI more but trusting it less, which means more time spent checking and verifying its outputs.
This Has Happened Before
Economist Robert Solow observed in 1987 that the computer age was visible everywhere except in the productivity statistics. Companies had invested heavily in technology but the expected gains were not showing up. It took roughly a decade for those gains to materialise, as organisations finally learned to redesign their processes around the technology rather than simply adding computers to existing ways of working.
The AI productivity paradox of the mid-2020s looks similar. The tools are here. The adoption has happened. The redesign of processes to genuinely take advantage of them is still, in most organisations, catching up.
What Deep Integration Actually Looks Like
Workday's research identifies a meaningful divide between organisations that have added AI tools around the edges of their processes and the 23 per cent that have embedded AI directly into their core business systems and workflows. The latter group reports substantially better outcomes.
The difference is straightforward: when AI is embedded into the system where work actually happens, information flows without people manually moving it. When AI sits alongside existing systems as a separate tool, someone has to act as the bridge. That someone is usually an employee who already has enough to do.
The research also highlights that employees are effectively showing organisations what good looks like: they want AI integrated directly into their workflows, surfacing insights proactively rather than requiring them to go and find them.
What This Means for Your Business
The gap between AI adoption and AI value is real, but it is not permanent. The organisations closing that gap are the ones treating AI as a reason to rethink how work flows, rather than simply as a faster way to do the same things.
For smaller and mid-sized businesses, that does not necessarily mean a wholesale transformation. It can mean starting with a clear-eyed look at where the friction actually sits in your day-to-day operations and choosing AI tools that connect to the systems your team already uses, rather than adding another layer on top.
If you are trying to work out where AI could genuinely reduce workload rather than redistribute it, our AI Consultancy page covers practical options for businesses that want results rather than just adoption numbers.
The Takeaway
AI is not failing. But the way most organisations have implemented it is delivering a fraction of the value it could. The research is clear that the problem is rarely the technology itself. It is the gap between buying tools and redesigning work around them.
The businesses that figure that out first will not just be more productive. They will be considerably harder to compete with.