Viewsยท

Organisations struggle with using new tools correctly

One in four UK workers spends more than seven hours a week moving information between AI tools and internal systems. Why that is a leadership and architecture problem.

I have been writing code, advising clients and watching enterprise software for twenty-five years. The pattern is one I have seen before. New tools land in an organisation faster than the organisation can absorb them. They do not talk to each other. The work of moving information between them, the copying, pasting, reconciling, summarising the summary, rewriting the rewrite, all of it falls onto humans. We become the connective tissue that makes the tools work.

What the data says

Workday's The Copy-Paste Economy AI Research Report, published this month, found that one in four UK workers spends more than seven hours every week moving information between disparate AI tools and internal systems.

  • Nearly 80% of UK workers are held up by admin tasks every week.
  • 75% say decisions are delayed because the information they need is missing or unclear.
  • More than 60% of UK workers have days that are "busy but unproductive" often or very often. The global figure is 43%.
  • 77% report higher stress as a direct consequence.

Three of those four numbers describe psychosocial conditions: demand, ambiguity, lack of progress. The fourth is the outcome those conditions produce.

What it actually feels like

It looks like this. You start a task. You open a tool. The tool needs context. You copy the context from the system that holds it. You paste it into the AI. The AI returns something almost right. You fix the parts that are wrong. You paste the corrected version back into the second system. You then write a Slack message explaining what you have done, because the system the AI was meant to update did not update.

Multiply that across a forty-five minute period, six times a day, two hundred days a year, and you have what an anonymous IT director told the Workday researchers. His days, he said, were busy, but did not feel like real progress.

That is exactly what can be described as low-autonomy, high-demand work. The framework for it is the Job Demand-Control model, first published by Robert Karasek in 1979 and used in occupational health research ever since to describe the conditions that produce chronic stress.

Why this is the predictable failure of bad AI strategy

The mid-2010s SaaS boom did this to mid-sized companies. Salesforce, Workday, Jira, Slack, Notion, three project tools per team, ten dashboards no-one read. The integrations promised by every vendor were partially real. The gaps were filled by humans.

AI is the same pattern just faster. Every team has bolted on an AI tool. Most of them do not connect to the systems where the actual work lives. Workday's own data says only 23% of UK organisations have AI deeply embedded in their core systems. In the other 77%, AI sits in pockets, and the human becomes the integration layer.

For the average UK worker, the result is more admin done at a faster pace, packaged as productivity. Workday's UK lead calls it the human middleware problem.

Why it matters for wellbeing

The simplest way to predict whether a workplace will burn its people out is to look at the relationship between demand and control.

High demand with high control is hard work, but it is sustainable. High demand with low control is the condition that produces chronic stress, exhaustion and turnover.

The human middleware condition is a textbook low-control, high-demand environment. Workers are required to keep up with a tool stack they did not choose, in an architecture they did not design, with poor communication between the two. They do not control the tools. They do not control the integration. They control whether they copy and paste fast enough. All in the age of AI where speed has become a substitute for quality of work.

It isn't surprising then that 77% of the people doing this are stressed.

This is not something a mindfulness app or a yoga session is going to fix. The modern workplace needs better data as to the why, not the what of stress and burnout.

Who is responsible

This is a leadership and architecture problem.

The companies in Workday's data with AI deeply embedded in core systems show 57% of employees reporting that AI has cut their task time by 25% or more. The companies without that integration are the ones whose people are middleware. The difference between those two groups is a leadership choice about how AI gets adopted: as a deliberate piece of system architecture, or a bolt on so someone can simply say 'we have AI'.

The real cost of getting it wrong won't appear on a Finance dashboard. It appears as stress and staff turnover.

Closing

Those that the article described as 'Human Middleware' would be better described as effectively 'Human Guinea Pigs' for the new AI boom. Whilst their productivity is being measured, their wellbeing is not.

Sources

  1. ITPro: "Too many employees are serving as the human middleware: workers are wasting a full work day each week switching between disparate AI tools and internal systems", Emma Woollacott, 15 May 2026
  2. Workday: The Copy-Paste Economy AI Research Report, May 2026
  3. Karasek, R. A. (1979) "Job demands, job decision latitude, and mental strain: Implications for job redesign." Administrative Science Quarterly, Vol. 24, No. 2, pp. 285-308. DOI: 10.2307/2392498

First published on LinkedIn in May 2026 as issue 2 of The Measured Workplace, the Occlio newsletter.

More in ViewsAll articles