Call our Glasgow IT support team: 0141 255 1617

Are Your Employees Becoming AI “Middleware”?

AI was meant to take the repetitive work off your team’s plate. In plenty of offices it has done something else: turned people into the go-between for software that can’t talk to itself.

Quick answer

“Human middleware” is what happens when your staff manually carry information between systems that aren’t connected. AI tends to make it more obvious, because people end up pasting data into an AI tool, explaining context it should already have, checking the result and then copying it somewhere else.

If AI is adding handoffs rather than removing them, look at the workflow and the connections between your systems before buying another tool.

Watch: Is AI creating more work for your employees?

You’ve probably seen it. Someone copies notes from one platform into another. Someone else cross-checks two applications because neither can be trusted on its own. A third person rewrites the same prompt for the fifth time because the AI doesn’t know what the business already knows, then tidies up an output that was nearly right.

People have started calling this human middleware: employees acting as the glue between disconnected systems. Once you know what to look for, it’s hard to stop noticing it.

Workday’s 2026 UK research found that one in four employees surveyed spends seven or more hours a week moving information between applications, reconciling data and manually giving AI tools context.

Source: Workday UK research, 2026.

What human middleware looks like day to day

Picture an office manager who needs a customer summary. The CRM can’t feed the AI assistant directly, so she exports a report, pastes it into the assistant, adds a few paragraphs of background, and gets a decent draft back. She doesn’t quite trust it, so she checks the figures against a spreadsheet. Then she pastes the final version into the system where the work actually lives.

No single step is slow. But multiply that by every task, every day, across a whole team, and you’re looking at a lot of lost hours.

Common examples include:

  • Copying customer details between a CRM, email and another business application.
  • Pasting information into an AI assistant because it can’t reach the original system.
  • Retyping context into prompts that already exists elsewhere in the business.
  • Checking AI output against another system before anyone will rely on it.
  • Exporting a spreadsheet or report just to upload it somewhere else.
  • Pasting an AI-written response back into the application where the work happens.

AI can save time and still create admin

Both things can be true at once. AI really can draft an email in seconds, summarise a long document, pull out the key points from a meeting or get a report started far faster than a person could. But around each of those wins, new chores appear: finding the source material, loading it into the tool, explaining the background, reviewing the result and moving it on.

The task got quicker. The process around it may not have.

Why does this happen?

Mostly because AI tools have arrived faster than the systems underneath them have caught up. A business adds an assistant here, an automation there, and an AI feature bolted onto an app it already uses. Each one does something useful. None of them share information with the others or understand the wider process.

So people fill the gaps by hand. It’s worst where information is scattered across email, spreadsheets, Microsoft 365, cloud apps and older line-of-business software. The AI is capable enough, but the information it needs isn’t where it needs to be.

The AI usually isn’t the problem

If your information sits in several disconnected systems, someone still has to move it before any AI tool can use it. A new subscription won’t change that.

The place to start is the workflow itself.

Trace how information travels through your business. Where does it come in? Where does it need to end up? Which systems already hold it? And how often does someone copy, paste, retype, export or double-check it along the way? Those are the spots where your technology should be doing the work for you.

If you already use Microsoft 365, it is worth checking how Outlook, Teams, SharePoint and OneDrive share information before adding another standalone product. Our Microsoft 365 support is a good place to start if that sounds like you.

Ask these questions before buying another tool

New AI products appear almost weekly, and it’s tempting to solve every headache by adding one. But each extra application also means another login, another store of information and another place data has to be moved to and from. Before you sign up, ask:

  • What exact problem is this supposed to solve?
  • Where will it get its information from?
  • Can it reach that information securely, without someone copying it across?
  • Where does the output need to go afterwards?
  • Will people still have to check it or retype it?
  • Do we already own something that does much the same job?

The answers will tell you whether the tool simplifies a process or just moves the work somewhere else.

Signs of hidden AI admin

Human middleware rarely shows up on an invoice or a report. You spot it in how people work: constant switching between applications, personal spreadsheets kept because the main systems don’t hold everything, and the same information copied again and again between email, cloud apps and AI assistants.

You’ll also hear it in the office:

  • “I just copy it across manually.”
  • “The AI does most of it, but I have to fix a few things.”
  • “That system doesn’t connect to this one.”
  • “I export it to Excel first.”
  • “I have to give Copilot the information again every time.”

Any one of these may seem insignificant on its own. If you’re hearing them regularly, across different people and teams, the combined cost is probably bigger than you think.

Good AI should cut handoffs, not add them

Your people shouldn’t spend their day helping one piece of software talk to another. They should be solving problems, looking after customers and making decisions.

That doesn’t mean automating everything. Human judgement still matters, especially when AI output needs checking or the decision carries real consequences. But there’s a big difference between someone reviewing work because their judgement adds value, and someone copying data between two applications because nobody ever connected them.

Keep people involved where they make a difference, and remove the handoffs where they don’t.

Start with what you already have

You rarely need to rip everything out. Often it comes down to understanding the systems you already use and finding where manual steps have crept in. That might mean tidying up your Microsoft 365 setup, connecting the cloud applications you already pay for, rethinking where files and information are stored, or automating one repetitive workflow.

A wider managed IT services relationship can help here too, because someone is looking at how everything fits together rather than one application at a time. And if your team relies on lots of online platforms, our cloud services can help you review how they connect.

Is AI reducing your workload, or just moving it?

Measuring the time the AI saves only gives you half the picture. Look at what your people have to do before and after it runs. If they’re constantly switching apps, correcting outputs and stitching workflows together by hand, the AI probably isn’t at fault. It’s more likely a sign that the technology underneath needs attention.

Could your technology be working harder for your team?

If your staff are losing time moving information between systems, working around applications that don’t connect or grinding through repetitive tasks, Acu IT Solutions can review how your technology is being used and show you where to simplify things.

Talk to Your Local IT Experts