AI was supposed to remove repetitive work. In some businesses, something stranger is happening instead.

People are spending large parts of their day moving information between systems so the AI can function: copying notes from one platform into another, checking whether data matches across apps, rewriting prompts to give a tool more context, and manually correcting outputs that almost worked. But not quite.

There is a name for this: human middleware. Employees end up acting as the glue holding disconnected systems together.

Hi, I am Andy from Initial IT in Lichfield. We help small businesses across Staffordshire and the West Midlands get real value out of Microsoft 365 and AI, and once you know what human middleware looks like, you start seeing it everywhere.

What does human middleware look like?

Someone downloads information from one system because another cannot access it directly. A team member pastes customer details into an AI tool to generate a response, then copies the result somewhere else. Data gets checked by hand because nobody fully trusts what the systems produce automatically.

None of these steps looks like a problem on its own. Each takes a minute or two. Added up across a team and a week, they eat serious time.

Why busy days hide the problem

The strange part is that a business can feel more productive while this is happening. AI does help in plenty of places: emails get drafted quicker, reports take less time, information is easier to summarise.

But new layers of admin grow around it. You end up with days that feel productive on the surface while a growing share of the effort goes into coordination rather than progress. That is also exhausting for the people doing it, because translating between systems all day is nobody’s idea of meaningful work.

How did businesses end up here?

AI tools arrive faster than the systems underneath them evolve. A company adds one assistant here, an automation there, an AI-powered feature somewhere else, and none of it naturally connects. So people bridge the gaps by hand.

If systems do not integrate properly, if data quality is inconsistent, or if processes rely on manual handoffs, AI can layer extra complexity on top rather than removing any. The tool is not the problem. The plumbing is.

How to fix it: connect first, automate second

Instead of adding isolated tools everywhere, look at how information moves through the business. Where does data live? How do systems connect? Where are people still acting as the translator between technologies?

For most small businesses on Microsoft 365, much of the glue already exists and is already paid for: shared data in one place, Power Automate joining systems together, and Copilot working across documents and email it can actually see. That connected groundwork is the heart of our automation and CRM work and the reason our Copilot and AI rollouts start with the data, not the chatbot.

Your team should not spend their day helping software talk to other software. They should be solving problems, helping customers and doing the work that creates value.

Want a second pair of eyes on how information moves through your business?

We will map where your team is doing middleware work by hand, show you what should be connected, and give you a plain-English priority list. No jargon, no obligation.

Talk to Andy

Frequently asked questions

How do I know if we have a human middleware problem?

Ask each person in your team to list what they copy, paste or re-enter between systems in a normal week. If the list is long, or if anyone says “I just keep the spreadsheet in sync myself”, you have your answer. The people doing the gluing always know where it is.

Is the answer more AI tools?

Usually the opposite. Fewer, better-connected tools beat a drawer full of disconnected ones. Every new standalone tool adds another boundary someone has to carry data across. Connect what you have first, then add capability where a gap remains.

Where does Microsoft 365 fit into this?

For most small businesses it is the natural backbone. Your documents, email, calendars and Teams conversations already live there, Power Automate can join systems without code, and Copilot works far better when the data it needs is connected and tidy rather than scattered across apps it cannot see.

How long does fixing this take?

Mapping the flows takes days, not months, and it usually surfaces two or three quick wins straight away, often a single automation that removes a daily copy-paste job. The deeper integration work can then be phased around business as usual.

– Andy Price, Founder & MD, Initial IT · 01543 524 594 · hello@initialit.co.uk