When AI Creates More Work: The Cost of Adding It to a Disconnected Process
A reply that used to take ten minutes to write now takes two.
That is a genuine gain, and it is easy to see. What is harder to see is what happens next.
Someone reads the draft to make sure it is accurate. Someone copies the details into the system where the record is kept. A colleague asks which version is the right one. A customer receives a reply that does not reflect what was agreed last week, so somebody has to correct it.
The writing got faster. The work did not necessarily get smaller.
This is the pattern we want to describe: what can happen when AI is added to a process that was already disconnected. It is not an argument against AI. It is an argument for noticing where the time goes.
Faster output is not the same as finished work
Most everyday AI tools are strong at a moment in the process. They draft, summarise, sort and suggest. That moment can be quicker.
But the work around it still has to happen. Information has to arrive in the right place, reach the right person, be checked, be stored, and trigger whatever comes next. If those steps depend on people copying, chasing and remembering, a faster draft simply arrives sooner at the same bottleneck.
It can even make the bottleneck heavier. More drafts, notes and summaries mean more things to review, file and act on.
We looked at the wider gap in AI Doesn’t Automatically Make a Business More Connected. This article is about the practical cost that gap can create.
Five ways AI can add work to a disconnected process
1. A new checking step appears, and nobody planned for it
AI can produce confident answers that are wrong. Sensible businesses check what it produces. In the government’s AI adoption research, 84% of AI-using businesses applied at least some human input or checking to AI outputs or decisions.
That is good practice. It is also work. If nobody has decided who checks, how long it should take, or what happens when it is wrong, the checking happens informally, in people’s spare minutes, and its cost never shows up anywhere.
2. The output has to be copied somewhere else
A draft in one tool, a customer record in another, a conversation in the inbox, a note in a spreadsheet. If the AI sits in one place and the record sits in another, someone becomes the connection. They copy, paste and retype.
Every copy is a chance to introduce an error, and every error is a chance to create more checking.
3. Everyone uses it differently
One person has a favourite prompt saved in a document. Another uses a personal account. A third has stopped using it because the answers did not suit the way the business talks to customers.
Quality and tone become inconsistent, and useful knowledge belongs to individuals rather than to the business. Someone then has to harmonise the results, or customers notice the difference.
4. More output, but nobody owns the next step
AI can produce notes, summaries and action lists very quickly. If there is no agreed place to keep them and no named person responsible for acting on them, speed produces volume rather than progress.
A meeting summary that nobody reads is not saving time. It is a new document to ignore.
5. The exceptions come back to a person, without the context
Routine requests are easy to handle. The unusual ones need judgement. If an AI assistant has handled the first part of a conversation but the person picking it up cannot see what was said, the customer repeats themselves, and the member of staff starts again.
The AI did what it was asked to do. The process around it did not connect.
What the UK research suggests
The Department for Science, Innovation and Technology’s AI Adoption Research gives a useful backdrop. Among UK businesses using AI, 75% reported improved workforce productivity. At the same time, 77% had not yet seen a change in revenue.
That does not mean AI is failing. It does suggest that time saved on individual tasks does not automatically become a business result.
The same research found that 23% of businesses already using AI said data complexity was hindering wider adoption, and 26% pointed to the difficulty of integrating and scaling AI projects.
The research does not measure rework, so we should not claim it proves the pattern in this article. Our reading is that these barriers describe the same underlying issue: information that is hard to reach, and tools that do not join up with the rest of the business.
|
A note on evidence The figures above describe UK businesses using AI in general. They do not show that AI creates more work, and they are not a Davina Connect result. The point of this article is to help you check whether it is happening in your own business. |
An illustration (hypothetical)
|
Imagine a small professional services firm The team starts using AI to draft replies to client queries. Each draft takes two minutes instead of ten. But the draft needs checking against the client’s file, which sits in a different system. The details are retyped into the practice software afterwards. The reply is saved in the inbox, and a note is added to a spreadsheet because that is where someone tracks outstanding items. Two weeks later, a client asks about something discussed in an email. Nobody is sure whether the system, the inbox or the spreadsheet is current. This is an invented example to show how the pattern works. It is not a description of any real business. |
In that example, the time saved at the drafting step has been spent several times over elsewhere. The cost has not disappeared. It has moved to places nobody is measuring.
How to see the hidden cost
Most businesses measure the moment AI is used. A better measure is the whole piece of work, from the first request to a finished, recorded outcome.
Pick one process where AI is already used, such as answering client queries or following up leads. Then trace it from beginning to end and note:
Where does the information first enter, and where is it finally stored?
How many times is it copied, retyped or moved by a person?
Who checks the AI output, and how long does that take?
What happens when the output is wrong or the situation is unusual?
Who owns the next step once the AI has finished?
Does the customer ever have to repeat themselves?
Then compare time to finished outcome, rather than time to first draft. If the first has not improved, the gain is probably being absorbed by the surrounding work.
What a better order looks like
When AI creates extra work, the answer is rarely to remove it. It is usually to connect the work around it.
That might mean one reliable place for customer information. It might mean an enquiry that reaches the CRM without being copied. It might mean a clear owner for checking, and a defined route for exceptions to reach a person with the full conversation attached. Sometimes it means a simpler form or a clearer handover, with no new technology at all.
Once the work is connected, AI has something solid to work with. It can prepare a response from reliable information, and the result goes where it needs to go.
If you are deciding what to put in place first, When Does a Small Business Actually Need a CRM? covers the questions to ask before choosing a system. Why Enquiries Get Lost Even When a Business Is Busy, Not Badly Run describes how information gets stranded, and AI Enquiry Assistants: Where They Fit and Where They Don’t looks at one specific use of AI in more detail.
Frequently asked questions
Does AI always create more work?
No. Used on a well-defined task, with reliable information and a clear place for the result to go, AI can save real time. The risk is highest when it is added to a process where information is already scattered, ownership is unclear and people are holding the steps together manually.
Should we stop using AI if we notice this?
Not necessarily. First find where the extra work is landing. The fix is often in the surrounding process, such as a clearer owner, a single record or a defined handover, rather than in removing the tool.
How do we know whether AI is saving time overall?
Measure the whole piece of work. Compare how long it takes to reach a finished, recorded outcome, and include checking, copying and correcting. Time to first draft on its own can be misleading.
Is human checking of AI a sign it is not working?
No. Checking is sensible, and most businesses using AI do some of it. The point is to plan for it, rather than letting it grow quietly in the gaps.
Where should a small business start?
With one process that regularly causes delay or frustration. Trace it end to end, mark every copy, check and chase, and decide whether the first improvement is a process, a system or AI.
Start with understanding
AI is not the cause in this pattern, and removing it is rarely the answer. The cost appears when it is added to work that was never connected: people, information, systems and processes each doing their part without joining up.
If you are not sure where the extra work is landing in your own business, the Davina Connect Business Review is a practical place to start. It looks at how work moves across your business now, and the recommendation does not begin with a tool.
