You bought the software to save time.

Now you've got three browser tabs open, two prompts that almost work and a colleague asking whether the AI-written email should really begin, “I hope this message finds you well.”

Welcome to the productivity revolution.

Please allow 20 minutes for setup.

The robot has arrived. Has anything improved?

AI is appearing in more workplaces, from small independent businesses to large organisations with entire departments dedicated to making technology work.

The promise is irresistible. Fewer repetitive jobs. Faster paperwork. Less time staring at a blank screen.

And occasionally, it delivers exactly that.

But there's an awkward difference between using AI and getting something useful out of it.

A business can subscribe to three AI platforms, generate hundreds of documents and still have staff working late.

In fact, it may have created an entirely new job: checking what the machines have produced.

The two-minute job that takes seventeen

Imagine a task that normally takes ten minutes.

An AI tool completes the first draft in two.

Brilliant.

Except someone then spends fifteen minutes checking the facts, removing invented details and making it sound like something a human being would actually send.

The ten-minute task now takes seventeen.

The robot hasn't earned its tea break.

That's not an argument against AI.

It's an argument for measuring the whole job rather than celebrating how quickly the first draft appeared.

The important number isn't how fast AI generates something. It's how long the finished, usable result takes.

Where AI actually earns its keep

Some of the most convincing applications are surprisingly ordinary.

Turning untidy meeting notes into a readable summary.

Preparing a first draft of a routine email.

Organising information from several documents.

Producing questions for an upcoming client meeting.

Helping someone who knows their subject but struggles to get started with a blank page.

None of these is likely to make a spectacular conference presentation.

But saving twenty minutes on a repetitive task, several times a week, can be more valuable than an impressive demonstration nobody uses again.

The trick is choosing work where AI genuinely removes effort, rather than adding another layer of administration.

Congratulations. You've hired yourself another supervisor.

Here's where things get interesting.

AI doesn't always replace work. Sometimes it moves it.

Instead of writing the document, you're reviewing the document.

Instead of answering the customer, you're checking the automated response.

Instead of researching from scratch, you're verifying whether the generated answer is accurate.

That can still be a substantial improvement.

But only when the checking takes less time than doing the original job.

And some tasks deserve particularly careful handling. Customer confidentiality, financial decisions, legal wording and medical information aren't places to accept an answer simply because it arrived confidently and quickly.

A convincing mistake is still a mistake.

The subscription nobody wants to cancel

Then there's the collection of software subscriptions.

One for writing. One for meetings. One for images. Another for customer enquiries. Something else that promises to connect all the others.

Each sounds inexpensive on its own.

Together, they become another monthly business expense and another collection of tools somebody has to learn.

Before buying the next one, there's a question worth asking:

What specific problem are we paying this to solve?

Not “How can we use more AI?”

Not “What are our competitors doing?”

Just: what work needs improving, and how will we know if this makes it better?

If nobody can answer that, perhaps the credit card should stay in the drawer.

The Friday-afternoon test

Pick one task your business repeats regularly.

Measure how long it takes now.

Then try doing it with AI, including the time spent preparing instructions, checking the result and correcting mistakes.

Compare the quality as well as the minutes.

If the new method works, keep it.

If it doesn't, change the process or cancel the subscription.

There's no shame in discovering that the old way was faster.

The objective was never to have the most AI tools in Cambridgeshire.

It was to run a better business — and perhaps, just occasionally, finish work on time.