Nobody asked for a longer email.
They asked whether the order had shipped.
That’s the awkward question behind Microsoft Copilot: is it helping people finish useful work, or making ordinary tasks look more impressive while taking just as long?
For UK business owners, the answer matters. An AI assistant can be useful, but another subscription needs to earn its place. Especially if its main contribution is turning ‘Yes, Tuesday is fine’ into three paragraphs about calendar alignment.
The sensible approach isn’t to dismiss Copilot or hand it the keys. It’s to test where it helps, measure the whole job and keep human judgement firmly involved.
More output isn’t necessarily more work done
AI tools make producing text easier. That doesn’t automatically make a business more productive.
A longer proposal might still leave the customer’s main question unanswered. A polished meeting summary might miss the decision that matters. An email drafted in seconds might need several minutes of editing before anyone should send it.
The useful measure is completed work, not generated words.
Did the customer get a clear answer? Did the spreadsheet help someone make a decision? Did the meeting finish with agreed actions and owners?
If Copilot helps achieve those outcomes with less effort, it’s doing something valuable. If it simply adds a drafting stage to a task that was already straightforward, the benefit is less convincing.
Start with an everyday irritation
Don’t begin with ‘How do we get everyone using AI?’
Begin with ‘Which recurring jobs take longer than they should?’
Perhaps your operations manager spends Monday morning turning updates into a weekly report. Perhaps someone regularly checks spreadsheet data for inconsistencies. Perhaps your sales team struggles to turn rough notes into a clear first draft.
These are better starting points than a general instruction to use Copilot more often.
Choose a task with a clear beginning, a clear finish and an obvious owner. Describe what a good result looks like before trying the tool.
For a weekly report, that might mean an accurate summary, outstanding issues and named next steps. Not a beautifully written introduction about the importance of operational excellence. Everyone can probably survive without that.
Measure the whole task, including the checking
A quick first draft is useful only if it leads to a quicker acceptable result.
Time the existing process, then compare it with an assisted version across several similar tasks. Include preparation, prompting, checking, editing and any corrections afterwards.
You don’t need an elaborate measurement system. A simple record can capture:
- What task was completed.
- How long it took from start to finish.
- Whether the result met the agreed standard.
- How much checking or reworking was needed.
- Whether it caused extra work for anyone else.
That final point is easily missed. Saving the sender a few minutes isn’t much of a win if several colleagues then have to read an unnecessarily long message.
Measure the effect on the workflow, not just the person pressing the button.
Give email a stricter brief
Email is an obvious place to try AI assistance because there’s usually plenty of it. Unfortunately, there’s also plenty of scope for making it worse.
The answer is to be specific about the job.
Instead of asking for a ‘professional email’, ask for a short reply that answers the customer’s question, states the next step and avoids repeating the background.
For example: ‘Draft a reply under 100 words. Confirm that we’ve received the request, explain that we need the invoice number and ask the customer to send it. Don’t promise a resolution date.’
Then check it. Has the draft invented a commitment? Does it sound like your business? Could you remove half of it without losing anything useful?
Sometimes the best use of Copilot is shortening something, not expanding it. Sometimes the best option is typing the two-sentence reply yourself.
Use spreadsheet help with sensible caution
Spreadsheet tasks can offer a more concrete test of usefulness than email drafting.
Depending on the Copilot product, licence and features available, assistance might help someone explore data, understand a formula or prepare an analysis. The important question is whether the result is correct and useful, not whether it appeared quickly.
Start with a familiar dataset where the person reviewing the output understands the expected result.
Check totals against the source. Look for missing rows, misleading assumptions and incorrect interpretations. Ask whether the suggested analysis answers the actual business question.
An attractive summary of the wrong figures is still wrong. It has simply dressed for the occasion.
For financial, contractual or other consequential work, agree who must review the output before it’s used.
Check access before adding convenience
Before a rollout, establish which Copilot product you’re considering and how it handles your business information. Products and configurations differ, so assumptions aren’t a reliable policy.
For Microsoft 365 Copilot, existing permissions matter because it can surface information a user is already permitted to access. That makes poorly managed sharing worth addressing before you introduce another way to find content.
Review shared folders, document permissions and sensitive material. Decide what staff may enter into approved tools, and what requires extra care.
People also need to know when to stop and ask. Customer information, employee records and confidential commercial documents shouldn’t become training exercises simply because someone wants to try a new feature.
Clear rules are more useful than a vague instruction to ‘be sensible’.
Run a small pilot with a real decision at the end
Choose a small group, a few repeatable tasks and a defined review period. Give people practical guidance and somewhere to raise problems.
At the end, review outcomes rather than enthusiasm alone.
Which tasks became easier? Which outputs needed too much correction? Did colleagues or customers receive clearer information? Were any steps removed, or did you just add another tool?
Keep the useful applications. Change the weak ones. Stop the ones that don’t justify the effort or cost.
You don’t need to prove that Copilot is brilliant or useless. You need to establish where it’s worth using in your business.
Make usefulness the standard
Copilot shouldn’t be judged by how often people open it, how much text it produces or how impressive the demonstration looks.
Judge it by accurate work completed, clearer communication and fewer avoidable steps. Keep the checks proportionate to the risk, and give staff permission not to use it when the ordinary method is quicker.
If you’re considering Copilot, start with the work that needs improving, then look at licensing, access and a manageable pilot. That gives you a firmer basis for a decision than buying seats and hoping productivity follows.
Let’s make IT one less thing to think about. morsenetworks.co.uk.