Almost every conversation about AI comes back to productivity. You watch a demo, a report appears in seconds, an email writes itself, and the admin that used to eat a morning is handled before the kettle boils.

It is easy to look at that and assume the savings follow automatically.

Sometimes they do. Often they do not, and the difference is worth understanding before you renew another subscription.

Hi, I am Andy from Initial IT in Lichfield. We look after Microsoft 365, cyber security and IT support for small businesses across Staffordshire and the West Midlands, and this is the question business owners ask me most often once the novelty has worn off: we are definitely faster, but are we actually better off?

Why does the saving disappear somewhere between the demo and the accounts?

Say someone on your team uses AI to draft a proposal. The document appears in a minute instead of taking half an hour.

That proposal still needs reviewing. Details need checking, figures need confirming, and the wording usually needs adjusting so it sounds like your business rather than everybody else. The finished result is often better, because a person has spent their time improving something rather than starting from a blank page.

That is not a criticism of AI. It is how most useful tools work. A spreadsheet does not remove the need for accounting knowledge, and a power tool does not remove the need for a skilled tradesperson. Good technology helps people work more effectively. It rarely removes the person.

So the twenty-nine minutes you saved are real. Whether they turn into money depends entirely on what happens to them next. Twenty-nine minutes spent on billable work or on a customer who was about to leave is money. Twenty-nine minutes absorbed into the day is not.

What does AI actually cost once it has spread through the business?

The first subscription is easy to account for. The problem is what comes after it.

One department adopts a tool for content. Someone else starts paying for a research platform. A third person finds an AI note-taker that joins meetings. Within a year a business can be paying for four tools that do broadly similar jobs, none of which appear on the same invoice or in the same budget line.

We see the same pattern from the other side when we audit a Microsoft 365 tenant. There is usually capability already paid for and switched off, sitting next to a third-party tool bought to do the same job. Before you buy anything else, it is worth knowing what your existing Microsoft 365 licences already include.

What happens when the approved tools do not fit?

This is the part that turns a cost conversation into a risk conversation.

When the sanctioned tool does not do what someone needs, people find a way round it. A personal account gets used. A free version fills the gap. Information starts moving between systems that were never designed to work together, and nobody set out to do anything wrong.

At that point the question is no longer whether you are getting value. It is where your customer data has ended up, who can see it, and whether you could answer that question if a client asked. That is squarely a cyber security problem wearing a productivity badge.

What should you point AI at instead?

Spend less time asking which tool to buy and more time asking where work is actually slowing down. In most businesses the honest answers are unglamorous:

• Customer enquiries taking too long to get a first response
• Reports that need hours of manual assembly every week
• A process that depends on somebody copying information between two systems all day

When AI is aimed at a problem you can describe in one sentence, you can tell within a month whether it worked. When it is introduced because the sector is talking about it, the technology arrives first and the business case gets written afterwards, which is how you end up with a renewal you cannot justify and cannot quite cancel.

Where this pays off best is when the repetitive step disappears completely rather than getting faster, which is usually an automation job rather than a chatbot one.

How do you tell whether it is working?

Not from a usage dashboard. A high usage figure tells you people opened the tool, which is the least interesting thing you could measure.

Six months in, three questions will tell you more than any vendor report:

• Are people getting through their work more easily, or just differently?
• Has anything a customer would notice actually improved, such as response times or accuracy?
• Is the business running more smoothly than it was, and can you point at where?

The encouraging part is that most businesses are still early in this. There is time to step back and look properly before the spend becomes something you inherit rather than something you chose.

Quick check: is your business ready for tools like this?

Our free AI & Copilot Readiness Quiz takes two minutes and shows you where your business stands: what is ready, what is missing and what to fix first. No marketing list.

Take the 2-minute AI Readiness Quiz

If you would like a straight answer on where AI is helping in your business, where it is quietly creating extra work, and what you are already paying for twice, we are happy to go through it with you.

Frequently asked questions

How long should it take to see a return on an AI tool?

For a narrow, well-defined job you should see something within a month, because you can measure the same task before and after. If a tool has been in the business for six months and nobody can name what it improved, that is your answer. Broad “make everyone more productive” rollouts are the ones that never resolve either way.

Is Microsoft Copilot better value than buying separate AI tools?

It often is, for the simple reason that it works inside the data you already hold in Microsoft 365 rather than requiring you to paste that data somewhere else. Whether it beats a specialist tool depends on the job. What we usually find is that businesses are paying for both and using neither properly.

How do I find out which AI tools my team is already paying for?

Start with the card statements and expense claims rather than asking around, because small subscriptions rarely go through a buying process. Then check your Microsoft 365 tenant for third-party apps that have been granted access. The second list is usually the more surprising of the two.

Should I stop people using AI until we have a policy?

A ban is very difficult to enforce and tends to push the activity somewhere you cannot see it. It is more effective to name a small number of approved tools quickly, say plainly what must never be pasted into a public tool, and then write the fuller policy while people have somewhere legitimate to work.

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