You are probably already paying for it.

The cheapest thing you can do about AI is usually not a purchase. It is a switch you already own.

Business software has changed a great deal in the last two years. The products used for email, documents, accounts, bookings and notes have all been adding features, and most of those features arrived quietly in an update. You chose your software once and set it up once, and it has been growing ever since on your money. Very few people have been back to look.

So before you price a build, spend an hour finding out what you already have. Here is how to do that properly.

Start with the bill, not the software

Open your bank or card statement and write down every software subscription you pay for. Include the ones that renew annually. Include anything bundled with something else, and anything that arrived with a device, a broadband package or an accountant’s recommendation. You are not looking for anything clever yet. You are making a list of suppliers who have been improving their products while you paid them.

That list is the whole basis of the exercise, and most people find it is longer than they expected.

Look in the settings, not in the app

This is the part that catches people out. New features almost never appear on the screen you open every morning. That screen is deliberately kept familiar, because a supplier does not want to disturb the thing you already know how to use. The new material goes into the administration area, the settings area, or a section labelled something like beta, preview, labs or what’s new.

So go there instead. In each product on your list, open the settings or admin area and read anything you do not recognise. Find the release notes or what’s new page, because most suppliers keep a running record of what they have added and when, and that record is where the last two years of change is actually written down. If your subscription has tiers, check what your tier includes today rather than what it included when you signed up, because those change too.

Two things are worth watching for. Some features are included but switched off by default, because they need an administrator to enable them, so they will not appear until somebody with the right access goes looking. And some are included for certain kinds of user only, so it is worth checking that the people who would actually use the thing can reach it.

What tends to turn up

We are deliberately not naming products or plans here. What is included, and at which tier, changes month to month, and anything specific written today would be wrong by the time you read it. Categories are more useful anyway, because they tell you what to search the settings for.

Those five cover a fair share of what businesses otherwise pay to have built. Not all of them will be sitting in your subscriptions, and the ones that are will vary in how well they work, so try each on something real before you rely on it. But if even one of them is there, you have avoided a purchase. What you switch on costs nothing extra, so it pays for itself the first time anybody uses it.

Write down what you find under each of the five headings. Then write down what nothing you own covers. That second list is the only one worth spending money on.

Why we are telling you this when it costs us money

This advice is bad for our business and we would rather say so than dress it up. Every reader who works through that list and finds the answer in something they already pay for is a reader who does not need to hire us. That is the honest consequence of publishing it.

We are publishing it anyway for two reasons. The first is that it is true, and you would find out eventually. Paying somebody to build a thing your software already does is the most avoidable way there is to waste money on AI, and it is a poor way to begin a working relationship.

The second is that doing the check first improves every conversation that follows. Anything you then decide to buy is something you have already proved you cannot get for free. That is a much more comfortable position to spend from, and it is far easier to justify to anybody who asks you why.

An hour, the list of what you pay for, and the settings page of each one. That is the whole job, and you can do it without any help. The worst outcome is that you finish knowing exactly where your gaps are.

Once you know what is already covered, the next questions are usually what to automate first and which AI tool actually fits the job — both are part of what we do before anything gets built.