It is almost always the first question, and the honest answer is more useful than a recommendation would be. The leading tools are converging, so the choice matters far less than what you do with it. Here is what actually decides it.
Which one should we get. It is a fair question and it deserves a straight answer, so here is one: the answer matters less than you have been led to believe, and getting to it in the wrong order is the most common way money gets wasted on AI.
Two reasons, and the second is the useful one.
First, we take nothing from any of them and we would like to keep it that way. Nobody pays us to put their name in an article.
Second, the leading tools are converging. A capability appears on one and turns up on the others within months. Any ranking we published would be out of date well before you read it. The same goes for prices, contract terms and how any given provider handles the information you put in — those move constantly, and we are not going to state them as fact when we cannot stand behind them next quarter. Check that yourself, on the day, from the provider.
What moves much more slowly is the thing that actually determines whether AI does anything for a business: how it is set up, and what it is pointed at. Two businesses with identical subscriptions get completely different results from them. The difference is almost never the tool.
Three things do most of the deciding, and none of them are about which one is cleverest this month.
What you already pay for. Most businesses already have a software suite — email, documents, shared files, calendars. The major suppliers of those suites have added AI features to them. Often that is the path of least resistance: nothing new to buy, nothing new to log into, and the assistant can already see the documents that would make it useful. Worth checking carefully before you buy anything at all. You may already be paying for a version of what you were about to purchase.
Where your information lives. An AI tool is only as good as what it can reach. If your work sits in shared documents, a tool that can read those documents beats a cleverer one that cannot see them. If your work sits in a system specific to your trade, the real question is whether anything connects to it, and what that connection would take. This is the question most people skip, and it is the one that decides how much you actually get out of the thing.
Whether you need the business version. The free and personal versions are genuinely capable. But how information is treated can differ between the consumer version and the business version of the same product, and those terms change. If anything you would be typing in is confidential — customer details, commercial terms, anything covered by an agreement you have signed — then that difference matters and it is the first thing to read, not the last. We are deliberately not telling you what anyone’s terms say today. We are telling you to go and look at them before you paste anything in.
The pattern is predictable enough to describe from a standing start. A tool gets chosen, usually the one somebody read about. Logins are handed out to everybody. There is a fortnight of enthusiasm and some genuinely impressive demonstrations. Then it goes quiet. Three months later it is a line on a card statement that nobody can quite account for.
There is usually nothing wrong with the tool. Nobody’s actual work ever changed, because the tool was never pointed at a specific piece of it. It arrived as a capability looking for a purpose, and capabilities looking for a purpose do not survive a busy week.
A tool chosen before the job is found is a subscription with nothing to do.
Find the job first. One real, repeating piece of work. The useful ones share a shape: they happen often, they take a fairly predictable amount of time, they follow roughly the same pattern each time, and the judgement sits at the end rather than at every step. Writing the same kind of reply again. Getting figures out of documents that always arrive in the same layout. Turning notes from a conversation into something that goes into the system properly.
Then choose the tool that fits that job and fits where the work already happens. Not the one with the best write-up. The one that can reach the information and sits where the person is already working, because a tool in a separate window is a tool that gets forgotten.
Then watch whether anybody opens it. In the first month that is the only measurement worth taking. A tool nobody opens has no features.
One honest consequence of doing it this way: you may well end up with more than one thing. That is normal, and it is fine. Very few businesses land on a single tool that covers everything, and forcing it is its own sort of waste. What matters is that each one was bought for a job you can name.
If you would rather not work through that on your own, that is the part we do. We find the jobs first, then we talk about tools.
That order is the whole of what we do: find the job, then fit the tool to it. What to automate first and where the hours actually go are the two pieces to read before this one, not after.
Working out which jobs are worth doing first is exactly what the audit is for.
Twenty minutes on the phone, free, no obligation. Tell us what keeps coming back round and we will tell you honestly whether AI is the right answer for it.