It's the question everyone asks, and it's the wrong place to start. "How much should we spend on Ai?" assumes the spend is the decision. It isn't. The real question is what a problem is costing you, and whether a tool would cost less than the problem. Get that right and the budget answers itself. Here's an honest way to think it through — without us quoting you a price we couldn't stand behind.
Don't start from the tool. Start from the problem.
Most bad Ai spending starts the same way: someone sees a clever tool, gets excited, and goes looking for a place to use it. That's backwards. The tool is the last thing you decide, not the first. Start instead with a problem that's actually costing you — the calls you miss, the quotes that go out late, the Friday that vanishes into admin. If you can't name the problem, you're not ready to spend a penny on the fix.
Spend where the payback is obvious
The good news is you don't need a clever way to work out whether something's worth it. The maths is simple. Take the time a job eats every week, multiply it by what your time is honestly worth, and you've got a rough yearly cost. Or, if it's about winning work, count the jobs you lose to being slow or unavailable and put a value on them. Now you have a number to weigh any tool against.
A quick way to do it on the back of an envelope:
- Hours saved. Say a job eats five hours a week, and an hour of your time is worth £40. That's £200 a week, or roughly £10,000 a year, going on one repeating task.
- Jobs won. Or say you lose two jobs a month because nobody answered the phone, and an average job is worth £300. That's £7,200 a year walking out the door.
You don't need either number to be exact. You need it to be honest. Once you have it, a tool that costs a fraction of that and clears most of the problem is an easy yes. A tool that costs about the same as the problem isn't worth the bother. And a shiny tool with no number behind it at all is just a hunch with a price tag.
Spend should follow the payback, not the hype. If you can't see the return, you can't justify the cost.
Start small and prove it
Here's the bit that saves the most money: don't buy the whole thing at once. Pick the single problem with the clearest payback, fix that one, and let it earn its keep for a few weeks before you spend on anything else. A small thing that works tells you far more than a big plan that might. If it pays off, you've got both the proof and the confidence to do the next one. If it doesn't, you've lost a little, not a lot.
This also protects you from your own optimism. It's easy to imagine a tool saving you a day a week. It's harder, and far more useful, to watch it actually do it. Let the first small win settle the argument before you scale the spend.
Be wary of the big "transformation" project
The riskiest way to spend on Ai is the all-at-once overhaul — the sprawling project that promises to change everything, costs a lot up front, and takes months before you see whether any of it works. By the time it lands, the need may have moved, the team may have drifted back to the old way, and you've spent the budget before proving a single thing. Big upfront projects tie your money to a bet you haven't tested yet.
You almost never need one. Real change tends to come in pieces — one problem solved, then the next, each paying for itself as it goes. If someone's selling you a grand transformation before anything small has been proven, that's a reason to slow down, not speed up.
Sometimes the honest answer is: spend nothing yet
This is the part most people selling Ai will never tell you. Sometimes the right move is to spend nothing for now. Maybe the problem isn't costing you enough to bother. Maybe your information is in too much of a mess for any tool to help yet. Maybe the process itself needs tidying first — and once it's tidy, the problem half disappears on its own. None of that is a failure. It's money saved and a sharper decision waiting for you down the line.
"Spend nothing yet" is a perfectly good answer, and it's one we give often. It keeps your budget for the day a problem genuinely earns it.
So, what's the number?
There isn't a magic figure, and anyone who gives you one without knowing your business is guessing. But the way to find yours is steady: name the problem, work out what it's costing you, spend a fraction of that on fixing it, prove it small, then move to the next. Spend follows payback. That's the whole thing.
If you'd like a straight, plain-English read on where Ai would actually pay off for you — and where your money's better kept in your pocket for now — that's exactly what we do. Have a chat with us, no pitch and no pressure, and we'll give you an honest steer either way.