Building a PIM Was Never the Hard Part.

Why the wave that's supposed to frighten software companies is, for the right kind of business, a tailwind.

There's a demo doing the rounds at the moment. Someone asks an AI to build a working spreadsheet application — a proper Excel clone, not a mock-up — hands it a goal, and walks away. Six days later it's still going: real cells, working formulas, the lot. It's genuinely impressive, and I'd encourage you to be impressed by it.

The conclusion most people draw is that software is finished. If a machine will build you a tool over a long weekend, why would you ever pay for one again?

I run a software company. You'd reasonably expect me to be nervous about that question. I'm not — and the reason turns out to be the most useful thing I know about this business, so let me share it.

The tool was never the deal

Here's what actually happens when a manufacturer or distributor comes to us. They don't arrive wanting software. They arrive with a problem: twenty people who need to work together to organise a mountain of product data and get it flowing — cleanly, and in sync — into their ERP, onto their website, and out to their customers.

They know something is wrong. What they almost never know is what right looks like. And that isn't a failing on their part — it's structural. You cannot see the shape of the solution from inside the problem. When you're one of the twenty people standing in the tangle, the tangle is all there is to see.

That — the seeing — is the deal. The software is the easy bit.

Three parts, and only one of them is getting cheaper

Break down what we actually do for a client and it comes in three parts.

First, we diagnose. We stand where you can't and work out what the solution genuinely needs to be — which of your ten thousand product attributes matter and which are noise, where the real bottleneck sits, what "integrated" has to mean for your business specifically. Thirty years of doing this is thirty years of pattern. We've usually met your particular flavour of mess before, even when it feels one of a kind from the inside.

Second, we build. We have a tool we built ourselves — OneTimePIM — designed to be shaped to any problem, with a facility to customise using Python agents and AI built directly into the apps. And yes, we use AI heavily to do it. We are not romantic about the build. If a machine can do a job faster and just as well, the machine should do it. This is the part that's getting cheaper, and we are glad of it, because it means we spend less of our time on plumbing and more on the parts that need a human.

Third — and this one hides behind a boring word — we stay. "Maintenance" makes it sound like patching. It isn't. It's being the people who are still there next Tuesday when something changes, still answerable, still on the hook. A system that runs smoothly is not one that was built well once and left. It's one with someone responsible standing behind it, over time.

The counterintuitive bit

Now here's the part that surprises people — including, for a while, me.

When the build gets cheap, the value doesn't disappear. It moves. It moves to the two ends — the diagnosis at the front and the staying at the back — because those are precisely the parts a clever machine, left to its own devices, can't do.

Point a capable AI at a vague goal and it will build you something remarkable. It'll just keep building — more features, more cleverness, more of everything — and it will never stop to ask whether it's building the right thing, because it can't see your business and it isn't the one who has to live with the result. It optimises for more. It has no way to optimise for right.

Which means the AI wave that's meant to frighten software companies is, for a business like ours, a tailwind. The cheaper it becomes to build, the more the whole game moves upstream — to knowing what to build in the first place, and to standing behind it once it's built. Those were always the hard parts. They were just easier to overlook when the software itself took months and cost a fortune.

What this means for you

So if you're sitting on a data problem right now — twenty people, an ERP, a website, a catalogue that fights you at every turn — here's the honest version.

Your problem was never that the right software didn't exist. Plenty of software exists. Your problem is that you can't see, from where you're standing, what the right answer is for you — and no tool, however clever, solves that simply by existing. It has to be worked out with you, by someone who has seen enough of these to know where the bodies are buried. And then it has to be looked after by someone who'll still pick up the phone.

That is the deal. It always was. The machines getting good at the middle bit just makes it easier to see.

An AI will happily build you a beautiful solution to the wrong problem. Working out the right problem is still a human sitting across a table from you, asking the awkward questions.

We're rather good at the awkward questions.