Use the ERP
You Already Paid For

When a business brings me a problem, the reflex, theirs and the whole industry's, is to reach for something new. Build a custom app. Bolt on an AI tool. Buy another subscription. There's always a vendor happy to sell you a new thing to solve a problem the thing you already own could probably handle.

So my first move is almost always the opposite one. Before we build anything, I want to know what the system you already have can do that you're not using.

Most companies use a fraction of their ERP. You bought a system that does far more than payroll and invoices, and then real life happened. Nobody had time to learn the rest, the reports were hard to build, and the capability just sat there. You're paying for it every month. It's the most expensive software you own and the least fully used.

Here's where AI actually earns its place, and it's not where the hype points. The exciting story is "build an AI app." The useful story is quieter: AI is the best tool yet for helping you finally use the system you already have. It can write the report you could never get out of the thing. It can read the manual you'll never read and tell you which module you already pay for does exactly what you need. It can help you configure a workflow the ERP already supports, instead of standing up a parallel one next to it. It turns the complexity of your own system into something you can actually ask questions of.

That is a completely different move than writing new software, and for most problems it's a better one. Nothing new to integrate. Nothing new to maintain. Nothing new to be locked into when the vendor changes their pricing. The answer stays inside your system of record, where your data already lives, instead of in some app bolted to the side of it. It's cheaper, it's faster, and there's far less that can break later.

I'll be honest about the irony here, because it matters. I build software for a living. I've built an AI agent, a handful of tools, a whole product. And I still start every engagement by trying not to build anything. I only reach for new software once I've satisfied myself the existing system genuinely can't do the job, not because building is the more interesting option. Most of the time, "can't do it" turns out to be "nobody figured out how." The problem wasn't missing software. It was locked capability, and the key was sitting right there.

This is the same discipline as everything else I believe about this work. You don't automate a broken process, you fix it first. You don't buy a strategy, you use tools for the things tools are good at. And you don't write new software to do a job the software you already own can do once someone shows it how. AI, pointed at that question instead of at building the next shiny thing, is genuinely useful.

So the first question was never "what should we build?" It's "what can the system I already have do that I'm not using?" Start there. Reach for AI to answer that before you reach for it to build. You'll be surprised how often the problem was never a software problem at all.

The cheapest, safest software you'll ever deploy is the software you already own, finally used the way it was meant to be.

← Back to all posts