Old systems don’t stop AI automation.

“Our systems are too old for AI” - the objection we hear most

Almost every company we talk to says a version of the same thing: the ERP went in years ago, half the business runs on Excel, orders arrive in a shared inbox, and the delivery notes are scanned paper. So AI must be for other companies - ones with modern software and clean data.

That belief costs real money, because the companies with the oldest systems usually have the most manual work - and the most to gain from automating it. So let’s answer the question properly.

Do old systems prevent AI automation? No.

The reason is simple: modern AI reads anything a person can read. Spreadsheets, emailed PDFs, scanned delivery notes, exports from a green-screen terminal, free-text notes in a CRM. Your staff make sense of those every day - AI can too, and it doesn’t get bored on row 4,000 or misread a total at 5pm on a Friday.

Old systems were a genuine blocker in the last wave of automation, which needed tidy databases and modern connections to work. That’s the wave most people formed their opinion in. It no longer applies.

How does AI actually connect to old software?

You build a layer in the middle - we call it the bridge. It reads whatever your systems produce, turns it into clean data, lets AI do the routine work within rules you set, and writes the result back into those same systems, the proper way. Your people and your processes keep working exactly as before, minus the drudge work.

Does AI need an API to work with our software?

No. A modern connection helps, but it isn’t required. Where there’s no API, the AI works the way your staff do: it reads the export, fills in the same fields, follows the same rules. Where records live on paper, scans are fine. Where the “integration” is a nightly file dropped on a server - that works too. If a person can operate your system, AI can be taught to operate it, with every action logged. We’ve answered this one in full, with a worked example from our own products, in Does AI need an API? No - here’s how it works without one.

What about our data - and do we have to replace anything?

Nothing gets replaced. Your ERP stays. Your data stays where it is. The bridge is a layer on top, built to be removed as cleanly as it was added. We also build on neutral tooling such as Langdock, so if a better AI provider comes along next year, you switch providers - not systems. And the big decisions still pass a human before anything happens.

What does this look like in practice?

Two products we built and run ourselves, both live today:

The full story of the first one is in our DM Taxi Assistant case study.

Where do you start?

Not with a rebuild. Start by seeing where you stand: our five-question check, Is our software too old for AI? A 5-minute test, tells you in five minutes whether AI can work with what you run.

Then come and talk it through - over a coffee at our Farringdon office, at yours, or on a call. You get a short written plan: what to automate first, a fixed price, what it should save, and how it connects to what you already run. If the numbers don’t stack up, the plan says so.