It is one of the first questions we hear from owners. The business runs on an accounting package from the early 2000s, a stock system a local firm built years ago, and a lot of spreadsheets that hold it all together. Everything works, mostly. The people who use it know its quirks. So when someone mentions AI, the natural reaction is: that sounds like something for businesses with modern systems, not us.
In most cases, that isn't true. Older software is rarely the thing that stops AI being useful. Here is why, and where the real limits are.
AI mostly works on what goes in and comes out
A lot of the time-consuming work in an established business doesn't happen inside the core system at all. It happens around it. Someone reads a supplier email and types the details into an order screen. Someone pulls a report out, pastes it into a spreadsheet, tidies it up and sends it to a customer. Someone reads a PDF contract to find the renewal date.
That is where today's AI tools are strongest: reading documents and emails, pulling out the details that matter, drafting replies, summarising reports, and putting information into a consistent shape. None of that needs your core system to change. The AI does the reading and preparing; your team, or a simple connection, puts the result where it needs to go.
Three ways AI can connect to older systems
When we look at a business's software, the question isn't how old it is. It's how information gets in and out. There are usually three options, from simplest to most involved.
- Files and exports. Almost every system, however old, can export a report or a spreadsheet, and many can import one. That alone is often enough. An AI step can read the export, do the work, and produce a file ready to import or review.
- The database or an existing interface. Many older systems sit on a standard database, or have a way for other software to talk to them that nobody has used in years. With care, and with read-only access to start, these can feed information to an AI tool directly.
- Working through the screen. Where there is truly no other way in, software can be set up to use the system the way a person does: opening screens, copying values, filling in fields. It works, but it is more fragile, so we treat it as a last resort.
Most first projects use the first option. It is the quickest to set up, the easiest to check, and the least likely to disturb anything that already works.
Where the real limits are
Old software can make some things harder, and it is better to know that up front.
Messy data. If customer names are spelled three different ways, or important details live in free-text notes, AI can often help tidy things up, but someone who knows the business needs to check the result.
Support and access. If the company that wrote your system no longer exists, or your contract doesn't allow third-party access, some routes are closed. That is worth finding out before anyone spends money.
Security. Older systems often have weaker controls over who can see what. Connecting anything new, AI or not, needs a clear view of what information it can reach and where that information ends up.
None of these is usually a reason to stop. They are reasons to start small and choose carefully.
You probably don't need to replace anything first
Owners are sometimes told that before they can benefit from AI, they need a new system. Occasionally that is right. More often, replacing a core system is a large, risky project that ties up your best people for a year or more, and the AI benefits could have been had much sooner.
A better order is usually: find one piece of repetitive work around your current system, make it faster with AI, and learn from that. If a replacement is genuinely needed, you will make that decision with much better information about what you actually need.
How to find out for your business
The quickest way to get a real answer is to look at real examples. Pick two or three tasks your team does every week that involve copying, checking or rewriting information. Note which systems the information comes from and where it ends up. That short list is usually enough for someone experienced to tell you what is practical, what isn't, and roughly what it would take.
That is exactly what our Executive AI Session does: half a day with you and two or three of your key people, working through your own documents and systems. If you already know the task you want to tackle, an Operational Improvement Project takes one of them from idea to something working.
Your software may be old. That rarely means AI can't help. It just means the first step should fit around what you already have.
