Does AI need an API? No.
“Our software has no API, so we can’t use AI” is one of the most common reasons businesses rule themselves out. It’s wrong — and we can prove it from our own products, both of which are live without one.
Why does everyone assume AI needs an API?
An API is simply a doorway that lets one piece of software talk to another. For the last decade, almost every “integration” you’ve heard about went through one — so the idea has stuck that no doorway means no automation. Software vendors reinforce it, because integration projects are what they sell. And if your booking system, job sheet or stock program went in fifteen years ago, the chances are it has no API at all, or the supplier charges heavily for access.
So plenty of business owners have done the maths in their head: no API, no AI, conversation over. That conclusion costs real money, because it’s built on a false premise.
Does AI actually need an API? No.
An API is one way in. It is not the only way in, and it’s often not even the best one.
AI can work through whatever interface already exists. The emails your customers send. The files your system exports. The documents and PDFs in your shared drive. WhatsApp messages. Calendars. The same screens your staff read every morning. If information moves through your business in any form a person can handle, AI can be put to work on it — without your software knowing anything has changed.
We’ve written before about why old systems don’t stop AI automation. The missing API is just the sharpest version of that worry, so let’s deal with it using two products we built and run ourselves.
How did we ship a taxi booking assistant without touching dispatch software?
Our DM Taxi Assistant is live with real UK taxi operators. It sits on the operator’s website chat, WhatsApp and social channels; it quotes the fare, books the job, takes card payment through Stripe, and hands the confirmed booking to the operator’s existing dispatch. It doesn’t answer phone calls, and — this is the part that matters — it is not plugged into any dispatch system’s API.
That was a deliberate design decision, not a limitation we tripped over. Dispatch software in the taxi trade ranges from modern platforms to systems old enough to vote, and every one speaks a different language. If we had made a dispatch integration the price of entry, we could only have served operators whose software happened to cooperate.
So version one does something simpler: it captures and confirms the booking, then hands it off the way the operator already works. The AI does the conversation — the quoting, the back-and-forth, the payment — and the finished, confirmed job lands in the operator’s existing routine. Nothing about their dispatch changes.
That one decision meant any operator could go live no matter how old their dispatch software is. The age of the system stopped being a question we ever had to ask. One operator, Matthew McElhinney, put the result plainly: “I can now do a week’s worth of work in a day.” His dispatch software didn’t change. The work arriving at it did.
What if there’s no API anywhere near the job?
Our other product, Olly, is built for tradespeople. A plumber finishes a job, records a WhatsApp voice note describing what he did and what it costs, and Olly turns it into a branded PDF quote or invoice, synced to Xero or QuickBooks.
Look at where the APIs are in that build. On the tradesman’s side: none. His “system” is WhatsApp and his own voice. There is no software to integrate with because there is no software — and that’s fine, because the AI meets him where he already is. The only API in the whole product sits at the accounting end, where Xero and QuickBooks offer good, modern connections and it would be silly not to use them.
That’s the honest shape of most real automation: APIs where they exist and earn their keep, and ordinary human channels — messages, documents, voice — everywhere else.
So when is an API actually worth having?
We’re not anti-API. Where a system offers a good one, use it: the connection is cleaner, faster and has fewer moving parts, which is exactly why Olly talks to Xero and QuickBooks that way. If your next software purchase offers a well-documented API, that’s a genuine point in its favour.
The point of this article is narrower and more useful: an API is optional, not required. It should never be the thing that decides whether your business can use AI. If you want a quick way to judge what you’re actually working with, our companion piece — Is our software too old for AI? A 5-minute test — walks through it.
What if a system has no API and no exports either?
Some software gives you nothing: no doorway, no export button, no files. Even then, ask one question — can a person operate it by reading the screen and typing? If yes, AI can be taught to do the same job the same way: read what’s on screen, follow the rules your best member of staff follows, and enter the result.
We put the same controls around that as around everything we build. It’s tested on your real cases before it goes live. Decisions above a threshold you set wait for a human. And every action is logged, so you can always see what was done and why. The AI gets the boring volume; your people keep the judgement calls.
Where do you start?
Not by asking your software vendor about their API. Start by listing where your hours actually go — the quoting, the chasing, the retyping — and asking which of those an AI could take on through the channels you already use. That’s precisely what our AI opportunity audit does: we look at how your business really runs today, find what’s worth automating, and then our in-house developers build it into the systems you already have. We’ve spent years making equipment and software work with whatever was already installed — that habit came from our hardware days building RTK networks and mapping systems, long before AI — and it’s exactly the habit this work needs.
If the honest answer is that AI won’t pay for itself in your business yet, we’ll tell you that too.
Quick answers
Our booking software is fifteen years old and has no API. Can we still use AI?
Yes. The AI works in front of your software, not inside it. It handles the customer, confirms the details, and passes the finished job to your system the same way your staff already do. Your software doesn’t change at all.
Is working without an API less reliable?
Not if it’s built properly. Every automation is tested on your real cases before it goes live, decisions above a threshold you set wait for a person, and every action is logged. The route in matters less than the checks around it.
Will AI make changes in our systems without us knowing?
No. You set the rules for what it can do on its own, anything bigger waits for a human to approve it, and there’s a log of every action. You can see exactly what it did and when.
Do our staff have to learn new software?
No. The whole point of working without an API is that the AI fits around how you already operate. Work arrives in the same place, in the same format, as it does today — there’s just less of it to do by hand.