Someone sold you a big platform. It handles the core workflow. But every day your team still copies figures between screens, reads PDFs to pull out a handful of numbers, and chases the same confirmations by email. That gap is not a flaw in your system. It is just where the vendor stopped building. You can close it without replacing anything.

Why the platform you have will never do this for you

Enterprise software is sold on breadth. The vendor needs the product to work for thousands of customers, so they build the common path and leave the edges to you. Document parsing, exception triage, status-chasing — these vary too much per business to standardise. So they ship an API, write it up in the docs, and move on. That is not a criticism. It is just how the economics work.

The result is a capable platform sitting next to a pile of manual work your team absorbs quietly. Nobody raises a ticket about it because it has always been that way.

What “AI on top” actually means

It means a lightweight layer that connects to your existing system through its API or export, does one job the platform skips, and writes the result back — or sends it wherever it needs to go. Your team keeps working in the same screens. The AI handles the repetitive step they used to do by hand.

Three patterns come up most often:

  • Document extraction. A PDF arrives — a supplier invoice, a customs declaration, a delivery note. Someone opens it, reads it, and types the relevant fields into the platform. An extraction layer reads the document, pulls the structured data, and posts it directly. The human reviews the exception queue, not every single document.
  • Inbox triage. Emails land in a shared mailbox. Someone decides what each one is, who it belongs to, and what needs to happen. A classifier reads each message, tags it, routes it to the right place, and drafts a reply for approval. The human approves or edits. They stop triaging from scratch.
  • Status and exception monitoring. The platform holds live data. Nobody is watching it continuously. A monitoring layer checks for conditions you define — a shipment overdue, a stock level below threshold, a document missing — and sends an alert with context attached. The human acts. They stop checking dashboards to find problems.

A worked example you can run on your own numbers

One UK supply-chain operator was processing a high volume of freight documents each day. Each document required a person to open it, find the relevant fields, and key them into CargoWise. Conservatively, that took 30 minutes per document. We built an extraction layer that reads the document, structures the data, and posts it to the platform. The human now reviews exceptions only.

Run the same sum on your operation. Count the documents your team processes in a day. Estimate the honest time per document, including interruptions. Multiply by your blended hourly cost. That is the floor of what the problem is worth. The build to fix it starts at £4,000.

What you do not need to do

You do not need to migrate. You do not need a new licence. You do not need to retrain your team on a different interface. You do not need to wait for your platform vendor to ship a feature. The system you already pay for keeps doing what it does. The AI layer handles the work it was never going to do.

The honest limits

This approach works when the manual work is repetitive and the inputs are consistent enough to model. If every document is genuinely different in structure, extraction accuracy drops and the exception queue grows. If the “manual” work is actually judgement — a negotiation, a relationship call, a complex decision — automation is not the right tool. We will tell you that before we build anything.

Where to start

Pick the one task your team does most often that feels like copying or checking. Estimate the time. Ask whether the input is structured enough to automate. If the answer is yes to both, you have a candidate.

We run a short audit — from £750 — that maps exactly this: where the manual work sits, whether it is automatable, and what a build would cost against what it would save. No commitment beyond that.

If you already know what you want to automate, see how we price system builds.