A new price list arrives from a supplier. Someone opens the website admin, the stock system and the master spreadsheet, and types the same numbers into all three. Then they check their work. Next month, they do it again.
What the job actually costs
Manual jobs survive because the status quo is free and nobody has to sign anything. But attach a number to it and the picture changes. Take a medium catalogue, checked properly: say three hours a month of one person’s time. At a loaded cost of around £30 an hour (the wage, plus the employer costs that ride along with it), that’s £1,000 a year.
Your number will be different. Bigger catalogue, more systems, more frequent updates: multiply it up. The point is that it’s only a couple of hours is never only a couple of hours, and the job never shrinks on its own. Add the cost of the errors (a mistyped price is either a margin leak or an awkward phone call) and the boring job starts looking expensive.
What the automated version looks like
Instead of typing the same rows into three places and hoping nobody typos, you get the prices entered once, landing everywhere they need to be, with a log you can actually check. The prices can come straight out of the supplier’s email, attachment or website, so nobody re-keys them at all. Built with scripts that read the list, direct connections into your website and stock system, and, where the supplier sends a messy PDF or buries the numbers in an email, a language model that turns the mess into structured rows.
Is this the AI part?
Sometimes barely. If the supplier publishes a tidy spreadsheet, the whole thing is plumbing: no model required. The AI earns its keep when the source is human-shaped, like a PDF laid out for reading, an email with numbers buried in prose, or a manufacturer’s website with the useful figures scattered across product pages. That extraction is what language models are good at.
Where a model is involved, your data goes to a major provider’s business-grade API, on paid terms that exclude it from their training data, and comes back as a finished result. Nothing is parked there, and nothing is sold on. The full answer, including what happens if your data genuinely can’t leave the building, is in the FAQ on the AI automation page.
Proof it works: Ham Imports
Every week, someone at Ham Imports had to gather product content from a manufacturer’s website: images to save and format, technical specifications to pull out, a write-up to produce for every product. Ten hours a week of it. The automation we built does the gathering, the formatting and the first-draft writing, and the job now takes twenty minutes, most of which is spot-checking the output.
Don’t take our word for any of this. Go and use the site. The product pages, the specs, the summaries: that’s the automation working, live, today. You can’t do that with a testimonial.
The questions you’re probably asking
What does it cost? The AI Opportunity Assessment is £300, and it comes off the build if you go ahead. Small data-sync builds typically land between £1k and £3k; your assessment report puts a firm indicative range on your specific job before you’ve committed to anything.
What happens when the supplier changes their format? The tool flags it instead of quietly guessing, and the old manual way still works while we fix it. No automation should leave you with only one way to do the job.
Is our data safe? See above. The boring answer is that price lists are among the least sensitive data a business holds. If yours involves something confidential, say so at the assessment and it gets handled there.
Where to start
The assessment is a defined piece of work with a price on it, not a sales call in disguise. We meet for an hour, watch how the job is actually done, and you get back a short report listing three tasks worth automating, ranked by time saved against cost. The report is yours to keep whoever builds the thing: us, another developer, or your own team.
If re-keying price lists isn’t your boring job, there are others on the automation page: copying data between systems, writing the same summary every Friday, processing files one at a time. Or skip the reading and give us a ring. That’s a landline in Wymondham, answered by the person who writes the code.