One team for the systems
your business runs on.

Off-the-shelf software makes your team work its way. We do the opposite: we build software shaped around the way your company already works, running on your own servers, using your own data.

Some of that happens at a desk. Some of it happens at a receiving door, with a scanner in one hand. It’s the same system either way, and that’s the point — the moment the floor gets its own separate tool, the numbers start to disagree.

At the desk

• One screen instead of the six-system copy-and-paste job every time you add a product
• A buying tool that works out reorder points from what actually sold, instead of gut feel
• A statement importer that reads spreadsheets, CSVs and scanned PDFs and sorts the lines for you
• A client intake form with no login, that turns a finished questionnaire into a tracked account
• Dashboards that read live from the systems you already run

On the floor

• Scan-and-check for incoming deliveries, so a short shipment is caught at the door instead of at month end
• Bin and shelf tracking, with quantities adjusted right there in the aisle
• Two-scan pack and check stations, with every scan tied to the person who made it
• Cycle counts on a phone, written to a proper record instead of a clipboard

How they’re built

Whatever suits the job. Your existing database or a new one. Docker, a one-line installer, and updates that apply themselves.

The floor tools are ordinary web apps, not phone apps — so there’s no app store, no device enrollment, and no worrying about which handheld is running an old version. They run on the scanners and phones you already own, and an update is a page refresh.

Where they need a login, they check against the staff list already in your ERP. A new hire is added once, in the place your team already manages people. Someone switched off there loses floor access immediately. Every change carries a name and a time, so you can always see who did what.

You get the code, the installer, and documentation your next hire can actually read.

Have a look at the warehouse floor suite for four of these running in one building.

Most companies don’t need new software. They need the software they already have to talk to each other. We build the piece in the middle that moves the records a person is currently carrying by hand.

Things we’ve connected

• Two back-office databases and a Shopify store, worked into one true stock number and pushed on a schedule
• Orders pulled from Shopify and from an invoicing system into one shipping desk, with tracking written back to both
• Barcodes and prices kept identical across several back offices and several storefronts at once
• A new product entered once and delivered to the ERP and the store, with images, variants and per-store pricing

Why ours keep running

Sending the same value twice leaves the same result, so a retry can never corrupt a count. We plan for the limits a platform puts on how often you can call it, instead of finding out the hard way. And anything that costs money is ordered carefully: the step you can’t undo happens first, and everything after it can be retried on its own. Buying a shipping label never depends on some other system being up at that moment.

Everything that moves gets written down. So “why did this change at 3pm?” is a ten-second answer instead of an argument.

Every business has jobs that eat hours and add nothing: the Monday report, the reorder spreadsheet, the copy-paste between two screens, the chase for a status update. Those are the things worth automating, and they’re usually the cheapest work we do.

Automations already running

• Reorder levels worked out from what actually sold, with overrides per product, exported as a spreadsheet or PDF the supplier can read
• Bank and card statements pulled in from CSVs, spreadsheets and scanned PDFs, sorted using what you corrected last time
• Catalog pages scored against clear rules and rewritten by AI, with a person approving every change before it goes live
• Listings hidden and brought back automatically as stock runs out and returns

Where the judgment stays

Automation is only worth trusting if it knows what it doesn’t know. Ours suggest rather than decide anywhere a mistake would be expensive. An AI-written product description is a suggestion until someone approves it. An unusual reorder quantity gets flagged instead of sent. You trust it because you can see how it got there.

We start with the one job costing your team the most hours, automate it end to end, and only move on once it’s been boring for a month.

Most AI pitches start with the model. We start with the job, and with what it costs you when the answer is wrong.

The work worth pointing it at is the work that never gets done: too big for the hours you have, too varied to write rules for. Nine hundred product pages with no descriptions. A drawer of scanned invoices. Three years of transactions nobody has sorted.

Anything a plain rule can still handle, a rule keeps handling. It’s faster, it’s cheaper, and it can’t surprise you.

What we’ve actually put it on

• Product pages written for you. Titles, descriptions and image text across thousands of products, written from what people actually search for, and scored before and after so you can see it worked.
• Paperwork typed up for you. Supplier statements, invoices and scanned PDFs turned into rows you can use — including the ones that arrive as a photo of a page.
• Endless sorting, done once. Transactions, products and tickets filed automatically. Correct one and it stops asking you about that merchant for good.
• Getting found when customers ask an AI. Whether the assistants your customers now ask are recommending you, what they say when they do, and what it takes to change it.

What stops it going wrong

It won’t invent things about your business. Everything it writes comes out of your own records — your products, your numbers, your search data. If a fact isn’t in there, it doesn’t reach the page. No made-up specifications for a customer to find and no invented claims to answer for.

Nothing changes until you say so. Every suggestion waits for a click, with its reasoning beside it. Your catalog, your prices and your books are never edited by a machine working alone. A bad suggestion costs you five seconds instead of a week of cleaning up.

The bill goes down, not up. It remembers every correction, so it asks the model less often as it goes. Month three costs less than month one for the same work, which is the opposite of how a subscription behaves.

A bad day at the AI company isn’t one at yours. The AI runs last, not underneath everything. If a model goes down or slows to a crawl, the rest of the system carries on and you approve a few more things by hand.

Why it stays affordable

Most of your records never reach the model at all. Your own rules take the obvious ones. What you’ve already corrected once is remembered. Your catalog and your history answer the next batch. Only what’s left — usually a small slice — costs you anything, and that slice shrinks every month.

What this isn’t

We won’t sell you a chatbot for your website, and we’ll tell you when AI is the wrong tool. A good deal of what gets pitched as AI is a lookup table with a subscription attached. If a rule solves your problem, we’ll write the rule and charge you less.

Your data already holds the answers: what’s selling, which products earn their shelf space, which customers are actually profitable, how the season is going. A reporting system puts all of that in front of you without anyone exporting a spreadsheet.

• Dashboards with trends, totals and period-on-period comparisons
• Profit and margin by product, customer or channel
• Charts that read live from your ERP, your store and your accounting system
• Scheduled reports delivered to your inbox as a PDF or a spreadsheet

Built on the systems you already run, so the numbers always match the source — and keep themselves up to date.