You get your own email address for documents the day you start. After that there's nothing to learn and nothing to click — the paperwork arrives, gets read, gets checked, and turns up where you need it. You only hear from us when something genuinely needs you.
Forward the PDF to your own pagesift address — or drag it onto the page. You can give the address to your suppliers and let the bills arrive on their own. Nothing to install.
We work out whether it's an invoice, a statement or something else, and read every line off the page. Then we do the sums, the way you'd do them with a calculator if you had the time.
Documents that check out arrive in your accounting system, as a spreadsheet, or in a folder — laid out exactly the way that system expects. The rest wait for you, with the problem already pointed out.
They all end up in the same place and get treated exactly the same way. Most people end up using the email address and forgetting the others exist.
Something like [email protected]. Forward bills to it, hand it to your suppliers, or put it on your invoices so the paperwork comes to it directly. It just fills up with documents that are already read, checked and filed. If you look after several businesses, each one gets its own.
And when a bill gets forwarded on to you, we still work out whether it really came from who it says it did — the check most systems lose the moment an email is passed along.
Got a stack on your desk? Drop them straight onto the page. If you already know what a document is, you can say so and skip the guessing.
If you've got someone technical, your system can hand documents over directly and collect the results back. Same reading, same checks, no person in the middle.
An invoice and a bank statement don't just look different — they're wrong in different ways. Reading them the same way is how wrong numbers end up in your books.
Before anything else, a quick look at the file itself — is this a proper digital PDF, or a scan of a piece of paper? That changes how carefully we go about the rest.
An invoice, a statement, a repair order, a credit note. This decides how we read it — the questions worth asking of a bank statement are not the ones worth asking of an invoice.
Every line item, every transaction, every number. Money is taken off the page exactly as printed — we never round it or tidy it up on the way in. That comes later, and only if you asked for it.
And here's the safety net underneath all of it: working out what a document is only ever decides how we read it. It never decides how we check it. The sums we do are chosen by what's actually on the page. So if we ever get the type wrong, the worst that happens is a document lands on your desk — never that a wrong number goes quietly through.
You'll get some. That's the design, not a fault. What matters is that it's quick, and that you're never left hunting for what's wrong.
The page sits on one side and the numbers on the other, with the problem already pointed out — “these lines add to 1,204.50, but the subtotal says 1,240.50”. You're not reading the invoice to find the error. You're looking at the error.
Change it, tick it off, done. And if you want it checked again after your fix, that costs nothing — re-checking the sums doesn't involve reading the document again.
If it's something bigger — the whole thing was read badly, or it's not the kind of document we thought it was — you can say so and have it read again properly.
Tick it off and anything you'd set up to happen next carries on by itself, from exactly where it stopped. How that works →
Most software that “learns from you” quietly starts trusting itself. This doesn't. A remembered fix still has to survive the same sums as everything else.
Correct the same thing twice for a supplier and it stops happening. The kinds of mistakes worth remembering are the structural ones — a value that keeps landing in the wrong place, or a supplier whose invoices have the buyer and seller the wrong way round.
This is the important part. A remembered fix is applied before the sums are done, not after. So the maths independently confirms every fix it's learned, every single time.
A new fix is applied but you still see the document. Only after it's been proved right on three separate documents does it stop bothering you about that one.
Turn it off and documents still arrive and sit safely, costing nothing, until you say go. Turn it back on and it can work through the backlog. You always see what each one cost — per document and per client.
If a document looks doubtful — a rough scan, or nothing on the page to check against — we can put a second, more careful reader over it and compare the two. Where they disagree is exactly where you should look.
Got paperwork nobody else has? Set it up yourself — in plain language, picking from your own sample document. No code, and you can try it on a real file before you keep it.
Look after a dozen businesses from one screen — each with its own address, its own documents and its own filing. What one client's paperwork teaches us is never shared with another.
One screen showing what needs you and what doesn't, filtered how you like. You can choose which details show in the list for each kind of document — so you're looking at the numbers that matter to you, not ours.
Every login, every correction, every document that moved — with a name and a time against it. So the question “who changed this, and when?” always has an answer.
Honestly, that's the fastest way to judge this. Pick a document you already know the right answer to, and we'll show you what comes back — and what it flags. No commitment, no setup.