// Document automation

Stop retyping invoices

Send the PDFs, the scans and the phone photos. Get back a clean Excel, CSV or database table — line items included, dates in one format, and the doubtful ones flagged instead of guessed.

// The problem

Somebody is doing this by hand right now

Usually the person you can least afford to have doing it. A few hundred invoices is a week of somebody’s life, and the output is a spreadsheet nobody fully trusts anyway.

Why it never gets fixed

Every supplier has their own layout. Templates work until the fifth supplier, and then somebody is back to copying by hand.
Half of it is not even digital. Scans, faxes, photos taken on a phone at an angle in bad light.
The tools stop at the totals. Which is the number you already knew. The line items are what would have answered the question.
Nobody has a spare fortnight. So the backlog grows, and it is still growing next quarter.

// What comes back

Columns you agreed, before the work started

What can be pulled out

Document fields — invoice or receipt number, dates, supplier, customer
Money fields — subtotal, tax, total, currency, discounts, payment status
Line items — description, quantity, unit price, per row
Anything else on the page that appears reliably enough to be worth a column

In the format you actually use

Excel, for a person
CSV, for a system
Google Sheets, for a team
A SQL-ready table, for a database

The columns are agreed first. It arrives in the shape your process already expects, rather than one you have to spend an afternoon reshaping — which is the step that usually eats the time saved.

// How it works

Four steps, and a sample first

You send a sample

Ten or twenty documents, including your worst ones. Not your cleanest — the awkward ones decide whether this works.
We agree the columns and the checks

What you need out, what the output format is, and what should happen when a field is unreadable. In writing, before the batch runs.
The batch is processed and validated

Totals checked against line items, dates normalised, currencies made consistent. Anything doubtful is flagged in the output rather than quietly filled in.
You get the file, plus the short list

The data, and a list of the few documents worth a human glance. That list is the honest part — a tool that returns no exceptions on a folder of phone photos is not being truthful with you.

// Who it is for

Three situations this keeps solving

Accounting and bookkeeping

Reconciliation, expense tracking and month-end that currently runs on manual entry.

Supplier invoices into one table
Receipts for expense claims
Consistent dates and currencies

Most common

A backlog nobody has time for

Years of documents in a folder or a filing cabinet, waiting for a spare fortnight that never arrives.

Bulk processing in one pass
Mixed formats and layouts
Scans and photos, not just PDFs

Reporting and BI

A dashboard that cannot be built because the data is trapped in documents.

Database-ready output
Line-item detail for real analysis
Repeatable on each new batch

Volume and document quality decide the price, which is why there is not one on this page. Current packages are on Upwork, or send a sample and get a straight quote.

Send me your worst documents.

Not your cleanest ones. Tell me roughly how many you have, what they are, and what you need out of them — then you get an honest read on what comes back and what it costs, before you commit to anything.

Send me a message ↓

Prefer to talk? Book a free call ↗ · Or hire me on Upwork ↗ · Typical reply within one business day.

// FAQ

Questions

My documents are photos and bad scans. Does that still work?

Usually, and that is most of the point. A clean digital PDF is easy and almost anything can read it. Phone photos at an angle, faxed copies, scans with a coffee ring on them and fifteen different supplier layouts in the same folder are the actual job. Send a sample of your worst ones before you commit, and you will get an honest answer about what comes back rather than an optimistic one.

Do I get the line items or just the totals?

Line items, if you want them. Totals are the easy part. Product descriptions, quantities and unit prices per row are where the value is, because that is what lets you ask what you actually spent on a category rather than what you paid a supplier. It is also the part most extraction tools quietly skip.

What format does it come back in?

Whatever you are going to use. Excel if a person is opening it, CSV if a system is reading it, Google Sheets if the team lives there, or a SQL-ready table if it is going into a database. The columns are agreed before the work starts, so it arrives in the shape your process already expects instead of one you have to reshape.

How do I know the numbers are right?

Extraction is checked rather than trusted. Totals are validated against the line items, dates are normalised to one format, and anything the process is not confident about is flagged in the output rather than quietly guessed. You get a short list of documents worth a human glance instead of a spreadsheet that looks complete and is not.

Is this a one-off, or can it keep running?

Both. Most people start with a backlog — a folder or a filing cabinet that has been waiting for someone with a spare fortnight. Once the columns and the checks are settled, the same process can run on each new batch, which is usually where it stops being a project and starts being one less job on somebody's week.

What happens to my documents?

They are used for the extraction and nothing else. They are not published, not resold, and not used as examples anywhere. If you want them deleted at the end of the job, say so and they are deleted at the end of the job. Invoices carry supplier terms, bank details and prices, so treating them casually is not an option.