Document processing
Document and email processing, without the retyping
Invoices, purchase orders, delivery notes, timesheets and application forms arrive as PDFs and photographs, and somebody types them into a system. That work can largely be read automatically, with the uncertain ones flagged for a person rather than guessed at.
- Read, checked, then filed
- Uncertain ones flagged
- Audit trail kept
The job this replaces
In most small businesses there is somebody who opens attachments and types what is in them into something else. Supplier invoices into the accounts package. Purchase orders into the job system. Delivery notes checked against what was ordered. Timesheets into payroll. It is quiet, repetitive, and it consumes a surprising number of hours a week.
It is also the work most prone to small errors, because it is boring. A transposed figure on an invoice is not caught by anybody until a reconciliation weeks later, and finding it costs more than typing it did.
Automated extraction reads the document, pulls out the fields that matter, and puts them where they need to go. Where it is confident, it proceeds. Where it is not, it flags the document for a person instead of guessing, which is the design decision that makes the difference between a time-saver and a source of new problems.
The general approach we take to this work is on our page about AI automation.
What this is used for most often
- Supplier invoices read and matched to purchase orders, with mismatches raised rather than posted
- Purchase orders from customers turned into jobs or sales orders without re-entry
- Delivery notes reconciled against what was ordered, so shortages are caught on the day
- Timesheets and job sheets photographed on site and turned into records that payroll can use
- Application and onboarding forms read into the system that needs them
- Emails classified and routed, with the attachment filed against the right customer or job
- A full audit trail: the original document kept and linked to whatever was created from it
The confidence threshold is the whole design
Automated extraction is not perfect and any supplier who tells you otherwise is selling something. Handwriting, poor photographs, unusual layouts and scanned faxes all produce uncertainty, and the question is what the system does when it is unsure.
Ours flags it. A document below the confidence threshold goes to a person with the uncertain fields highlighted, which takes seconds to check and is dramatically faster than typing the whole thing. The threshold is set with you and tuned once there is real data behind it.
- Confident documents processed automatically
- Uncertain fields highlighted for a quick human check
- Nothing silently guessed and posted
Count the documents before deciding
This pays for itself on volume. A business handling forty supplier invoices a month will save less than the build costs. One handling four hundred is a different conversation entirely.
The free review counts what actually arrives and what handling it costs in hours, then gives you a fixed quote and the arithmetic to decide with.
Get in touch
Talk to us about document processing
A few lines is enough. We will tell you straight whether we can help, what it would cost and how quickly we can be there. No hard sell, and no follow-up sequence if you decide against it.
- We reply within one working day
- A straight answer, including when it is no
- Free process review available, before any money changes hands
Would rather just ring? 01623 354250, Mon–Fri, 9am–5:30pm.
Frequently asked questions
If the answer is not here, ask us. You will get a straight one, from someone who does the work.
Mon–Fri, 9am–5:30pm
How accurate is automated document reading?
On clean, consistent documents it is very good. On handwriting, poor photographs and unusual layouts it is less certain, which is exactly why the system is built to flag low-confidence documents for a person rather than posting a guess. Accuracy on its own is the wrong question; what matters is what happens when it is unsure.
What does it cost?
A build cost quoted up front, plus a running cost that scales with volume. It only makes sense above a certain number of documents a month, and we will tell you if you are below that line rather than selling you the project anyway.
Where does the data end up?
In whichever system already needs it: the accounts package, the job system, payroll. The original document is kept and linked to the record created from it, so there is always an audit trail back to the source.
Is our data used to train anything?
We set that up so it is not, and we put the arrangement in writing. For most businesses handling supplier and customer paperwork, that is a requirement rather than a preference.
Can it handle our specific forms?
Usually, and we test that on your actual documents before quoting rather than after. A sample of twenty real examples tells us far more than any specification.