The paperwork reads itself, and lands where it belongs.
Supplier invoices, purchase orders, enquiry forms and customer emails all hold details someone retypes. AI can read them, check them against what you already know and file them, with a person checking anything it isn't sure of.
From accounts@supplier.example
INVOICE 30418 To: Skerrin Ltd PO ref: 4471 40 x Fell Rolltop, olive @ £31.50 24 x Field Tote, sand @ £14.20 Delivery £45.00 Total inc. VAT £1,976.88 Due 30 days
Into: Accounts and stock
Nothing filed yet. Press Read with AI.
Fictional documents. Details that don't match your records are flagged for a person, never filed.
Who this is for
The documents differ between a shop and a trade. The job is the same: read it, pull out what matters, put it in the right place, and flag what looks wrong.
Online shops
Supplier paperwork and product data, from inbox to store and accounts.
- Supplier invoices matched to purchase orders and posted to your accounts
- Product descriptions, attributes and categories drafted from supplier sheets
- Wholesale orders by email turned into orders in your system
- Customer emails sorted by topic and urgency
Trades and local services
Enquiries, job notes and receipts, without the evening of typing.
- Every enquiry summarised with the job type, area and urgency
- Voice notes turned into job sheets and invoice drafts
- Receipts and merchant invoices read and added to the job
- Addresses and contact details tidied and checked
What we actually do
- Find what's retyped
- We list every document and message that someone copies from by hand, how often it arrives and how long it takes.
- Decide what it may do alone
- Some details are safe to file directly. Others, such as prices or bank details, always go to a person. We agree which is which first.
- Check against what you know
- Extracted details are checked against your own records: a supplier's usual prices, a known customer, a real postcode. Anything that doesn't match is flagged.
- Fill the gaps
- Where details are missing, it drafts them for you to approve: a product description from a spec sheet, a category, a corrected address.
- Count the accuracy
- We measure how often it was right and how often a person changed something, and share the figures with you each month.
How an engagement runs
Collect examples
A few weeks of real documents and messages, so we build on what you actually receive.
Build and test
We set up reading and filing for one document type and test it against those examples before anything goes live.
Run alongside
It runs next to the manual process for a short period, with a person checking every result.
Hand over
Once it is reliable it runs on its own, with uncertain cases still going to a person.
What you get
- A list of every retyped document, with time spent on each
- Automated reading and filing, one document type at a time
- Checks against your own records, with flags for anything odd
- A review queue for the cases that need a person
- Monthly accuracy figures
- A register of every automation and what to do if it fails
The /how-we-work page shows this working on enquiry forms: the AI brief, the routing and the checks run live on our own contact form.
How the return works
The saving is time you can count before and after. The bigger win is fewer mistakes from retyping.
Say a shop receives 60 supplier invoices a month and each takes eight minutes to check and enter. That is eight hours a month. If nine in ten are filed automatically, about seven hours come back.
Say a trade writes up 30 jobs a week and each write-up takes ten minutes in the evening. That is five hours a week. Voice notes turned into drafts can cut each one to a quick check.
Worked examples with assumed figures, for illustration. We measure your real volumes and times before building anything.
Questions we get asked
How accurate is it?
It depends on the documents, which is why we test on your own examples first and share the figures. Anything below the confidence level you set goes to a person instead of being filed.
What about handwritten or messy documents?
Clear handwriting and photos usually work. Poor scans and unusual layouts are flagged for a person rather than guessed.
Where does our data go?
It stays in your own systems. The AI step uses providers whose business terms say they don't train on your data, and only the document being read is sent.
Do we need new software?
Usually not. It reads from your inbox or shared folder and writes into the store, accounts or job system you already use.
Can it handle documents in other languages?
Yes, for most common languages. It can also translate the details it pulls out.
Send us the document your team retypes most.
A sample invoice, order or enquiry with personal details removed is enough. We'll show you what AI would pull out of it.
Join the waitlist