Read & Process
Clients send documents in every format imaginable. Scanned PDFs. Phone photos of receipts. Contracts in Word, invoices in Excel, payroll reports from three different providers. Archie reads all of it, and actually understands what he's reading.
Book a working session
More Than Reading Text
Every skill on the platform does a narrow job properly rather than everything vaguely. This is where this one earns its place.
Turn a stack into a file
Reads whatever arrives — PDFs, photographs, forwarded emails — and puts it where the cycle expects it, named the way your firm names things.
Find what is missing
Knows what a complete file looks like for this client and this cycle, and chases the gap before the deadline makes it urgent.
Keep the source attached
Every figure it lifts stays linked to the page it came from, so review is a click rather than a hunt.
What Read & Process Handles
Ingest
Arrives by upload, email, integration or a shared folder you already use. Nothing needs renaming first.
Classify
Works out what the document is and which client and period it belongs to, using your naming conventions rather than generic ones.
Extract
Lifts the figures and terms the cycle needs, with a confidence flag on anything it would want a human to glance at.
Reconcile
Checks what it read against the ledger and last period, and raises the differences rather than smoothing them over.
Hand on
Passes a structured, sourced result to the next step of the WorkStream — or to you, if that is where the gate is.
How Archie Describes It
A reader that guesses confidently is worse than one that stops. Anything below your confidence threshold is flagged, not filled in — and the threshold is yours to set.
Where Firms Use Read & Process
The reasoning is visible while it runs, not summarized afterwards — which is how people come to trust it.
“It reads the scanned ones. That sounds small until you remember half our clients still photograph things.”
Where This Fits in Archie's Workflow
What formats can it handle?
PDFs, images, spreadsheets, email bodies and attachments. Poor-quality scans are read and flagged rather than refused.
Does it work in other languages?
Yes, with the same citation behavior. The output is written in your workspace language.
Where do documents go?
They stay in your region and are never used to train a shared model. Retention follows your setting.
Can I correct what it read?
Correct it in the thread. The correction is held against the cycle, so the next client inherits it.
See Read & Process on Real Client Documents
Bring the client whose paperwork is a nightmare. That is a far better test than a clean demo set.
Book a working session