Blueprint
Document Extract and Load
The right method for every document
Rules where they work, AI where it helps
Take the reading, interpreting and rekeying out of your operation. Not every document needs AI. Structured files such as spreadsheets and statements are read by rules, the same way every time. Scans, images and PDFs are read with OCR and AI models chosen for the job, with every value checked for confidence.
1
Receive
Documents arrive by email, post, portal upload, spreadsheet or from other systems. Each one is logged against whoever sent it and routed to the right method of extraction.
2
Extract
Structured files are read by import profiles, using rules rather than AI. Scans, images and PDFs go through OCR to capture the text, then AI models identify the document and the values that matter, each with a confidence score.
3
Map
Extracted values are normalised and mapped onto your own fields and the right record. Anything below your confidence threshold goes to Triage for a person to check first.
4
Act
Once the data is saved, Workflows take over: running checks, updating records, sending data to other systems and keeping everyone informed.
Read by rules. No AI needed
Spreadsheets, CSV files and statements already have their data in rows and columns. An import profile maps each column to the right field and applies your validation rules, so the same file gives the same result every time.
Read by OCR and AI, checked by people
Where the data is locked in a page, OCR captures the text and layout, and AI models work out what the document is and which values matter. Every value carries a confidence score, and anything uncertain goes to a person.
A working version of every stage
Every document is received, its data extracted by the method that suits it, mapped onto your fields and put to work. The Blueprint gives you a working version of each stage, linked by Workflows. You decide which documents, fields and rules it uses.
Linked into one process
We provide the tools. You decide how they fit together, and Workflows link them into one process.
Add the steps you need
Need another step? Add an approval, a duplicate check, a translation or a hand-off to another team. The process is yours to shape, and you can change it later without a development project.
What you can do
We provide the tools. You build the process
Every tool below already works with the others on one platform. Pick the ones your documents need, and Workflows link them into one process, with every document traceable from arrival to the data it produced
Take documents in from anywhere
Email attachments, scanned post, portal uploads, spreadsheets and files from other systems all arrive in one place, matched to the person or company that sent them.
Structured files, read by rules
Spreadsheets, CSV files and statements go through reusable import profiles that map each column to the right field and check it against your rules. No AI involved, and the same result every time.
Scans and PDFs, read by OCR and AI
OCR turns scans, images and PDFs into text and layout. AI models then identify the document type and pull out the values you need, translating them where required.
The right model for each document
Extraction models are configured through Integrations, so each document type can use the approach that reads it best. We help you test accuracy before you go live, and you can bring your own AI account.
Confidence on every value
Every value extracted by OCR and AI carries a confidence score. Set the threshold, and anything below it goes to Triage with its values pre-filled, so your team checks rather than types.
Land data in the right place
Extracted values are normalised and mapped onto your own fields, in a consistent structure from the first document to the thousandth, and attached to the right record.
Put the data to work
The moment data is saved, workflows take over: updating records, running checks, generating responses, requesting anything missing and sending data on to the other systems you use.
See volumes and accuracy
Track how many documents arrive, how quickly they're processed, how many needed a person to check, and where work is waiting.
Keep AI costs under control
See exactly what your AI usage costs, with a clear breakdown by document type, process and model. Spot where spend is growing and why, so AI costs never run away from you.
See it in action
Processing foreign-language invoices
Here's how a financial services organisation processes high volumes of German supplier invoices on Volume, from the email arriving to the data reaching their accounts. Every step is automated by Workflows, using the tools and integrations shown
1
Receive
Invoices arrive
Supplier invoices arrive by email as PDFs, in German, in a different layout from every supplier.
2
Extract
Text captured by OCR
OCR reads each PDF, capturing the text and where it sits on the page, whatever the supplier's layout.
3
Extract
Values identified and translated
AI models identify the invoice number, dates, supplier details, line items, VAT and totals, translate them into English and score each value for confidence.
4
5
Map
Checked where needed
Any invoice with a value below the confidence threshold goes to Triage with every value pre-filled, so your team only confirms.
6
Act
Checks run automatically
A workflow checks for duplicate invoices and that the line items add up to the total, and routes anything unusual to your team.
7
Act
Sent on for payment
The invoice data is passed to your accounting platform, ready for payment.
8
Act
Tracked in dashboards
Volumes, processing times and anything waiting for a check are visible in a live dashboard.
FAQ
Common questions
Do you use AI for everything?
No. Structured files such as spreadsheets, CSV files and statements are read by import profiles using rules, because that's faster, cheaper and gives the same result every time. OCR and AI are used where the data is locked in a page, such as scans, images and PDFs.
How do you decide which method to use?
By the type of document. We look at what you receive, test the options against real examples and configure the method that reads each type most accurately, then measure it once it's live.
What kinds of document can it read?
Spreadsheets, CSV files and statements, as well as invoices, forms, letters and identity documents that arrive as PDFs, scanned post or images.
Do documents need to follow a set layout?
Structured files are mapped once with an import profile, then reused. Scans and PDFs don't need a template, so you don't need one for every supplier, lender or customer that sends you a document.
How accurate is it?
Rule-based imports either match your rules or are rejected with a reason. For OCR and AI, every value carries a confidence score, and anything below your threshold goes to a person to check before it's saved.
How do we keep AI costs under control?
Every document read by AI is tracked, with a clear breakdown of usage and cost by document type, process and model. Structured files use rules rather than AI, so you only pay for AI where it's needed.
Can we use our own AI account?
Yes. Extraction models are configured through Integrations, and you can bring your own AI account if you'd rather use one you already have.
Can it handle documents in other languages?
Yes. Documents can be translated as they're read, so the data reaches your team in English.
Can we choose which fields are extracted?
Yes. For each type of document, you decide which fields to capture and where they're saved. Add or change fields at any time through configuration.
How long does it take to set up?
Because you start from a working Blueprint rather than a blank page, most extraction processes go live in weeks, not months. We'll agree a timeline with you after discovery.
How is our data protected?
Data is hosted in the UK, held in your own encrypted database and protected down to individual fields. Volume holds Cyber Essentials and Cyber Essentials Plus certification and is committed to ISO 9001 & 27001.
Proven in practice
Running for real customers
Case Study – Legal Services
One platform in place of a costly Salesforce stack
Alawco moved from a costly Salesforce setup to one Volume platform, with incoming documents read and loaded alongside its casework, customer engagement and finance.
Read the Case Study
Case Study – Litigation Financing
Evidence to settlement for German gambling actions
Munster Holdings stopped reading, translating and rekeying German invoices by hand. Volume now extracts and translates each one as it arrives, as structured English data ready to match to its claim.
Read the Case StudyPrevious
