[ Case study ]Document processing

    Intelligent document processing: one pipeline, any industry

    Documents from every channel read, identified, extracted and filed as structured records. Proven across financial services, wealth management and healthcare, with a person confirming anything uncertain.

    • Any industry
    • OCR + LLM extraction
    • Human in the loop

    Skilled people spend their days opening attachments

    The problem

    Every business that handles documents has the same job hiding in plain sight. Someone opens an attachment, works out what it is and whose it is, retypes what is on it into a system, and files it where they hope the next person will look. It is done by people qualified to do far more valuable work, and it peaks exactly when there is least capacity to spare.

    The documents are never tidy. They arrive locked behind passwords, photographed on phones, bundled together, and in dozens of near-identical formats. A single missed or misfiled document can hold up a whole client, claim or case.

    Ingest anywhere, normalise once

    How it solves it

    Documents arrive by whatever channel suits the sender. The pipeline unlocks and reads each one, identifies it against a catalogue of document types the business owns, pulls out the fields that type defines, and writes the result into the systems already in use. Anything it is unsure about becomes a visible "needs review" item for a person.

    Arrives by

    Email attachments
    Chat uploads
    Watched folder
    Client portal

    One pipeline

    Unlock protected files
    Read the pages
    Identify the type
    Pull the fields

    Lands as

    Filed against the right case
    Checklist item ticked
    Figures feed reporting
    Unsure? Needs review

    Three industries, one pipeline

    Proof in production

    Financial services and tax

    High volume, every channel, a deadline

    Income certificatesTax assessmentsReturnsSupporting schedules

    Documents arrive by email, chat, a watched folder and a client portal. Protected files are unlocked, read, classified against the business's own list of types and filed against the right case, with the checklist ticked off. Income certificates return dozens of fields, and the extracted figures feed revenue data and client segmentation.

    Anything the system cannot place becomes a "needs review" item and the consultant is notified.

    Wealth and advisory

    Records that stay current

    Proof of IDProof of addressProof of bankingTax compliance certificates

    An assistant sits beside the client record. A document is dropped in, the type is proposed with its reasoning, and the values read are shown next to what the record holds. Confirmed values update the CRM, and the file is stored under a consistent name, tags and folder.

    Nothing is saved until a person confirms, and a value that disagrees with the record is shown side by side, never overwritten quietly.

    Healthcare

    Sensitive records, mandatory checking

    Scan and imaging reportsReferral lettersClinical records

    Reports and letters are read into structured fields such as scan date, facility and test type on the patient record. Identifiers are redacted before any data leaves the environment.

    100,000+

    documents processed

    99.3%

    accuracy, checked by hand, extracting structured data from tens of thousands of clinical records

    ~$216

    compute for that entire run

    A person validates everything read by OCR before it is relied on.

    Here is the wealth and advisory case up close. A document is read, lined up against what the record already holds, and a person confirms in one screen. Only then is the file stored and the record updated.

    Proof of address

    Recognised with 96% confidence: a municipal account with a residential address and an issue date.

    Awaiting your confirmation
    Residential address 68% confident · Changed, please check

    Already on the record

    12 Oak Avenue

    Read from the document

    14 Oak Avenue

    Date issued 88% confident · New, will be added

    Already on the record

    Not held

    Read from the document

    12 Aug 2026

    Issued by 79% confident · New, will be added

    Already on the record

    Not held

    Read from the document

    City Municipality

    Will be filed as ClientRef_ProofOfAddress_2026-08-12.pdf in the client's Address folder

    The review screen: illustrative

    Configured, not rebuilt

    Any industry

    What changes from one industry to the next is not the engine. It is the catalogue of document types, the fields each one defines, the rules that settle the ambiguous cases, and the system the result lands in. Those are settings held as data, which is why a new industry or a new document type does not mean starting again.

    You configure

    Document types
    Fields to read per type
    Rules and confidence lines

    One pipeline, unchanged

    Ingest from any channel
    Unlock and read
    Classify and extract
    Person confirms the unsure

    Lands in your systems

    CRM or case system
    SharePoint or file store
    Finance or ERP system
    Review queue

    A different folder of paper, the same pipeline

    Where else it applies

    If your team opens it, identifies it and retypes it, it is a candidate. These are the same steps pointed at other industries.

    Finance and retail

    Supplier invoices

    Read, matched to the purchase order and queued for accounts payable, with exceptions flagged.

    Insurance

    Claim bundles

    Split a mixed bundle into its parts, classify each, and pull the claim fields.

    Banking and lending

    KYC and onboarding packs

    Identify each proof document, extract the details and tick the compliance checklist.

    Property and legal

    Leases and contracts

    Abstract dates, parties and obligations into fields the business can search and report on.

    Logistics

    Delivery notes and proof of delivery

    Read the signed paperwork and close the job against the order automatically.

    Any business

    HR onboarding documents

    File identity, qualification and bank documents against the new starter and chase what is missing.

    Public sector and NGOs

    Grant and compliance submissions

    Check each submission against its required documents, extract the figures and flag what is missing.

    Accounting

    Bank statements and receipts

    Categorise transactions and pull the totals into the books, with unusual lines queued for review.

    Your industry

    Whatever your team opens and retypes

    If a person opens it, identifies it and keys it in, it is a candidate. Bring us a sample folder.

    Bring us a folder of your worst documents

    Whatever the industry, we will show you what the pipeline makes of them.