Wherever paperwork is the bottleneck.

Different industries, the same problem: structured data trapped in documents no two senders format the same way. What follows is the documents, the fields teams actually ask for, and the condition every one of these workloads shares.

See it read.

Five of these workloads, each shown the way its paperwork actually arrives. Pick one; every scene uses sample documents.

A fixed supplier list sending the same forms every month is the best case this product has.

Documents

Invoices · Purchase orders · Remittance advice · Credit notes

Why the layouts repeat

Your vendor list barely changes. A few hundred layouts cover years of invoices, and each one is learned the first time it arrives.

docRefinery Engine · Accounts payable

Invoices arriveRecognised by supplierA new supplier, learned onceIllustrative · sample documents

Seven workloads, in detail.

Each entry states why its layouts repeat, because that is the condition this product depends on. Where a credible third-party benchmark exists for the cost of doing it by hand, it is cited; where one does not, none is invented.

Accounts payable

AP clerks and controllers

A fixed supplier list sending the same forms every month is the best case this product has.

Documents

Invoices · Purchase orders · Remittance advice · Credit notes · Supplier statements

Fields teams ask for

Invoice number
Supplier name and tax ID
PO reference
Line items and quantities
Tax
Total due
Payment terms
Bank details

Why the layouts repeat: Your vendor list barely changes. A few hundred layouts cover years of invoices, and each is learned the first time it arrives.
What it costs today: Ardent Partners puts the average cost of processing a single invoice at $9.40, against $2.78 for best-in-class teams, with an average of 9.2 days per invoice and a 22% exception rate against 9% best-in-class.¹ The gap between those columns is almost entirely re-keying and chasing exceptions.

Lending and mortgage

Loan processors and underwriters

Income and asset verification across packets that arrive in the same handful of formats, under a clock the borrower is watching.

Documents

Pay stubs · Bank statements · W-2s and tax forms · Closing packets · Employment verification

Fields teams ask for

Employer
Gross and net pay
Pay period
Account balances
Deposit history
Statement dates
Loan and property identifiers

Why the layouts repeat: Payroll providers and banks issue documents from a small number of templates. The applicant changes; the layout does not.
What it costs today: Onboarding friction is measured in lost applications, not hours: ABBYY’s State of Intelligent Automation report found 90% of companies lose potential customers during digital onboarding, and industry reporting puts financial-service application abandonment near 68%.² Every document a human has to re-read is time on that clock.

KYC and onboarding

Compliance and onboarding analysts

Identity and address evidence, where extracting something that is not there is the expensive failure.

Documents

Identity documents · Proofs of address · Certificates of incorporation · Registry extracts · Beneficial-ownership declarations

Fields teams ask for

Legal name
Document number
Issue and expiry dates
Issuing authority
Registered address
Company registry number
Ownership percentages

Why the layouts repeat: Issuers are institutions, and institutions do not redesign their forms often. A passport from one country is the same layout every time.
Where it goes wrong: This is the workload where a confident wrong answer is worst. A field the document does not contain is marked absent and never asked for, which is the difference between an empty cell a reviewer can see and a plausible value nobody questions.

Insurance claims

FNOL intake and adjusters

First notice of loss, where speed matters and a wrong value costs more than a slow one.

Documents

FNOL forms · Claims packets · Adjuster notes · Repair estimates · Police and incident reports

Fields teams ask for

Policy number
Date and place of loss
Claimant and insured
Loss description
Estimate line items
Estimate total
Coverage references

Why the layouts repeat: Claims arrive on the carrier's own forms and on a repeating set of third-party estimates, so the same layouts come back constantly.
What it costs today: J.D. Power’s 2026 benchmark puts the cycle from FNOL to final payment at an average of 40.7 days, and the NAIC’s 2024 figure for auto property & casualty claims at 14.7 days.³ Intake is the front of that queue.

Logistics and freight

Coordinators and customs brokers

Paperwork that has to reconcile against a shipment record before anything moves.

Documents

Bills of lading · Customs declarations · Manifests · Delivery notes · Certificates of origin

Fields teams ask for

Consignor and consignee
Container and seal numbers
Gross and net weight
HS commodity codes
Incoterms
Port of loading and discharge
Freight charges

Why the layouts repeat: Carriers and forwarders each have their own form and send thousands of them. Learning one carrier covers every shipment they handle.
Where it goes wrong: Every field re-keyed from a bill of lading is a field that can disagree with the shipment record downstream. The reconciliation is the work; the typing is the tax on it.

Healthcare administration

Intake and revenue-cycle teams

Intake and benefits paperwork that has to become a clean record before anyone bills against it.

Documents

Intake forms · Explanations of benefits · Lab reports · Referrals · Prior-authorisation forms

Fields teams ask for

Patient and member ID
Payer and plan
Service dates
Procedure and diagnosis codes
Allowed and billed amounts
Patient responsibility

Why the layouts repeat: Payers issue explanations of benefits from a small set of templates, and a practice deals with the same payers month after month.
What makes it hard: Handwriting, checkboxes and dense tabular remittance sit in the same document. Confidence thresholds are set per field here more often than anywhere else, because a coded field and a free-text note do not deserve the same trust.

Contracts and legal

Legal operations and commercial teams

Obligations and dates buried in agreements nobody re-reads until something goes wrong.

Documents

MSAs · Amendments · Order forms · Schedules and exhibits · NDAs

Fields teams ask for

Parties
Effective and expiry dates
Renewal and notice periods
Payment terms
Liability caps
Governing law
Assignment and change-of-control clauses

Why the layouts repeat: A counterparty's paper is their paper. Once their MSA template is learned, every order form on it follows.
The honest caveat: The honest caveat: a bespoke negotiated agreement is not a repeating layout. Where contracts are one-off, the advantage narrows to governance and confidence — see the fit test below.

Sources

1Ardent Partners, Accounts Payable Metrics that Matter, 2025. Source
2ABBYY, State of Intelligent Automation Report, Q4 2022; abandonment reporting via Fintactix. Source
3J.D. Power 2026 claims benchmark; NAIC 2024 Claims Processing Benchmark. Source

Third-party industry benchmarks, cited as published. They describe the cost of the work today, not docRefinery’s measured performance.

The fit test.

The advantage comes from layouts repeating. Where they do not, we would rather say so before a trial than after one.

ConditionFit

Same senders, same forms, month after monthStrong

A few hundred layouts covering tens of thousands of documentsStrong

Fields your organisation wants governed rather than improvised per callStrong

A queue where a wrong value is worse than a slow oneStrong

A one-off collection where every document is a different layoutPoor

Getting a corpus into a vector store for retrievalPoor

Spreadsheets and presentations as the primary formatPoor

Do not see your document type?

If it has a consistent structure, chances are docRefinery can learn it — and if it is outside the catalogue, we will train a classifier for it at no charge.

Book a demo How the pipeline works