Where Understanding
Leads

Built on deep expertise in payment operations, document structure, and intelligent capture, enabling accurate, scalable, and dependable processing across real-world scenarios.

Explore Capabilities arrow_forward

Envisioning the Future

Explore Solutions →

Purpose-Built Document Models

PowerDox Expertise applies One Intelligence to real operational document types, including checks, invoices, remittance coupons, claims forms, receipts, signatures, and envelopes. Each domain incorporates purpose-built field models, layout interpretation, amount logic, validation rules, and exception signals aligned to payment, lockbox, and enterprise intake workflows.

These documents vary by issuer, format, print quality, and completion style. Fields may be printed, handwritten, optional, or inconsistently placed. PowerDox accounts for this variation at the domain level, delivering consistent, traceable outputs even as document structures shift.

Expertise also means understanding what drives processing. Amounts must reconcile. Identifiers must match. Signatures may gate workflows. Line items must balance. PowerDox embeds these controls directly into extraction and validation so outputs support posting, routing, reconciliation, and review, not just data capture.

Checks

PowerDox One Intelligence for Checks extracts and validates key data from checks used in personal, business, government, and institutional payment workflows. The system interprets checks as structured financial documents, supporting reliable extraction of fields, amounts, MICR data, and non-authoritative signature signals to enable automated posting, review, and exception handling.

PowerDox Check Intelligence extracts core check fields using layout-aware analysis.

Extracted check fields include:

  • Payee name, including printed or handwritten text

  • Account holder name and address, when present

  • Check date, including printed or handwritten formats

  • Check number, extracted from printed locations

  • Bank or financial institution name, when present

  • Memo field content, including printed or handwritten text

Each field is returned with confidence and source evidence, with both raw and normalized values preserved for downstream processing and review.

This approach enables consistent extraction across personal, business, cashier’s, and government check formats without issuer-specific configuration.

PowerDox interprets check amounts as financial values governed by payment rules, not as isolated OCR strings.

The system independently extracts and interprets:

  • Courtesy amount, captured from the numeric amount box

  • Legal amount, captured from the written amount line, including fractional cents notation

Amount interpretation includes:

  • Normalization of punctuation, separators, and currency indicators

  • Interpretation of handwritten legal amounts, including extended dash patterns and partial text

  • Extraction of fractional cents components (for example, xx/100 formats)

  • Preservation of raw amount text alongside normalized numeric representations

After independent interpretation, courtesy and legal amounts are reconciled:

  • Agreement or mismatch is explicitly determined

  • Multiple candidate amounts are ranked when ambiguity exists

  • Mismatches are surfaced as structured outcomes rather than silently overridden

This ensures amount discrepancies are handled deterministically and routed appropriately in downstream workflows.

PowerDox detects and extracts MICR data from checks.

MICR extraction includes:

  • Routing number

  • Account number

  • Check serial number from the MICR line

  • Printed check number, when present elsewhere on the check

  • Raw MICR text, preserved for audit and review

MICR data is normalized to address common OCR ambiguities and evaluated for internal consistency. When both MICR serials and printed check numbers are present, values are compared for consistency.

Optional validation can be applied when required, including routing number checksum validation and basic account number sanity checks.

PowerDox detects and interprets handwritten marks in expected signature regions on checks.

Signature detection includes:

  • Identification of expected signature regions based on check layout

  • Interpretation of handwritten ink strokes consistent with a signature mark

  • Location evidence and confidence scoring indicating the presence of a signature

Signature interpretation is non-authoritative and is not used for identity verification or biometric comparison. It is intended to indicate the presence and placement of a signature-like mark only.

Signature results are returned as structured signals to support workflow gating, exception routing, and audit requirements.

PowerDox applies check-specific validation logic and produces explicit exception signals for downstream decisioning.

Validation and exception handling include:

  • Courtesy and legal amount agreement or mismatch

  • Missing, unreadable, or ambiguous amount fields

  • Missing or invalid routing information when validation is enabled

  • Missing or ambiguous check date

  • Missing or incomplete payee information

  • Missing signature when required by workflow rules

  • Duplicate or conflicting check identifiers within a submission or batch

  • Structural anomalies, such as partial scans or inconsistent front- and back-side submissions

Validation outcomes are returned as structured results with confidence indicators, enabling deterministic routing to straight-through processing, review queues, or exception handling without manual document inspection.

Check Fraud Detection

PowerDox One-Intelligence evaluates checks not only for data extraction, but for fraud risk.  The engine validates structural integrity, reconciles critical fields, verifies MICR check digit values, and detects visual inconsistencies such as ink variation within sensitive regions. These controls produce structured fraud signals and risk indicators as part of the engine response.  When deployed within PowerDox Intelligent Capture, the returned fraud signals can be combined with historical duplicate detection and optional consortium verification services to form a complete risk assessment.

One-Intelligence produces the intelligence. The host application applies the operational decisioning.

One-Intelligence evaluates whether the check follows expected structural composition.

Analysis includes:

• Layout alignment of key regions
• Presence of required elements such as date, payee, and signature
• Proper positioning of amount fields
• Structural consistency of the MICR line

Structural irregularities are identified before transactional validation occurs.

This ensures that the instrument itself meets expected financial document standards.

Critical fields are validated both logically and mathematically.

The system verifies:

• Written amount matches numeric amount
• Printed check number aligns with MICR serial number
• Routing number passes check digit verification
• Required data elements are present

Failures in field reconciliation or checksum validation are flagged immediately.

These controls prevent posting of structurally invalid or internally inconsistent instruments.

The engine evaluates visual consistency within sensitive regions of the check.

This includes analysis of:

• Ink color variation across amount fields
• Differences in text appearance within key regions
• Irregular visual characteristics that diverge from surrounding content

These signals may indicate potential modification and are evaluated alongside structural and field validation controls.

Visual analysis strengthens fraud detection without relying solely on manual inspection.

When One-Intelligence operates within Intelligent Capture, extracted check data can be compared against processing history.

This includes:

• Duplicate serial number detection
• Repeated presentment identification
• Previously processed item comparison

Historical checks add transactional context to document-level signals.

Intelligent Capture can optionally call consortium verification services, such as TrueChecks, using structured check data produced by One-Intelligence.

The consortium response may:

• Confirm instrument validity
• Identify mismatches
• Return risk indicators based on external network data

These responses are incorporated into the overall fraud assessment.

Invoices

PowerDox One Intelligence for Invoices extracts structured data from invoices across a wide range of formats and layouts. The system supports header fields, line items, amounts, and validation signals without relying on rigid vendor-specific templates, enabling consistent downstream reconciliation, matching, and review workflows.

PowerDox Invoice Intelligence extracts core invoice fields using layout-aware analysis.

Extracted invoice fields include:

  • Vendor name and remit-to details, when present

  • Customer bill-to and ship-to information, when present

  • Invoice identifiers, such as invoice number, purchase order number, and reference IDs

  • Invoice dates, including invoice date and due date

  • Payment terms, including discount and net terms

Extracted values can be evaluated against configured patterns and expected formats. When values do not match configured expectations, explicit exceptions are surfaced.

All fields are returned with confidence scoring and source evidence. Both raw extracted text and normalized values are preserved to support auditability, review workflows, and downstream reconciliation.

Vendor-specific templates are not required, and the system uses learning-based methods that continually improves over time, with targeted refinement applied when needed.

PowerDox Invoice Intelligence extracts and normalizes monetary amounts from invoices.

Amount-related extraction includes:

  • Invoice-level amounts, such as subtotal, adjustments, and total amount due

  • Numeric normalization, including handling of currency symbols, separators, and formatting variations

  • Preservation of raw and normalized values for traceability and audit

Amounts are interpreted in relation to other financial components on the invoice. When inconsistencies are detected between extracted values or expected relationships, the system produces explicit validation signals rather than forcing reconciliation or altering detected values.

This ensures downstream systems receive transparent and reliable financial data.

PowerDox extracts invoice line items as structured rows, preserving the logical relationship between item descriptions, quantities, pricing, and extended amounts.

Line item extraction includes:

  • Item or service descriptions, including multi-line entries

  • Quantities and units of measure, when present

  • Unit prices and extended line amounts

  • Part numbers, service codes, or SKUs, when present

  • Line-level discounts or adjustments, when present

The system supports common invoice table layouts and variations, including multi-page tables, abbreviated headers, and uneven column alignment. Extracted line items are returned as structured rows with confidence and source evidence for downstream processing and review.

PowerDox extracts invoice summary totals and related charges.

Totals extraction includes:

  • Subtotal, when present

  • Invoice-level adjustments, such as discounts or credits

  • Additional charges, such as shipping or handling

  • Tax amounts, when present

  • Total amount due

Totals are compared against line item sums and related values when applicable. When discrepancies are detected, explicit validation outcomes are produced.

PowerDox validates extracted invoice data using internal consistency checks, with optional pattern-based rules applied when available.

  • Pattern-based validation of invoice numbers, purchase order numbers, and reference identifiers, when patterns are known

  • Required field checks for mandatory invoice fields

  • Arithmetic checks comparing line item sums to subtotals and totals, when applicable

  • Duplicate detection for invoice identifiers within a submission or processing window

  • Date checks for missing or inconsistent invoice dates

Validation results are returned as explicit signals to support straight-through processing, review, or exception handling.

Remittance and Coupons

PowerDox One Intelligence for Remittances extracts payment-related information from retail remittance coupons and payment stubs commonly used in lockbox and bill payment workflows. The system focuses on OCR line interpretation, payer-completed payment fields, and selection indicators to support accurate payment application and exception handling.

PowerDox Remittance Intelligence processes retail remittance coupons used in lockbox and related payment processing workflows.

These coupons are designed to identify the customer or account associated with a payment and to capture the payment amount submitted by the payer.

Remittance coupons may also include payer-completed selections or annotations, such as change-of-address indicators, enrollment or subscription checkboxes, and other marked responses. PowerDox detects and captures these selections and associated handwritten input (often referred to as mark sense or MarkSense detection) as structured signals for downstream processing.

Extracted fields include:

  • Customer or account number

  • Invoice or billing reference, when present

  • Coupon or stub identifier, when present

  • Customer or account name, when present

Fields are extracted from printed content and returned with confidence and source location to support downstream posting and review.

Vendor-specific templates are not required, and the system uses learning-based methods that improve over time, with targeted refinement applied when needed.

Retail remittance coupons typically include a payment amount box completed by the payer.

PowerDox extracts and normalizes payment amounts from:

  • Printed or handwritten payment amount boxes on the coupon

  • Amount values encoded in the OCR line, when present

Amount extraction focuses on capturing the payer-submitted value accurately, preserving both the raw captured value and a normalized numeric representation.

The OCR line is the primary machine-readable signal on retail remittance coupons.

PowerDox detects and reads OCR lines printed in OCR-A or OCR-B fonts and extracts encoded values such as:

  • Customer or account number

  • Invoice or billing reference

  • Payment amount, when encoded

  • Check digit values (CDV), when present

When CDV values are present, they are used to validate the integrity of extracted OCR fields.

OCR line values are treated as authoritative for coupon identification and posting, and are correlated with printed fields when both are present.

PowerDox applies remittance-specific validation to ensure coupon data is suitable for automated posting.

Validation includes:

  • OCR line integrity checks, including CDV validation when applicable

  • Consistency checks between OCR line values and printed fields

  • Payment amount validation, ensuring captured values are complete and well-formed

  • Missing or unreadable field detection

Validation results are returned as structured signals, enabling straight-through posting or targeted exception handling without manual interpretation.

Claims and Forms

PowerDox One Intelligence Claims and Forms processing extracts structured data from standardized and semi-structured forms used for claims submission, enrollment, requests, and updates. The system supports field extraction, selections, amounts, and validation signals to enable efficient intake, routing, and review.

PowerDox Claims and Forms Intelligence extracts structured fields from standardized and semi-structured forms used for claims submission, enrollment, requests, and updates.

Extracted fields may include:

  • Claim, form, or case identifiers

  • Customer, member, or policy information

  • Provider, submitter, or organization details, when present

  • Dates, such as service dates, submission dates, or effective dates

  • Form-specific identifiers or reference numbers

Fields are extracted from printed and handwritten content and returned with confidence and source evidence for downstream processing and review.

Vendor-specific templates are not required, and the system uses learning-based methods that improve over time, with targeted refinement applied when needed.

PowerDox extracts and normalizes monetary amounts present on claims and forms.

Amount extraction may include:

  • Claimed amounts

  • Requested reimbursements or adjustments

  • Deductibles, copays, or contribution amounts, when present

  • Totals or summary amounts, when explicitly stated

Amounts are normalized into consistent numeric values. When multiple related amounts are present, they are evaluated for internal consistency, with discrepancies surfaced as validation signals.

Claims and forms frequently include payer- or submitter-completed selections, such as checkboxes, indicators, and marked responses.

PowerDox detects and captures:

  • Checkbox and selection fields (for example, yes/no questions or option selections)

  • Enrollment, eligibility, or coverage indicators

  • Change requests, such as change of address or plan updates

  • Marked responses and associated handwritten annotations

Detected selections (often referred to as mark sense or MarkSense detection) are returned as structured signals with position and confidence information.

PowerDox validates extracted claim and form data using internal consistency checks, with optional pattern-based validation applied when available.

Validation includes:

  • Required field checks for mandatory claim or form fields

  • Format and pattern checks, when reference patterns are known

  • Amount consistency checks, when multiple related values are present

  • Duplicate detection, such as repeated claim or form identifiers

  • Date checks for missing or inconsistent dates

Validation results are returned as explicit signals to support straight-through processing, review, or exception handling.

Receipt and Expense Capture

PowerDox One Intelligence for Receipt and Expense Capture extracts key transaction data from receipts and expense documents used in reimbursement, audit, and expense management workflows. The system supports field extraction, itemized amounts, and validation checks to enable consistent downstream reporting and review.

PowerDox Receipt and Expense Capture extracts core fields from receipts and expense documents using layout-aware analysis.

Extracted fields may include:

  • Merchant or vendor name

  • Transaction date

  • Receipt or transaction identifier, when present

  • Location information, when present

  • Payment method indicators, when present

  • Expense category indicators, when present or inferred

Fields are returned with confidence and source evidence for downstream expense reporting, audit, and review workflows.

Vendor-specific templates are not required, and the system uses learning-based methods that improve over time, with targeted refinement applied when needed.

PowerDox extracts and normalizes monetary amounts from receipts and expense documents.

Amount extraction includes:

  • Subtotal, when present

  • Taxes, when present

  • Tips or gratuities, when present

  • Fees or surcharges, when present

  • Total amount paid

Amounts are normalized into consistent numeric values. When multiple related amounts are present, they are evaluated for internal consistency, with discrepancies surfaced as validation signals.

Receipts and expense documents may include itemized purchase detail.

PowerDox extracts line items as structured rows when itemization is present.

Line item extraction includes:

  • Item descriptions

  • Quantities, when present

  • Unit prices, when present

  • Extended line amounts

The system supports common receipt layouts and itemization patterns. Extracted line items are returned as structured rows with confidence and source evidence for downstream processing and review.

PowerDox validates extracted receipt and expense data using internal consistency checks, with optional pattern-based validation applied when available.

Validation includes:

  • Required field checks, such as date and total amount

  • Arithmetic checks, comparing line item sums to totals when applicable

  • Duplicate detection, identifying repeated receipts or transactions

  • Date validation, flagging missing or inconsistent transaction dates

Validation results are returned as explicit signals to support straight-through processing, review, or exception handling.

Signature Detection

PowerDox One Intelligence for Signatures detects and interprets handwritten signature-like marks across supported document types. Signature signals are non-authoritative and are intended to indicate presence and placement only, supporting workflow gating, exception routing, and audit requirements without identity verification.

PowerDox detects and interprets handwritten marks in expected signature regions across documents.

Signature detection includes:

  • Identification of expected signature regions based on document layout

  • Interpretation of handwritten ink strokes consistent with a signature-like mark

  • Differentiation between signature marks and non-signature handwriting, when possible

Detection is focused on identifying the presence and placement of a signature mark, not confirming identity or authorization.

For each detected signature, PowerDox returns structured evidence to support downstream workflows.

Evidence includes:

  • Location data indicating where the signature mark appears on the document

  • Confidence scoring associated with signature presence

  • Source region information to support visual review or audit

When multiple handwritten marks are present, contextual information is used to associate marks with expected signature regions.

PowerDox applies signature-specific validation checks to support workflow decisioning.

Validation includes:

  • Presence checks, confirming whether a signature-like mark is detected

  • Placement checks, evaluating whether the mark appears in an expected region

  • Consistency checks, when multiple signature regions or document sides are present

Signature validation is non-authoritative and does not perform identity verification or biometric comparison. Validation results are returned as structured signals for workflow gating, exception routing, and audit support.

Postal Mail

PowerDox One Intelligence for Postal Mail processing focuses on envelope-level intelligence for inbound mail. The system supports envelope classification, address extraction, detection of postal markings, and validation signals to enable reliable routing, sorting, and intake workflows.

PowerDox Envelope Intelligence classifies envelopes based on physical structure, layout, and visible envelope features.

Classification includes:

  • Envelope type identification, such as windowed, non-windowed, or multi-window envelopes

  • Mail category indicators, such as remittance, correspondence, claims, or notices, inferred from envelope characteristics when possible

  • Detection of special handling envelopes, when identifiable

Classification results are returned as structured signals to support routing, batching, and downstream intake workflows.

PowerDox extracts and structures address information visible on envelopes.

Address data extraction includes:

  • Recipient address, including name, street, city, state, and postal code

  • Return address, when present

  • Window-extracted address content, when visible through envelope windows

Address data is returned with confidence and source evidence to support routing and downstream processing.

Envelopes often contain postal and carrier-applied markings that convey routing or handling information.

PowerDox detects and extracts common envelope markings, including:

  • Postmarks and cancellation marks, when visible

  • Barcodes, such as IMb or routing barcodes, when present

  • Stamps, meter marks, and indicia

  • Handwritten or stamped handling notations, when present

Extracted markings are returned as structured signals for sorting and exception handling.

PowerDox applies envelope-specific validation checks to support reliable mail intake.

Validation includes:

  • Address completeness checks

  • Detection of missing or unreadable recipient information

  • Confidence checks on envelope classification

  • Duplicate envelope indicators, when applicable

Validation results are returned as explicit signals to support routing, review, or exception handling.