Plain-language automation guide

What is exception-based processing?

Instead of paying employees to search every record for problems, use a tool to perform the repetitive checking and direct employee attention to the records that actually need judgment.

Exception-based processing means: let the tool check everything that follows known rules, then send only the records that do not agree to a person.

Why checking everything manually wastes employee time

Many office processes treat every record as if it requires the same attention. An employee checks hundreds of invoice lines, orders, receipts, rates, or transactions even when most are correct. The employee's valuable work begins only when something does not agree.

A simple example

100 RECORDS
↓
TOOL CHECKS ALL 100
↓
96 MATCH → NO MANUAL SEARCH
4 EXCEPTIONS → EMPLOYEE REVIEWS

The exact numbers will vary. The principle is that employee attention is directed toward the smaller group of records requiring investigation or judgment.

What counts as an exception?

An exception is a record that does not satisfy a defined business rule or cannot be matched confidently. Examples include a price difference, missing receipt, quantity mismatch, invoice above an expected rate, duplicate record, missing order, conflicting identifier, or incomplete document.

Exception-based processing is not automatic approval

This distinction matters. A tool can identify records that satisfy matching rules without being given authority to approve payment or make accounting decisions. Likewise, every discrepancy can remain visible until a person reviews and corrects it.

Automate the search. Preserve the decision.

Where this approach works

ProcessRoutine checking the tool can performException left to a person
Invoice auditingCompare published or expected charge with actual chargeCharge above expected amount
Three-way matchingCompare PO, receipt, and invoicePrice, quantity, receipt, or document difference
ERP / ecommerce reconciliationMatch transactions across systemsMissing, duplicate, timing, or dollar difference
Recurring reportingApply the same calculations and data checksMissing or unusual data requiring investigation

Real example: invoice auditing

In the UPS invoice-checking workflow, the invoice provides both a published shipment rate and the rate actually charged. Instead of an employee searching through the entire invoice for differences, the tool identifies charges above the published rate. The employee can then verify dimensions and determine whether the difference is accurate or an overcharge.

The manual review took about four 8-hour workdays each month. The tool-assisted process takes approximately 30 minutes. The employee's judgment remains; the repetitive search is reduced.

See the invoice-checking case study →

Real example: reconciliation

The same principle applies to purchase orders, receipts, invoices, and ERP/ecommerce records. Matching records do not need the same attention as discrepancies. The tool performs the comparison and makes differences visible for review.

See how PO, receipt and invoice matching works →

When exception-based processing is a good candidate

  • The business checks a large number of similar records.
  • Most records usually follow known rules.
  • Employees spend substantial time finding the smaller number that do not.
  • The matching or checking rules can be documented.
  • Exceptions still need employee judgment.
  • The business wants to preserve review and approval controls.

What needs to be defined before building a tool?

The business needs to explain what a normal record looks like, what conditions make a record an exception, what information the reviewer needs to resolve it, and what should happen after correction. If those rules cannot yet be explained, the process may need to be clarified before it is automated.

Questions businesses commonly ask

Does exception-based processing use AI?

It can, but it does not have to. Many valuable workflows use straightforward matching, calculations, tolerances, and business rules. The important question is whether the repetitive checking can be performed reliably.

What happens to records that match?

That depends on the business process. They may proceed to the next controlled step, be included in an output file, or simply require less manual review.

What happens to an exception?

It stays visible for an employee to investigate, correct, approve, or otherwise resolve according to the company's normal controls.