The Complete AP Workflow: From Invoice Receipt to Payment

Every purchase invoice follows the same journey through your organization. It arrives, gets read, coded, matched, reviewed, approved, and paid. Whether you process 500 invoices per month or 5,000, the workflow is fundamentally the same. What changes is the cost of doing it badly at scale.

This guide maps the complete accounts payable workflow, identifies where things typically go wrong, and shows how AI changes each step.

The 7 Stages of the AP Workflow

┌─────────────────────────────────────────────────────────┐
│                    INVOICE ARRIVES                        │
│            (email, portal, EDI, paper)                   │
└───────────────────────┬─────────────────────────────────┘
                        ▼
              ┌─────────────────┐
              │   1. INTAKE     │  Receive and register
              └────────┬────────┘
                       ▼
              ┌─────────────────┐
              │  2. EXTRACTION  │  Read header + line items
              └────────┬────────┘
                       ▼
              ┌─────────────────┐
              │   3. CODING     │  Assign GL, cost center, entity
              └────────┬────────┘
                       ▼
              ┌─────────────────┐
              │  4. MATCHING    │  PO ↔ GR ↔ Invoice
              └────────┬────────┘
                       ▼
              ┌──────────────────────┐
              │ 5. EXCEPTION HANDLING│  Resolve mismatches
              └────────┬─────────────┘
                       ▼
              ┌─────────────────┐
              │  6. APPROVAL    │  Authorize for payment
              └────────┬────────┘
                       ▼
              ┌─────────────────┐
              │ 7. PAYMENT PREP │  Batch and release
              └────────┬────────┘
                       ▼
┌─────────────────────────────────────────────────────────┐
│                  PAYMENT EXECUTED                         │
│              (bank transfer / batch)                     │
└─────────────────────────────────────────────────────────┘

Stage 1: Intake

What happens: Purchase invoices arrive from suppliers through various channels. Email (PDF attachments), supplier portals, EDI feeds, e-invoicing (Peppol/UBL), and occasionally still paper.

Common pitfalls:

• Invoices sent to personal email addresses instead of a central AP mailbox
• Portal invoices only checked weekly, creating artificial delays
• Paper invoices scanned but not registered until someone gets to them
• Duplicate invoices arriving through multiple channels (email and portal)

What good looks like: A single intake point that captures invoices from all channels within minutes of arrival. Every invoice gets a timestamp, a unique reference, and enters the processing queue immediately.

Volume context: A mid-market company processing 8,000 invoices per year receives roughly 30-35 invoices per business day. At that volume, even a one-day intake delay means 30+ invoices sitting unprocessed.

Stage 2: Extraction

What happens: Someone (or something) reads the invoice and captures the key data: supplier name, invoice number, date, due date, PO reference, line items, quantities, unit prices, totals, and tax amounts.

Common pitfalls:

• Manual data entry errors: transposed digits, wrong PO numbers, incorrect amounts
• Inconsistent supplier names (is it "Van der Berg B.V." or "VD Berg" or "Berg, van der"?)
• Line item extraction skipped entirely, only header data captured
• Free-text PO references missed ("ref: your order 2024-1234")

What good looks like: All header and line item data extracted accurately, with confidence indicators for uncertain fields. Supplier automatically matched to the master record. PO references identified even from unstructured text.

Error impact: One transposed digit in an invoice amount means a matching failure downstream. One wrong PO reference means the entire matching step fails. Extraction errors cascade through every subsequent stage.

Stage 3: Coding

What happens: Each invoice (or invoice line) gets assigned to the correct general ledger account, cost center, project code, and entity. This determines where the cost lands in your financial reporting.

Common pitfalls:

• Coding knowledge lives in one person's head. When they are out, others guess.
• Inconsistent coding across team members. Same supplier, same product, different GL accounts.
• Cost center assignments based on "what we always did" rather than current organizational structure
• No validation against the chart of accounts, leading to postings to inactive or wrong accounts

What good looks like: Coding follows documented rules that are applied consistently. Historical patterns are visible and reusable. New invoice types get flagged for human decision, and that decision is remembered for next time.

Scale impact: At 500 invoices per month, coding inconsistencies create reporting noise. At 2,000+ invoices per month, they create material misstatements in management reporting and cost allocations.

Stage 4: Matching

What happens: For PO-based invoices, the system compares the invoice against the purchase order and the goods receipt. This is the 3-way match: did we order this (PO), did we receive it (GR), and does the invoice reflect what we received?

Common pitfalls:

• Binary matching: either it is a perfect match or it fails completely. No tolerance for minor variances.
• Goods receipts not posted on time, causing invoices to fail matching even though the goods arrived
• Multi-line invoices where some lines match and others do not, but the entire invoice gets blocked
• Partial deliveries that require cumulative matching across multiple goods receipts

What good looks like: Matching with configurable tolerance rules (e.g., 2% price variance auto-approved). Partial and cumulative matching supported. Missing goods receipts trigger a retry cycle rather than a dead stop.

Time impact: Matching is typically the most time-consuming step in AP. For a company processing 10,000 PO-based invoices per year, matching consumes 40-60% of total AP processing time.

Stage 5: Exception Handling

What happens: When matching fails or something looks wrong, the invoice becomes an exception. Someone investigates: Why did it not match? Is the price different? Is the quantity wrong? Is the PO missing?

Common pitfalls:

• Exceptions pile up in a shared queue with no prioritization
• Investigation requires manually checking multiple systems (ERP, email, supplier portal)
• Resolution depends on responses from purchasing or warehouse teams, creating email chains that take days
• No visibility into how many exceptions are open or how old they are

What good looks like: Exceptions are categorized by type (price variance, quantity mismatch, missing GR, missing PO, duplicate). Each exception is routed to the person who can actually resolve it, with all relevant context attached. Auto-resolvable exceptions (like a missing GR that gets posted the next day) clear themselves.

Bottleneck reality: In most AP departments, 60-70% of total processing time is spent on the 15-25% of invoices that become exceptions.

Stage 6: Approval

What happens: Invoices that pass matching (or have exceptions resolved) go through an approval workflow. Budget holders or managers authorize the payment based on amount thresholds, cost center ownership, or other business rules.

Common pitfalls:

• Invoices routed to the wrong approver, requiring re-routing and delays
• No escalation rules: invoices sit with absent approvers for weeks
• Threshold confusion: approver does not know their limit, or limits are outdated
• Approval treated as a formality rather than a control, leading to rubber-stamping

What good looks like: Clear approval matrix based on amount, cost center, and entity. Automatic escalation after defined timeframes. Delegation rules for absences. Low-risk, fully matched invoices auto-approved within policy.

Delay impact: Approval is the second-biggest bottleneck after matching. Average approval cycle time in companies without automation is 5-8 business days. That is 5-8 days added to every invoice's processing time.

Stage 7: Payment Preparation

What happens: Approved invoices are grouped into payment batches. Due dates are checked, early payment discounts are evaluated, and the batch is sent to the bank for execution.

Common pitfalls:

• Invoices approved but not moved to "ready for payment," creating a gap between approval and actual payment
• Early payment discounts missed because invoices were not processed fast enough
• Duplicate payments when the same invoice exists in multiple systems or was processed twice
• Manual batch creation that happens weekly instead of dynamically based on due dates

What good looks like: Approved invoices automatically move to payment-ready status. Batches are created based on payment terms and due dates. Duplicate detection prevents double payments. Early payment discounts are captured when available.

How AI Changes Each Stage

The workflow above does not change when you introduce AI. The seven steps remain the same. What changes is who (or what) performs each step.

The difference is not incremental. AI turns a process that requires 15-30 minutes per invoice into one that takes seconds for the majority of invoices. The remaining portion still needs human judgment, but the human gets all the information they need in one screen rather than hunting through three systems.

Where to Start

If you are evaluating your AP workflow, start by measuring where the time actually goes. Most organizations underestimate how much time matching and exception handling consume relative to the other steps.

A quick assessment:

1. Count your exceptions. What percentage of invoices require manual intervention after extraction?
2. Measure matching time. How long does it take to match a complex, multi-line PO invoice?
3. Track approval delays. What is your average time from "ready for approval" to "approved"?
4. Identify your key-person risks. If your most experienced AP clerk is out for two weeks, what breaks?

These four data points tell you where AI will have the biggest impact on your specific workflow.

Want to map your AP workflow against what AI can handle today? Get a free workflow assessment -- bring your invoice volumes and we will identify which steps are ready for automation.