Direct procurement accounts payable (AP) automation is software that processes supplier invoices for goods a business produces, resells, or consumes in core operations. One invoice can map to blanket releases, delivery schedules, advance ship notices (ASNs), and several ship-to locations. Governed AI agents find the right purchasing record, match the invoice, and resolve exceptions through documented standard operating procedures (SOPs).

Key takeaways

  • Suppliers bill for the raw materials and components in a company's products on direct procurement invoices. Each of these invoices often bills a high-value shipment, so direct procurement invoices make up a larger share of AP spend than of AP invoice volume.
  • Traditional AP software built on two-way and three-way matching assumes the matching path is known before the invoice arrives. But direct procurement suppliers often cite a parent PO without the release, or they bill one PO across several shipments and locations. In The State of AP 2026: AI Rising, Ardent Partners reports a nearly 20% industry-average exception rate across all invoice types.
  • Agents that follow governed SOPs resolve exceptions, whereas routing software only passes exceptions to a reviewer. Agents also record each step of the resolution for audit. 
  • A direct procurement matching agent resolves five layers of context before it commits a match. It checks the purchase order (PO) type, the release or schedule, the ASN or packing slip, the ship-to location, and the receipt status.

An invoice for office chairs clears in one pass. AP software routes an invoice for 1,500 drums of industrial solvent to a review queue because the supplier shipped the drums in three batches to three plants, against the third quarterly release of an annual agreement in Oracle or SAP. The difference between those two invoices illustrates why AP teams automate fewer direct procurement invoices. They need AI agents rather than additional matching rules to raise their touchless invoice processing rates.

Why direct procurement AP automation is a fulfillment problem

Direct procurement invoices cover the goods that become the product. Suppliers bill for those goods as they ship them, so AP must match each invoice with the physical fulfillment of the order. These invoices account for some of the largest payments AP makes to suppliers, such as manufacturers, retailers, healthcare systems, distributors, and utilities.   

In indirect procurement, AP usually checks that the supplier delivered a service or that a manager approved the purchase. The matching path is predictable. The system matches the invoice to a PO, checks price and quantity, and routes an exception to a reviewer when the records disagree.

A direct procurement invoice bills for goods that a supplier physically shipped. The buyer often orders the goods against a long-running agreement and releases them in batches. The supplier ships them to several locations and bills across multiple delivery events, and the buyer receives them in partial deliveries. Even when the supplier, PO, and item repeat, AP still faces several plausible match candidates. AP picks the right one by checking the ship-to location, schedule, release, ASN, packing slip, or receipt status.

AP creates errors outside the department when it matches an invoice to the wrong receipt. The enterprise resource planning (ERP) system records incorrect inventory and accrual entries, and the AP team can no longer reconcile the supplier’s statement against the ledger. The suppliers involved are typically strategic; they bill weekly, and the volumes are high. Direct procurement invoices therefore make up a larger share of AP spend than of AP invoice volume.

Procurement and supply chain teams also depend on how quickly AP handles these invoices. When AI agents process direct procurement invoices as they arrive, AP no longer delays supplier payments or supplier changes. Finance can accept invoices from newly selected vendors right away, so supplier onboarding takes less time. Procurement can then add or change sources without waiting for integration projects. When supplier capacity is tight, a buyer that pays on time keeps its allocation and its payment terms. AP also sees exceptions such as short shipments as they happen, so procurement hears about supply problems before month-end close.

In McKinsey’s Supply Chain Risk Pulse 2025, 82% of the 100 supply chain leaders surveyed said new tariffs affect their supply chains. Among respondents facing tariff impacts, 39% are pursuing dual sourcing for components or raw materials, and 33% are developing supplier nearshoring or onshoring plans. Each new source adds a supplier that AP must onboard and pay, and procurement cannot rely on that source until AP has done so. When finance onboards and pays new suppliers quickly, procurement can change sources during a disruption. Fast supplier onboarding in AP is therefore part of supply chain resilience.

Why traditional PO matching breaks down in direct procurement

Traditional two-way and three-way matching rules assume that AP knows the matching path before the invoice arrives. The system expects a valid PO number, an identifiable PO line, a matching receipt, and a tolerance check at the end. In The State of AP 2026: AI Rising, Ardent Partners reports an industry-average invoice exception rate of 19.9%.

Suppliers break these assumptions in several common ways. One supplier references the parent PO and omits the release. Another includes an ASN but no receipt number. A third bills the same PO and item as the previous week, for goods shipped to a different warehouse. Buyers also set up purchases in different structures. A blanket purchase agreement includes multiple releases, and a scheduling agreement sets monthly delivery schedules. In a drop-ship flow, AP relies on end-customer delivery evidence rather than ordinary receiving.

Data extraction is largely solved. Optical character recognition (OCR) and large language models read invoice fields reliably. But an extraction tool can read every field and still not know which purchasing context the invoice belongs to.

AP teams therefore fall back on manual research. A reviewer opens the invoice, searches the ERP, compares ship-to addresses, inspects receiving history, and selects a candidate. AP pays for that research in cycle time. Ardent Partners’ 2026 report puts the industry average time to process a single invoice at 11.6 days.

Five checks a direct procurement PO matching agent runs before it matches an invoice

A direct procurement matching agent starts with the purchasing pattern, then works through the evidence available for that patter. The agent checks five layers of context to decide whether a match is defensible.

1. PO type and purchasing pattern

The first question is what kind of purchase this invoice represents. AP matches each purchase type differently. The types include a standard PO, a blanket release, a scheduling agreement, a multi-shipment PO, a drop-ship order, and a recurring supply flow. Each ERP system records the pattern differently. Oracle structures blanket releases with shipment-level context, and SAP uses scheduling agreements and call-offs. Workday records changes as PO revisions, and procurement networks such as JAGGAER add their own collaboration references. The underlying task is the same in each system. The agent must resolve the correct purchasing context for the invoice.

2. Parent PO, release, schedule, and call-off context

Under a long-running agreement, the buyer releases goods in batches. The supplier’s invoice may cite the parent agreement, the specific release, or both. The agent must read whichever reference is present and locate the invoice within the agreement's release and schedule structure.

3. ASN, packing slip, and delivery evidence

AP often finds the strongest evidence in shipping documents. An ASN or packing slip number links the invoice to one specific delivery event. The agent relies on that link most in drop-ship flows, where the buyer never physically receives the goods.

4. Ship-to, warehouse, and end-customer context

When supplier, PO, and item all repeat, the agent identifies candidates by destination. The agent must also normalize location evidence, because a site recorded as "Newark Plant NJ" on the PO may appear as "Newark, New Jersey" on the invoice.

5. Receipt status, open quantity, and tolerance checks

Suppliers often deliver goods in partial shipments and invoice each shipment separately. The agent must confirm what was received, what quantity remains open on the line, and whether price and quantity fall within tolerance before committing the match.

From static rules to governed SOPs

Finance teams need governance before they let an AI agent act. Analysts expect many agentic AI projects to be scrapped for that reason. In 2025, Gartner predicted that over 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear business value, or inadequate risk controls. For AP leaders, the lesson is that he value of an AI agent in PO matching comes from governed reasoning over a wider evidence set, applied consistently on every invoice.

Finance teams write that governance into an SOP. In a direct procurement matching SOP, then, the agent follows a documented sequence.

  1. Classify the PO type and purchasing pattern.
  2. Select the standard or direct procurement matching path.
  3. Search candidate PO and receipt records.
  4. Weigh references such as ASN, packing slip, schedule, release, ship-to, warehouse, and item.
  5. Apply quantity, price, and open-balance checks.
  6. Commit the best-supported candidate.
  7. Resolve the exception according to the SOP, or route genuine judgment calls to a human with the evidence already assembled.

With the last step in place, the AP specialist no longer starts an exception from a blank screen. The specialist sees what the agent found, what it matched, and where it stopped. The specialist confirms the agent’s work, and the system produces the audit trail as a by-product.

From routing exceptions to resolving them

At minimum, an agentic system should route each exception with the evidence attached. A stronger system resolves the exception. Software that routes exceptions hands the work to another queue. Agents that resolve exceptions finish the work.

When an invoice disagrees with the purchasing record, a governed agent applies the same SOP and judgment criteria a buyer or AP specialist would apply. It then takes the prescribed action and completes the resolution workflow. The exception record still exists, and the agent uses it as an audit trail. The record documents the issue, the action the agent took, and the reasoning behind that action. It’s no longer a task waiting for a human.

Finance leaders should therefore build the business case on more than AP productivity. Exceptions in direct procurement indicate issues within the physical supply chain. AP often sees short shipments, price disagreements, and receiving discrepancies in invoices first. Agents that resolve them quickly keep suppliers paid on time, keep accruals accurate, and give procurement earlier warning of disruption. In Ardent Partners’ 2026 report, 48% of AP leaders named a high percentage of exceptions as a top challenge. The same report finds that AP teams spend 29.4% of staff time on supplier inquiries, on average. In our experience, unresolved exceptions are a common cause of those inquiries.

How AI agents shorten time to value

AI agents shorten time to value because they match each invoice against evidence already in the invoice and the ERP. AP can automate a supplier's invoices without waiting for that supplier to adopt EDI or a portal.

Enterprises have spent the last two decades trying to standardize direct procurement invoices at the supplier end. They deploy electronic data interchange (EDI), supplier portals, and customized ERP workflows, and large buyers still rely on these tools for high-volume strategic suppliers. Buyers need quarters to deploy each one, and they get no benefit until suppliers adopt it. An EDI map or portal rule also rejects any document that arrives in an unexpected format. Ardent Partners' The State of AP 2026: AI Rising finds that AP teams still process only 38.3% of invoices straight through, on average, across all invoice types.

AI agents work in the opposite direction. They use the evidence already present in invoices and ERP records, which reduces the dependency on perfectly formatted supplier data and removes the need to wait for every supplier to standardize. Ardent Partners finds that exception rates fall from 22% to 9% in automated AP environments. Gartner expects 40% of enterprise applications to include task-specific AI agents by the end of 2026, up from less than 5% in 2025.

Variability, more than volume, is the obstacle to automating direct procurement AP. AP must adjust its matching path to how each supplier invoices, how the ERP structures the PO, and how the receiving team records deliveries. With an agentic approach, finance teams write that operational knowledge into a governed SOP. The agents then apply the SOP to every invoice, including invoices from suppliers that never adopted EDI or a portal.

How AppZen's AI Agents address direct procurement complexity today

Our Autonomous AP solution processes PO-backed invoices with governed, evidence-based decision-making. Fortune 500 companies run finance operations on our platform in manufacturing, life sciences, distribution, and other goods-driven industries. We built our Direct Procurement Matching Agent specifically for this complexity. It evaluates the purchasing context before choosing a matching path. The Agent matches standard POs against standard PO and receipt records. For direct procurement invoices, it follows a specialized path and weighs blanket releases, schedules, ASNs, packing slips, ship-to locations, warehouses, drop-ship references, receipt evidence, and open quantity or amount.

We apply the same governed approach to exception resolution. Our Agents evaluate an exception using the SOP and judgment criteria the business user would apply, then take the prescribed action. The Agent keeps the exception record as the audit trail. Every Agent operates within a documented SOP, so finance keeps control of the procedure and the audit trail. The result is governed autonomy. The Agent resolves exceptions where the evidence is sufficient and routes them, with its reasoning attached, where human judgment is required. Teams extend the approach to their own purchasing patterns through our AI Agent Studio.

The bottom line on direct procurement AP automation

AP teams can automate direct procurement invoices once their software understands the full fulfillment record. That record shows how the buyer ordered the goods, how the supplier shipped and billed them, and how the receiving team recorded them. When AP accepts invoices from new suppliers as they arrive, pays on time, and resolves exceptions quickly, the supply chain becomes more resilient and agile. Our Autonomous AP solution includes AI Agents such as the Direct Procurement Matching Agent. The Agent handles blanket releases, scheduling agreements, multi-shipment POs, and drop-ship flows. Finance leaders evaluating direct procurement AP automation should first ask: Does the system resolve purchasing context or does it only read the invoice? 

Ready to explore our Direct Procurement Matching Agent for AppZen Autonomous AP?

Request a demo today →

FAQ

What is direct procurement AP automation?

Direct procurement AP automation is invoice processing automation for the goods and materials a business produces, resells, or consumes in core operations. These invoices depend on purchase orders, receipts, shipments, ASNs, schedules, and inventory locations, so the automation must resolve fulfillment context in addition to reading invoice fields. Indirect procurement AP automation covers the goods and services that support operations, such as software, facilities, and professional services, where matching is simpler. The same agentic approach handles both, and direct procurement is the harder case.

Why is direct procurement PO matching difficult?

One invoice often relates to several purchasing and fulfillment records. The correct match depends on PO type, release, schedule, ASN, packing slip, ship-to location, warehouse, item, quantity, price, receipt status, and open balance, and suppliers reference them inconsistently.

How do AI agents improve PO matching?

AI agents follow a governed SOP that classifies the purchasing pattern, selects the right matching path, weighs the available evidence, and applies tolerance checks. They resolve matches that static rules miss and route genuine ambiguity to reviewers with the evidence already assembled.

How do AI agents handle direct procurement exceptions?

Governed agents evaluate an exception with the same SOP and judgment criteria a buyer or AP specialist would apply, then take the prescribed action. For a short shipment with a return on record, the agent rejects the invoice or flags the disputed quantity, initiates the return-to-vendor workflow, and notifies the supplier with the evidence attached. The exception record documents the issue, the action, and the reasoning. That record gives auditors a complete trail, and AP no longer works the exception from a manual queue.

Do AI agents replace EDI or supplier portals?

No. EDI and supplier portals remain viable, and AI agents reduce the dependency on them rather than replacing them. Under the older direct procurement AP approach, buyers standardized the input. Suppliers sent structured invoices over EDI or flipped the PO into an invoice on a portal. Buyers need time to onboard and configure each supplier on either channel. AI agents reduce dependency on portal adoption, EDI interchange, and perfectly formatted supplier data by using the evidence already present in invoices and ERP records. The agents help most when suppliers send incomplete or inconsistent references, and they let AP onboard new suppliers faster.

What should finance teams look for in direct procurement AP automation?

Look for automation that handles standard POs, blanket releases, scheduling agreements, multi-shipment POs, and drop-ship scenarios; reads ASN and packing slip references; matches against receipts with tolerance checks; and resolves exceptions through documented SOPs that preserve agent reasoning as an audit trail.