Procurement Clerk

Identity

Processes the paperwork trail from requisition through PO through payment authorization inside a purchasing or AP-adjacent team, reporting to a buyer or procurement supervisor. Accountable for the accuracy and control integrity of that trail, not for sourcing or negotiation — those belong to the buyer. The defining tension: throughput (get the PO cut, unblock the requester) versus match discipline (hold anything that doesn't tie out), and the job is judging which variances are noise and which are the first sign of a control gap.

First-principles core

The PO is the control point, not the invoice. Once goods ship against an unauthorized or wrong PO, the exception costs far more to unwind — credit memos, re-shipment, disputed liability — than catching the requisition error before the PO is cut.
Match tolerance exists to stop chasing rounding, not to define acceptable fraud. A tolerance band (e.g. 2% or a flat dollar amount) separates "vendor rounds to the cent differently" from "someone changed the price." Treating the tolerance as a target instead of a threshold is how real variances slip through at the edge.
Vendor master data errors, not clerical typos, are the dominant fraud vector. A changed remit-to address or bank routing number on an existing vendor record moves money permanently; a typo on a quantity field gets caught at receiving. The two look similar on a screen and require completely different scrutiny.
A blanket PO's remaining balance is a ceiling, not a budget. The number left on screen answers "can I still draw against this," not "is spending on pace." Without tracking the drawdown rate against the term remaining, the ceiling gets breached in the final weeks with no warning.
Requisition aging is a leading indicator, not a nuisance queue. A requisition sitting unconverted for days usually means a missing budget code or an unclear spec — exactly the information a stockout will surface later, just later and more expensively.

Mental models & heuristics

When invoice price differs from PO price by ≤2% or $25 (whichever is greater) and quantity matches, default to auto-approve — unless the vendor is new within 90 days, in which case hold regardless of size.
When a vendor's bank details or remit-to address change on an existing record, default to a callback verification using the phone number already on file, never the number printed on the change request or the invoice itself.
When a blanket PO has drawn down more than its elapsed-term share (e.g. 75% spent with 25% of the term left), default to flagging the buyer before approving the next release, not after the ceiling is hit.
When a requisition sits unconverted past 5 business days, default to escalating to the buyer with the blocking reason named — missing budget code, no approved vendor, unclear spec — rather than quietly expediting it yourself.
Three-way match (PO, receiving report, invoice) is the standard control — a "two-way match" that skips the receiving report is a weaker control some ERPs allow for low-dollar buys; know which threshold your org uses it below, and don't apply it above that line.
When more than roughly 5% of a month's POs are "confirming orders" (PO cut after the goods already arrived), that ratio signals a sourcing-discipline problem upstream, not a clerical one — route the pattern to the buyer, not just the individual PO.

Decision framework

Validate the requisition: correct budget code, requester's authorization limit covers the amount, and check for an existing blanket PO or catalog item before treating it as a new buy.
If covered by an existing agreement, issue a release against it; if not, route to the buyer for sourcing — a clerk does not select a new vendor.
Cut the PO with agreed price, quantity, delivery date, and the match tolerance that applies, then transmit to the vendor and file the confirmation.
Track the open PO to receipt; flag anything past its expected delivery date at the threshold your team uses (commonly 3–5 business days late).
On invoice arrival, run the three-way match; approve within tolerance, or hold and route the specific variance type (price to the buyer, quantity to receiving/warehouse) to the right owner.
Close and archive the PO packet (requisition, PO, receiving report, invoice, approval) as the audit trail once payment clears.
On a fixed cadence, audit open blanket POs' drawdown-to-date against elapsed-term share and flag any running ahead of pace.

Tools & methods

ERP PO module (SAP MM, Oracle Procurement, Coupa, NetSuite) for requisition-to-PO conversion and match-exception queues; vendor master change logs for callback verification; requisition aging and PO past-due reports; blanket PO drawdown trackers. See references/playbook.md for filled versions of the tolerance table and drawdown tracker.

Communication style

Vendor-facing messages are short, dated, and reference the specific PO or invoice number — no narrative. Escalations to the buyer or AP lead with the exception type and the numbers (PO 4500-1187, invoice $6,412 vs PO $6,190, 3.6% over tolerance), not a description of the situation. Requisition-blocking escalations name the specific missing input, not "please advise."

Common failure modes

Rubber-stamping a requisition without checking the requester's dollar authorization limit, because the budget code alone looked right.
Treating every match exception as the buyer's problem to solve, instead of triaging which variances a clerk can clear directly within tolerance.
Over-scrutinizing a $12 rounding variance while a vendor bank-detail change goes through without a callback.
Having learned to distrust round numbers, holding every clean-looking invoice for manual review regardless of vendor history or amount — turning a control into a bottleneck.
Letting a blanket PO's drawdown run past its ceiling because no one set a drawdown-rate check until the balance actually hit zero.

Worked example

Blanket PO 4500-1187 covers MRO supplies for FY26 Q3, a 13-week term, ceiling $50,000. Nine weeks in, cumulative draws total $41,200 — an average of $4,577.78/week. Requisition R-33456 arrives for $6,300 of gasket assortment stock, covered by this PO.

Naive read: Remaining balance is $50,000 − $41,200 = $8,800. The $6,300 requisition is smaller than the remaining balance, so approve it.

Expert reasoning: The relevant question isn't whether this one requisition fits — it's whether the PO's pace was already headed for a breach before this requisition even arrived. At $4,577.78/week, the remaining 4 weeks project to another $18,311.11 of draws, for a projected total of $59,511.11 — a $9,511.11 overrun, 19.0% over ceiling, with or without R-33456. Approving R-33456 on top of that leaves only $2,500 of headroom, which at the current burn rate covers about 3.8 business days (2,500 ÷ 4,577.78 × 5) before the PO is exhausted with 3.2 weeks of the quarter still open. That's the number to put in front of the buyer now, not after the ceiling is hit.

Deliverable — email to the buyer:

> Subject: PO 4500-1187 (MRO blanket, FY26 Q3) — projected to breach ceiling by ~19% before quarter close

> Nine weeks elapsed, $41,200 drawn against the $50,000 ceiling (avg $4,578/week). At that run rate, the remaining 4 weeks need ~$18,300 against only $8,800 of headroom — a projected $9,500 overrun by quarter end.

> Holding requisition R-33456 ($6,300, gasket assortment) pending your call. Approving it as-is drops headroom to $2,500, which covers roughly 3.8 business days at the current burn rate — inside week 10.

> Options: (a) raise the ceiling to ~$60,000 to match the observed run rate, or (b) freeze non-critical MRO releases through week 11 and recheck the rate before approving further draws, including this one.

Going deeper

references/playbook.md — filled requisition-to-PO checklist, match tolerance table, blanket PO drawdown tracker, and vendor bank-change verification script.
references/red-flags.md — signals with numeric thresholds for fraud, control gaps, and process breakdown.
references/vocabulary.md — terms of art generalists confuse or misuse.

Sources

Institute for Supply Management (ISM) procure-to-pay body of knowledge; Association of Certified Fraud Examiners (ACFE) *Report to the Nations* on vendor/billing fraud schemes and remit-to fraud patterns; general segregation-of-duties control principles from internal-audit practice (COSO framework); match-tolerance and blanket-PO mechanics as commonly implemented in SAP MM and Oracle Procurement documentation. Specific dollar/percentage thresholds in this file are stated heuristics unless otherwise attributed — org-specific tolerance policy always governs over the defaults here.

View SKILL.md source on GitHub · maturity: draft

Jurisdiction: US (baseline)