The Requisition That Coded Itself.

A multi-brand restaurant group. Thousands of locations. IT procurement flowing through a centralized intake process that was designed for volume but not for accuracy. Requisitions arrived from franchise operators, regional IT managers, and corporate departments — each one requiring GL coding, budget validation, and policy compliance before it could enter the ERP for processing.

The GL coding error rate was the quiet problem. Thirteen percent of requisitions were miscoded at intake — wrong cost center, wrong account, wrong project code. Each miscoded requisition created a downstream correction cycle: identification during reconciliation, research to find the correct coding, journal entry to fix it, re-approval through the authorization chain. The correction cost more than the original processing. And it happened thousands of times a year.

The GL coding errors were not random. They clustered around three patterns: cross-brand requisitions where the submitter used their own brand’s cost center instead of the shared services code, technology refresh projects where the capital-versus-expense distinction was ambiguous, and new vendor onboarding where no historical coding pattern existed to follow.

Budget validation was a separate step performed by a different team, days later. By the time a requisition was confirmed as over-budget, the purchase had often been verbally committed to the vendor. The policy existed. The enforcement happened too late to prevent the commitment.

The governance model encoded the GL coding rules, budget thresholds, and procurement policies as enforceable validation at the point of entry. Cross-brand coding logic, capital-versus-expense classification criteria, vendor onboarding requirements — each rule applied automatically before the requisition entered the system.

The intelligence layer analyzed the statement of work, matched it against historical coding patterns, and proposed the GL coding with a confidence score. Where the coding was unambiguous — a standard hardware refresh under an existing project code — it coded automatically. Where the classification was genuinely unclear, it held the requisition and presented the competing options with their implications.

Budget validation happened simultaneously. Not downstream. Not days later. At the same moment the coding was proposed, the budget impact was calculated and the authorization threshold was applied.

The requisition arrived pre-coded. GL accounts proposed, budget impact calculated, policy compliance verified. The analyst reviewed a recommendation — not a blank form. Where the system was confident, approval was a single click. Where it held for judgment, the competing options were laid out with their financial implications visible.

The correction cycle that had consumed days during monthly close was replaced by a validation that happened in seconds at intake. Errors didn’t enter the system. They were caught at the door.

IT procurement was the first intake point. But every procurement category — facilities, food supply, marketing, construction — passed through the same requisition workflow with the same coding challenges. The validation architecture was category-agnostic. The GL rules changed; the enforcement pattern did not. One intake point solved. The next question: how many more intake points have the same thirteen percent problem hiding in plain sight?