Purchase orders were supposed to bring control to indirect spend. Instead, they’ve created a compliance paradox — one that leaves most invoices stuck in manual review.
The premise is straightforward: if every spend item has a matching PO, every invoice can be automatically matched and approved. No human review required. But in practice, indirect spend doesn’t work like that.
The Problem with PO Coverage
Indirect spend is, by definition, unpredictable. It includes everything from office supplies to professional services, facilities maintenance to IT support. These categories share a common trait: they’re hard to forecast with a precise PO at the time of purchase.
When POs don’t exist — or don’t match precisely — invoices fall out of the automated flow and land back on someone’s desk. In most AP teams, that’s the majority of invoices.
Why Rules-Based Systems Can’t Scale
The standard response to this problem is to write more rules. If invoice X comes from supplier Y and references project Z, code it to cost centre 1234. The first few rules work well. Then the exceptions begin.
New suppliers. Changed invoice formats. Cost centres that have been restructured. Each change creates new exceptions. Each exception creates new rules. Eventually, the rules engine becomes more expensive to maintain than the manual coding it replaced.
The AI Alternative
Snowfox takes a different approach entirely. Instead of maintaining rules, it learns patterns from your historical invoice data — how your team has coded invoices for every supplier, every cost type, every context. It then predicts the correct coding for new invoices based on those patterns.
No rules to write. No templates to maintain. Just a model that improves every time your team processes an invoice.
The result: 90%+ automation rates from go-live, even for indirect spend categories that were previously considered too variable to automate.
