We built Spendaq because month-end close is still broken.
Yuki Tanaka spent six years as a Controller at Hadleigh Industrial Group in Charlotte — watching her team spend the first week of every month-end reclassifying the same categories: a ZOOM.US charge landing in Office Supplies instead of SaaS, a consulting invoice split between two cost centers every single time. The ERP was on a version from 2008. The model accuracy to fix this had existed for years. No one had applied it specifically to GL categorization for mid-market finance teams.
Finance teams shouldn't spend the close week fixing what software should have gotten right.
Most ERP categorization logic was written 15–20 years ago, for a world where every transaction fit neatly into one vendor, one cost center, one account code. Modern corporate spend is messier: SaaS subscriptions billed in three different ways by the same vendor, multi-line AP invoices that span departments, card swipes with truncated merchant codes that look nothing like the vendor's actual name. Spendaq applies current machine learning to the specific, unglamorous problem of GL mapping accuracy — not to the broader "spend management" category, just the categorization part that finance teams still do manually every month.
The team
Former Controller at Hadleigh Industrial Group (Charlotte, NC). Spent 6 years doing manual GL reclassification before building Spendaq. BA Accounting from UNC Charlotte, CPA licensed.
Machine learning engineer with a background in transaction classification and financial data pipelines. Spent three years building vendor categorization models at an AP automation company before co-founding Spendaq. BS Computer Science from NC State. Responsible for the GL mapping engine, confidence scoring, and the active learning loop that improves accuracy on account-specific corrections.
How we work
We ship categorization improvements only when they test above the current baseline. A bad AI suggestion that a Controller has to override is worse than no suggestion.
We don't think of this as an AI product that happens to touch finance. We think of it as a finance operations product that happens to use AI. The distinction matters in every design decision.
Spendaq does GL categorization. Not expense management, not corporate cards, not procurement, not AP payment processing. The depth of the GL mapping problem — handling every ERP's quirks, every company's chart of accounts, every vendor's billing format — requires that kind of focus. We haven't broadened the scope, and we don't plan to.
We show confidence scores. We don't hide low-confidence mappings behind a clean interface. If Spendaq isn't sure, you'll see that — not a false green checkmark.