Field notes on AI in finance.
Short, practical reads on what actually works when you put AI agents to work inside close, AP, AR, treasury and FP&A — written by the Vaku team.
Most finance teams can tell you exactly how long close takes — and it's almost never as fast as they'd like. The usual culprits aren't exotic: checklists live in spreadsheets, reconciliations are done by hand, accrual estimates are built from last quarter's memory, and variance commentary gets written the night before the board deck is due.
None of that is a tooling problem in the traditional sense — most teams already have an ERP, a BI tool, maybe a close-management product. The gap is execution: something has to actually run the checklist, chase the open items, flag the one reconciliation that's off by $4,200, and write the first draft of why revenue missed plan.
That's the shift AI agents make possible. Not a dashboard that shows you close is late — an agent that runs recs overnight, drafts the accrual journal entries from historical patterns, and tells you at 8am which three items need a human decision today. The median close time we see teams move to is 3 days, down from 7–10 — not because the close process changed, but because the busywork inside it got automated.