The Continuous Close: Start with Reconciliation

Every few years, accounting software vendors resell the same dream with a new adjective in front of it: real-time close, continuous close, now agentic close. The pitch is always some version of “why wait until month-end to know your numbers.” It’s a good pitch. It’s also, in the fully-automated form usually shown in the demo, not how most finance orgs actually operate today—and I’d argue shouldn’t, yet.

Why the full vision keeps not arriving

A real month-end close isn’t just arithmetic. It’s judgment calls: which accrual estimate is close enough, which intercompany balance needs a manual true-up, which vendor invoice is disputed and shouldn’t hit the books as-is yet. A genuinely continuous, fully automated close requires those judgment calls to either disappear (unlikely—the business keeps generating edge cases) or to be made by a system with enough context to make them reliably (not there yet, and I’d be nervous about anyone claiming otherwise for anything beyond the most mechanical entries).

So the full vision—close the books continuously, no month-end crunch, ever—keeps getting demoed and keeps not fully arriving. That’s not a knock on the vendors; it’s a mismatch between a genuinely hard problem and a marketing timeline.

What’s actually working right now

The narrower, more practical version is continuous reconciliation, not continuous close. Bank feeds, subledger-to-GL matching, intercompany elimination candidates, and routine journal entries can run more frequently with mature tooling. AI-assisted anomaly flagging may help identify unusual entries, but teams should measure its accuracy and review effort. Hypothetical: A team automates recurring matching and reviews exceptions throughout the month; month-end still requires judgment on estimates, disputes, and unusual transactions.

I’d take that version every time over a vendor promising a fully automated close, because the narrower version is honest about where the hard part of the job actually lives.

The trap to watch for

The risk I’ve seen teams fall into is treating “we automated the reconciliation” as “we can now compress the review cycle by the same proportion.” It doesn’t work that way. If reconciliation used to take 60% of close time and automation eliminates most of it, you haven’t necessarily freed up 60% of the calendar—you’ve freed up the mechanical time, and the judgment-heavy 40% still needs the same thinking time it always did, sometimes more, because reviewers now need to trust an automated match instead of having personally traced it.

Build the timeline savings estimate around that reality, not around the naive “we automated X% of tasks, therefore close is X% faster” math. The teams that get burned by continuous-close projects are usually the ones that promised the naive version to leadership and then had to explain, mid-quarter, why close day 3 still feels like close day 3 used to.

What I’d actually pilot

Pick your highest-volume, most repetitive reconciliation (bank feeds are usually the easiest real win) and get that genuinely automated and trusted before touching anything closer to judgment. Prove the pattern narrow and deep before it becomes a “close transformation” program with a name and a steering committee. The unglamorous version is the one that actually ships.