Lights-Out Finance:
what the CFO owns when agents run the close
Finance has always been the most automatable function in the enterprise and the least allowed to fail. High volume, rule-dense, deadline-driven, and audited — the exact profile where the economics of autonomy are sharpest. That is why the first question about Lights-Out Finance is never whether agents can run the close. They can. The question is what the CFO's organization owns once they do.
What changes
In an autonomous finance function, agents run reconciliation continuously rather than at month-end; journals post themselves with rationale attached; intercompany clears without email threads; the forecast updates when drivers move, not when the planning calendar says so. The close stops being an event and becomes a state — the books are always nearly closed, and quarter-end is a verification, not a sprint.
People do not disappear from this picture; they move up. Exception queues replace task lists. The controller's team stops preparing and starts challenging. FP&A argues with assumptions instead of consolidating spreadsheets. And the scarce senior judgment that used to drown in mechanics is finally applied to the decisions that needed it.
The CFO who hands processes to agents does not give up control. Done properly, they gain a kind of control no manual function ever offered: full population, every transaction, every night.
What the CFO must own
Three things cannot be delegated to the machine. First, the guardrails: which decisions agents take alone, which require a human, and where the thresholds sit — these are policy choices, and they belong to the CFO, not the vendor. Second, the exception model: an autonomous function is only as good as the quality of what it escalates, and designing that queue is operating-model work. Third, the attestation: SOX and its equivalents do not care that an agent did the work; they care that the control operated and the evidence exists. Explainability and audit trail have to be designed in at the start — retrofitting them is the most expensive mistake in this field.
Where to begin
Reconciliation and AP are the proven entry points — high volume, clear rules, measurable from day one, and already demonstrated at touchless rates above ninety percent at global scale. From there the path runs through close orchestration and reporting toward planning, where agents do the assembling and people do the deciding. Each step pays for the next; none requires a big-bang program.
The destination is a finance function that runs overnight, at quarter-end, and through Day-1 events — with a team that has moved from producing the numbers to governing them. That is not a smaller finance organization. It is a more senior one.