In most banks the finance calendar still runs on a monthly heartbeat: transactions accumulate, sub-ledgers are closed, reconciliations are chased, intercompany positions are agreed, journals are posted and commentary is written — then the cycle begins again. The work is essential, but much of it is assembly rather than analysis. And by the time management sees the numbers, the month they describe is already history.
The benchmarks
APQC's benchmarking of more than 2,300 organisations puts the median monthly close at 6.4 calendar days, measured from running the trial balance to completing consolidated statements. Top-quartile performers close in 4.8 days or less; the bottom quartile needs 10 days or more1. The main barriers APQC identifies are data quality, lack of a standardised chart of accounts and insufficient process documentation1.
How long it takes to close the month
Cycle time from trial balance to consolidated statements, calendar days (days)
Note: Top quartile is 4.8 days or less; bottom quartile is 10 days or more.
Planning tells a similar story. The 2025 FP&A Trends Survey found that high-value work such as decision support and storytelling had fallen to 31% of FP&A time, from 35% in 2024, while 46% of capacity was absorbed by manual effort including data-quality issues, reconciliations and preparation2. Just 15% of organisations can complete a forecast in under two days, 29% need more than ten, and only 17% report good data quality3. More than half of FP&A teams — 54% — describe themselves as merely coping with their workload, the highest level in four years2.
Where FP&A time goes
Share of FP&A capacity, %, 2025 (%)
Note: 'Other activities' is the remainder of the two reported categories (100 − 46 − 31).
AI adoption has stalled — for a reason
Finance teams have taken up AI, but adoption has largely plateaued, and the obstacles finance leaders describe are less about the models than about data literacy, technical skills, and data quality and availability.
The pattern is telling. Finance AI has found its footing in tasks that sit on top of documents and transactions. It has made less headway in the close and in planning, because those depend on a clean, reconciled, well-described view of the ledger that many banks still assemble by hand each month.
For bank CFOs the implication is uncomfortable but useful. Buying more AI tools will not, on its own, shorten the close or free analysts for planning. The gains come from changing when and how the underlying work is done — and from giving AI a single, reconciled version of the numbers to work with. That is a finance-operating-model decision as much as a technology one, and it belongs on the CFO's agenda rather than only the CIO's.
What a continuous close looks like
A continuous close does not mean closing the books every day. It means doing the close work — matching, reconciling, accruing, eliminating, explaining — as transactions arrive, so that period-end becomes a confirmation rather than a scramble. Four capabilities make that possible:
- A governed accounting data layer. Sub-ledger and general-ledger data flow continuously into a single model with lineage, so every number can be traced to source. This is the role of an accounting hub such as LEDGER360.
- Always-on reconciliation. Agents match and investigate breaks daily, escalating only genuine exceptions to accountants — the pattern behind CLARION, and consistent with the large task-time reductions reported for agentic reconciliation elsewhere in the industry.
- Automated intercompany. Intercompany positions are matched and disputes surfaced during the month rather than at period-end, as in NEXUS.
- Draft commentary on demand. Variance explanations are drafted by agents from the reconciled data and reviewed by finance, as NARRATOR does, instead of being written from scratch at quarter-end.
Banks face particular complexity here. A universal bank's ledger is fed by dozens of product systems — lending, deposits, cards, treasury, markets — each with its own sub-ledger, cut-off conventions and reconciliation to the general ledger. Intercompany flows between legal entities and branches add another layer, as do regulatory reporting requirements that draw on the same data. This is why close acceleration in banks has historically depended on large reconciliation teams, and why agents that can investigate breaks across systems are such a natural fit.
The sequencing matters. Banks that deploy AI commentary before they have fixed reconciliation end up with articulate explanations of numbers nobody trusts. Banks that start with the data layer and reconciliation find that commentary, forecasting and analysis become far easier to automate afterwards, because the inputs are already clean and traceable.
From faster close to better planning
The payoff of a continuous close is felt most in planning. When actuals are trustworthy mid-month, FP&A can move to rolling, driver-based forecasts and scenario analysis rather than rebuilding spreadsheets. AI then becomes useful in the way finance leaders hoped: answering ‘why did net interest income move?’ from governed data, simulating rate and volume scenarios, and drafting the board narrative for a human to refine. This is the design behind COMPASS, which links ledger, planning and performance management with agents working under named finance owners.
The fastest route to AI-assisted planning runs through the reconciliation queue. Fix the close, and the forecast follows. (SCIKIQ view)
Governance matters as much in finance as in any other domain. Agents that post or propose journals, clear breaks or draft disclosures must operate within segregation-of-duties rules, with every action logged and a human approving anything material — requirements that SOX-style controls already impose on people.