The augmented CFO: faster close, reporting and analysis

A slow close is a flow problem

In many SMEs and mid-sized companies, the monthly close finishes somewhere between the tenth and fifteenth working day. By the time figures for the past month reach the management committee, the next month is well under way. The CFO delivers information that is accurate but late, and decisions are made on impressions.

The common assumption is that the close is slow because accounting work takes time. In most of the cases we see, that is not the main reason. The calculation itself accounts for a modest share of the elapsed time. The rest is waiting: an invoice not yet entered, an intercompany confirmation that has not arrived, a journal corrected three times, an approval stuck with someone travelling. The close is a flow problem before it is a tooling problem.

Where the time goes: a worked example

Consider an illustrative example with explicit assumptions. An industrial group has revenue of CHF 80 million, three entities (Switzerland, France, Germany) and a finance team of eight. The close ends on working day 10 (WD+10), broken down as follows:

• WD1 to WD2: sub-ledger cut-off (purchases, sales, inventory) and bank reconciliations.
• WD3 to WD5: accruals, provisions, intercompany transactions, currency differences.
• WD6 to WD7: consolidation and adjustments.
• WD8 to WD10: variance analysis, commentary, management sign-off.

Assume that, of those ten days, close to half of the elapsed time is waiting or rework rather than productive effort. This is an assumption, but it is consistent with what we often observe. It has a consequence: speeding up individual tasks achieves little while the queues remain. A faster accountant does not compensate for an intercompany confirmation received on day six.

In this example, the analysis phase (WD8 to WD10) is also where the CFO adds the most value, yet it is squeezed into the end of the process when the team's energy is spent. Shortening the close is therefore not about accounting comfort. It is a way of giving time back to analysis.

The levers, in order

1. Close continuously rather than at month-end

Weekly bank reconciliations, invoices entered as they arrive, chasing missing documents before the month ends: these habits cost almost nothing and move a large part of the workload out of the critical window. It is the least technological lever and often the most effective.

2. Set a calendar, owners and thresholds

A close calendar with one task, one owner and one deadline per line, combined with materiality thresholds, prevents the same effort being spent on a CHF 200 difference and a CHF 200,000 one. For small amounts, a reasonable estimate is enough, with true-up the following month. A soft close at WD+5 followed by a final close later is a common and sound compromise for internal steering, provided statutory obligations (the Swiss Code of Obligations, French accounting rules and tax calendar, or US GAAP requirements for those who report under it) are still met for the formal accounts.

3. Settle intercompany balances before the close

Reconciling intercompany balances at WD-3 rather than WD+4 removes one of the most frequent causes of waiting in multi-entity groups.

4. Automate matching and data entry

Well-configured bank matching and ledger reconciliation rules handle the large majority of entries without intervention. AI enters here in a targeted way: extracting data from unstructured supplier invoices, suggesting account coding from history, flagging unusual entries. It is useful where a simple rule does not work. It is unnecessary where a rule will do.

5. Draft the analysis faster, validate it carefully

Language models can produce a first draft of variance commentary from ledger data: revenue movement by line, margin trends, unusual items. The draft saves time. It does not replace the judgment of someone who knows the commercial context, and it should be reviewed the way one reviews the work of a capable intern with no memory of the company.

What AI changes, and what it does not

AI lowers the cost of reading, sorting and first-draft writing. It does not fix an inconsistent chart of accounts, bypassed purchasing processes or poor master data. Where upstream data quality is weak, an AI layer produces errors faster and with more apparent confidence. This is why we put levers 1 to 3, which rest on common sense and organization, ahead of levers 4 and 5.

Expected gains should stay measured. Again with textbook assumptions: if the close consumes the equivalent of 35 person-days per month and both duration and effort fall by a third, about 12 person-days are freed each month, roughly half a full-time position. That time has value only if it is explicitly reassigned to analysis, budget preparation or cash monitoring. Otherwise it dissolves.

Control, data and compliance

Speed must not weaken internal control. Three principles seem structural to us.

• AI proposes, a person validates. Suggested entries or generated commentary remain the responsibility of an identified employee, with a record of what was proposed and what was retained.
• Segregation of duties holds. A tool that both enters and approves is a risk, however fast it is.
• Personal data is protected. Accounting documents contain names, expense claims, sometimes payroll elements. The revised Swiss Federal Act on Data Protection (FADP), in force since 1 September 2023, requires companies to identify these flows, control their processors and know where data is processed. GDPR applies in the same way to the group's EU entities.

On AI regulation, the EU AI Act entered into force on 1 August 2024 and applies in phases: prohibitions in February 2025, obligations for general-purpose AI models in August 2025, most of the remainder in August 2026. Ordinary accounting and reporting uses are generally not high-risk systems, but keeping a simple register of the tools used and their purpose is good practice, including for a Swiss company operating in the European market. In the US, listed companies already face internal control requirements; for private companies, it is often lenders or financial sponsors who set the deadline for monthly packages, and the logic is the same.

Implications for leadership

A few reference points for a CEO or CFO who wants to move.

1. Measure first. Time each step of the close over two or three months, separating working time from waiting time. Without this, automation is a guess.
2. Set a realistic deadline. Moving from WD+10 to WD+6 within six to nine months is a plausible order of magnitude for a reasonably organized SME. Aiming for WD+2 is rarely relevant.
3. Connect the close to decisions. A fast number has value only if it feeds something useful. This is the logic we develop in the executive dashboard that actually supports decisions, and it is visible in an automated executive dashboard.
4. Link it to cash. A faster close improves the data feeding cash flow forecasting, from spreadsheet to continuous steering.
5. Prepare the team. The shift from producing numbers to interpreting them cannot be decreed. It is a matter of organization and adoption: training, clear responsibilities, and acceptance that some tasks will disappear.

The augmented CFO is not the one with the most tools. It is the one whose team spends less time producing figures and more time explaining what they mean for management decisions. Our approach to Finance & performance starts from that idea: strategy, then deployment, then steering results, with AI where it helps and common sense elsewhere.

If you want to know where the waiting, the rework and the reachable gains sit in your own close, our diagnostic is a pragmatic starting point, and helps prioritize actions before any investment is made.