A report released earlier this month by Anthony Barbey, fractional CFO and founder of AI-native spreadsheet developer Layerz, set out to identify the key use cases of AI for fractional CFOs. He suggested that the hours AI has not yet reached are ones that sit upstream – such as chasing clients for information, repairing models and cleaning the books.
For smaller businesses, the biggest delay in any close is waiting on the client, agrees Anurag Kakar, CFO, Partner and Co-founder of fractional CFO firm Azurium.
“Chasing a missing receipt or a transfer nobody can explain immediately without digging through their inbox used to mean a team member chasing over and over again,” he says. “Now AI writes a specific, clear request with instructions if necessary, follows up on schedule and reads the reply when it comes in. It easily saves us an hour or two per client each month.”
Model repair is the same story. AI will find a broken link or a number hardcoded into a formula in seconds that a person could spend hours on, depending on how good they are with Excel – although Kakar says a human should still sign off on the fixes.
Building simple models is an equally important feature according to Jerome Le Louer, Interim Fractional CFO and Managing Partner of smeCFO.
“Business plans for a fundraise, a full set of financials with P&L, balance sheet and cash flow, operating budgets, scenario models – AI produces a clean first version in hours rather than days, which lets me spend the time on the assumptions rather than on the plumbing,” he says.
Five-year business plan
La Louer says he used the technology to build a five-year business plan for a Thai food and beverage company. He recalls that the structure, drivers and sensitivity tables were all in place by the time he sat down with the founder, so the meeting was about the numbers rather than the format.
“AI now drafts almost every email I send, from each client’s own history, tone and open items, so I review and hit send rather than start from scratch,” he adds.
“The same engine works for AR [accounts receivable]. The accountant gets a ready draft for every overdue invoice, every missing withholding tax certificate, every supplier statement the client has not returned, personalised and sent on a cadence, without the team having to remember who owes what.”
The caveat is pace. If the client is not yet using AI, this can feel like too much too fast – so part of the job now is calibrating how fast to run on each mandate, not just how much AI can do.
The numbers inside the spreadsheet
Despite massive growth in the use of Claude, the Layerz report found that four in five of the CFOs surveyed were still using Excel. For almost half of survey respondents, the AI sees the numbers inside the spreadsheet.
For now the change is in who builds the model, says Kakar. “AI will set up the model, write the formulas, check the logic and explain in plain English what a tab is doing,” he adds. “That is a big deal for a small business owner who has always found spreadsheets intimidating.
“But the output still has to live somewhere auditable and for the foreseeable future that is Excel, albeit building, running and checking it gets a lot quicker.”
Excel is the format auditors, banks, boards and investors expect to receive the numbers in and that will not change because the tool that builds the file has changed. What changes is who types, suggests La Louer.
“I increasingly describe the model I want and review the workbook that comes back, formulas and all,” he says. “When a fintech client needed an intercompany reconciliation for its Series A data room, I described the matrix I wanted and AI produced the workbook: matching logic, classification columns and corrective journal entries already keyed to the chart of accounts.”