The CFOs we work with describe the same trajectory. After two or three years of investment, the finance function does fundamentally different work — not because the brief has changed, but because the lower layers of it have moved off humans entirely. The question is no longer whether to automate; it is what the function looks like on the other side of the automation, and how to govern the transition. Below is the picture we now draw for clients planning that journey.
What an AI-first finance function actually does
Three concrete things. First, every routine reporting task that used to take a person time happens automatically — the weekly business review, the monthly close, the variance commentary, the cash-position update. The human work is no longer producing the report; it is reviewing what the system produced and challenging it where it looks wrong.
Second, every forecasting cycle becomes AI-augmented. The base forecast is produced by the model; the planning team's job is to identify what the model missed and adjust. Teams move from arguing about the numbers to arguing about the assumptions — which is where the strategic conversation actually lives.
Third — and the one that surprises CFOs the most — the team that emerges is roughly half the size and twice as senior as the team they started with. The mid-level analyst layer is largely automated. The senior finance partners are doing the partnering work the old function never had the bandwidth for.
The team transition is the hard part
It is tempting to skip past this section. The teams we have seen succeed in this transition have not. The honest picture: in a typical programme, about a third of the headcount reduction is involuntary. The remaining two-thirds is natural attrition the firm does not replace, plus internal moves into other parts of the business where the skills remain valuable.
People lose their jobs in this transition. It is largely unavoidable. How you treat the involuntary cohort is the most important thing about the programme — and the part most firms get wrong.
The cohort that leaves involuntarily deserves more care than the industry norm. The firms we admire spend on generous severance, long notice, and internal placement support, and they design the transition with that cost line in the business case from day one. It is more expensive in year one. It is much cheaper in years two and three, because the rest of the team keeps trusting leadership.
What to build versus buy
Buy the foundations. The ERP modernisation, the data platform, the BI layer — these are commodities now, and the firms that try to build them are spending money they should not be spending. Differentiation does not live there.
Build two things. First, a financial-controls AI that watches transactions in real time for fraud, policy violations, and unusual journal entries. The off-the-shelf products are improving but still trail what a careful in-house build can do against your specific control framework. Second, an internal copilot for the finance team — built on top of the data platform, with access to internal financial data under appropriate permissions, so that anyone in finance can ask questions in plain language and get answers with the source data linked. We use Development AI as the underlying engine on our engagements; the integration into client data is bespoke each time.
What does not work
Two patterns consistently fail. The first is trying to fully automate M&A diligence. The technology is capable, but the legal and reputational stakes of a wrong answer are too high, and the right architecture uses AI as one input to a human-led process — not the other way around. We have seen teams burn a year trying to reverse that arrangement before reverting.
The second is customer-facing AI that produces credit decisions without a human in the loop. The accuracy can be higher than the human-led baseline, but the regulatory questions are not yet answerable in most markets, and supervisors push back. The pragmatic design ships with a human-in-the-loop architecture; the fully automated version waits for the regulatory framework to mature.
What the CFO does next
When the routine work is automated, the CFO's role re-weights into three things in roughly equal measure. Capital allocation, with vastly better information than the function used to have — faster data, more granular, more reliable. Governance — model-risk management, audit, and assurance work — which becomes more important rather than less. And partnering with the business at a level the old finance function never had time for. When the senior team has the bandwidth, they can sit in operating meetings and help leaders make better decisions. That is the most rewarding part of the new shape of the role, and the part most CFOs underinvest in.
- The point is not headcount; it is what the team does. Half the size, twice as senior — and doing the partnering work the old function never had time for.
- Govern the transitions with care. Some people will lose jobs. How you handle that cohort is the most important thing about the programme.
- Build the copilot, buy the platform. Differentiation lives in the bespoke layer on top of standard infrastructure — not in the infrastructure itself.
If you are scoping work in this space and want a second pair of eyes on the design, talk to us.