When the plan is no longer enough: AI-assisted scenario planning in an interim CFO assignment

From practice · 17 October 2026 · 8 min read · Manfred Schönberger

The revenue plan is agreed. Then a customer postpones an order, margins come under pressure and receivables arrive later than expected. Each deviation would be manageable on its own, but together they can change liquidity within a few weeks. In an interim CFO assignment, finance leadership must recognise this effect early and prepare a decision.

Scenario planning connects business assumptions with the income statement, balance sheet and cash flow. Artificial intelligence shortens the preparation by reviewing data, grouping anomalies and quickly drafting revised scenarios. Its output remains a working version, approved by people who are responsible for both the business and its financial consequences.

Planning under uncertainty

In spring 2026, Deloitte surveyed 1,136 CFOs across twelve European countries: 76 per cent rated external financial and economic uncertainty as high or very high, while 51 per cent had used scenario analysis or impact assessments during the preceding year. These figures explain why an annual budget alone provides too little guidance.

The budget sets out a course of development, the forecast updates it with new information, and a scenario tests the combined effect of several changes. These might include weaker order intake and rising procurement costs alongside longer payment terms. Their combined effect shows finance leadership how much time remains before a shortfall and which corrective action could take effect in time.

In its “Futures Toolkit”, updated in 2024, the UK Government Office for Science explicitly describes scenarios as possible futures that are neither forecasts nor plans. This distinction matters to the CFO. Management tests a decision under different conditions, looks for early warning signals and prepares an appropriate response, rather than receiving a prediction with misleading precision.

The starting point is a decision question. “What happens in a recession?” offers little help, so I narrow the question in an assignment: Will the available financing be sufficient if order intake falls for two quarters and customers pay ten days later? Under which conditions can the planned investment continue, and at what threshold should a discussion with the bank be prepared?

Questions like these give the model direction and prevent the team from spending time on scenarios that have no bearing on the next decision.

The financial baseline

An assignment rarely starts with AI. Actual figures, unresolved closing issues and data sources must first be clarified. For example, an unreconciled inventory balance, order data without a binding contract status and project margins remain uncertain after automated processing. Sales and controlling define reliable data differently.

The work therefore starts with the current position. Closing, planning and reporting are connected with technical implementation. Data from ERP and upstream systems is brought together, metrics are clearly defined and processes are documented so that every figure can be traced to its source.

A small number of well-chosen drivers is enough for scenario planning, with the selection following the business model. In trading, sales volumes and selling prices are particularly influential, alongside purchase prices and payment periods; inventory turnover determines how much capital is tied up in stock. For a service provider, utilisation and project margins matter most, together with the timing of invoicing and cash receipts. As interim CFO, I agree this logic with sales and those responsible for purchasing and operations: finance owns the calculation; the assumptions come from the business.

Three working scenarios for a management decision

For an initial management discussion, a baseline and two adverse developments are usually sufficient. Clear distinctions between them matter more than their names.

The model processes all three scenarios under the same rules, keeping differences explainable. A separate sensitivity analysis also shows which driver has the greatest effect. When planning actions, this information is often more useful than another scenario.

AI's contribution

AI speeds up the work around the model by bringing together variances from several reports, mapping external signals to the affected drivers and drafting a coherent set of assumptions. When new data becomes available, it describes which changes have shifted cash flow and prepares initial commentary for management.

A controlled model remains responsible for the financial calculations. Excel is sufficient when formulas and inputs are managed properly and every version remains traceable; a planning or BI solution may be suitable for multiple entities or frequent updates. The use case determines the appropriate tool. A new system does not resolve unclear logic.

For its “Future Ready CFO 2026” study, Wolters Kluwer surveyed 1,672 finance leaders in 20 markets: 60 per cent expect AI and advanced analytics to significantly change scenario planning; for budgeting and forecasting, the figure is 62 per cent. Only 18 per cent describe their finance function as digitally advanced, leaving a substantial gap between expectations and operational maturity.

This gap also appears in the measurement of benefits. According to Protiviti's 2026 study, 77 per cent of the finance organisations surveyed use AI, and its use for financial forecasting rose from 58 to 76 per cent within a year; only 35 per cent consider their measurement of AI benefits effective. Metrics are therefore defined before every pilot, such as the time needed to update the outlook, the number of manual transfers, errors detected and the time from warning signal to decision.

Review and responsibility

Language models can present an incorrect explanation convincingly. Every proposed assumption therefore needs a source, a data reference date and an accountable owner. Changes to the model are logged; the CFO approves the financial logic, while management decides on actions and risk appetite.

Data that leaves the company requires particular attention. According to the Swiss Federal Data Protection and Information Commissioner, the Swiss Data Protection Act applies directly to AI-assisted data processing; the purpose and operation must be transparent, as must the data sources used. Payroll, customer and supplier data therefore belongs only in an approved environment with controlled access rights; anonymised or synthetic datasets are sufficient for early tests.

Professional oversight can be organised simply: AI prepares, the defined financial model calculates, and a responsible person reviews the results before approval. Postings and payments, critical master data changes and financing decisions continue to pass through the agreed controls.

From scenario to action

Without a threshold, a scenario remains an interesting calculation. Leadership needs a trigger: if order intake falls below the agreed level, the forecast is updated; if collection periods increase, cash forecasting is monitored more closely; if margins fall below their limit, the company reviews prices, purchasing terms or its product range.

Each trigger is linked to an action, a responsible person and a deadline, changing the nature of the discussion. Participants then spend less time debating whether a figure is plausible and more time deciding when to act. In a time-limited assignment, this discipline matters because the company must later continue the process itself.

BCG describes financial planning as a system that identifies weaker order intake, traces its effect through to the revenue forecast and proposes an adjusted scenario. Depending on the planning environment, such applications are at different stages of maturity, so the starting point should remain a use case with a verified data flow. Clear acceptance criteria are set in advance; after the pilot, the company decides on further expansion.

If leadership capacity is temporarily missing for these decisions, an interim finance leader can take responsibility for planning, actions and handover with your team.

Working rhythm and handover

Scenario planning is a recurring process. When liquidity is tight, actual payments and assumptions are reconciled weekly, while a monthly meeting may be sufficient for an investment decision. The frequency depends on the time between a warning signal and an effective response.

At every update, the report answers the same questions: How do earnings, the balance sheet and cash flow change? Which limit is reached first? Which action is prepared, and who decides by when? AI can prepare the comparison between scenarios. The management decision remains outside the system.

At the end of the interim CFO assignment, the team receives more than a file. Alongside the reconciled model, it needs an assumptions register, with documented data sources and thresholds. Responsibilities and meeting schedules are also part of the handover, because otherwise the planning soon becomes outdated after the assignment.

AI makes scenario work faster. The rules of financial leadership make it reliable.