{"id":237198,"date":"2026-10-12T17:00:35","date_gmt":"2026-10-12T17:00:35","guid":{"rendered":"https:\/\/swissfederalism.ch\/ia-finanza\/"},"modified":"2026-10-12T17:21:20","modified_gmt":"2026-10-12T17:21:20","slug":"ai-finance","status":"publish","type":"post","link":"https:\/\/swissfederalism.ch\/en\/ai-finance\/","title":{"rendered":"Artificial intelligence and finance: skills and accountability to lead change"},"content":{"rendered":"

Artificial intelligence and finance: skills and accountability to lead change<\/span><\/h1>\n

AI opens up new possibilities for Finance leaders. Turning them into value requires reliable data, well-prepared people and an operating model that integrates automation and control. This approach is at the heart of the EFFAS workshop developed by Andrea Schenone<\/em><\/span><\/h3>\n
\"EFFAS
EFFAS AI CFO Executive Workshop<\/span><\/figcaption><\/figure>\n

Where should artificial intelligence be used within the Finance function? What value should it generate? How much autonomy should an AI agent receive? And who remains accountable for the results?<\/span><\/p>\n

These questions directly concern business leadership. The answers influence how work is organised, the quality of decisions and the development of professional skills.<\/span><\/p>\n

For Swiss companies, from SMEs to international groups, this topic can be approached by starting with practical needs: making a process more efficient, improving an analysis or providing management with more timely information. The decisive step is to connect these opportunities with organisational readiness and responsible execution.<\/span><\/p>\n

On this basis, I developed my AI CFO Framework<\/strong>, structured around four pillars: AI, People, Governance and Leadership<\/strong>. Its purpose is to support Finance leaders in designing a Finance function that combines new technological possibilities with professional expertise and clear accountability.<\/span><\/p>\n

Assess organisational readiness before choosing tools<\/strong><\/span><\/h2>\n

A credible transformation begins by examining everyday work.<\/span><\/p>\n

Where does the team spend its time? Which processes cause delays? Where do errors recur? Which decisions could benefit from better analysis?<\/span><\/p>\n

These questions help identify opportunities. Choosing which ones to invest in then requires assessing the availability of reliable data, clarity of responsibilities, access to systems and people\u2019s ability to verify AI-generated outputs.<\/span><\/p>\n

In my approach, potential use cases are assessed against five criteria: business value, feasibility, data readiness, risk and speed to impact.<\/span><\/p>\n

This comparison makes priorities explicit. An application focused on financial forecasting might require months of work on data. A more narrowly defined reporting project could allow useful evidence to be gathered sooner.<\/span><\/p>\n

The initial choice should therefore be a project that the organisation can realistically implement, evaluate and use as a learning opportunity.<\/span><\/p>\n

AI agents: define their mission, autonomy and limits<\/strong><\/span><\/h2>\n

AI agents can pursue an objective through multiple steps and tools. Within Finance, this creates possibilities for forecasting, reporting, financial closing, treasury and risk analysis.<\/span><\/p>\n

To understand the implications, consider an agent tasked with analysing variances against a forecast. It could retrieve authorised data, compare it with a baseline, flag anomalies and prepare a draft commentary.<\/span><\/p>\n

Before introducing it into the process, the organisation must establish which information it can consult, which actions it may perform, when human review is necessary and how exceptions should be handled.<\/span><\/p>\n

Identifying a variance, explaining it and approving a report for management are distinct responsibilities. The ability to perform one activity does not automatically justify authorisation to perform all subsequent activities.<\/span><\/p>\n

To make these boundaries explicit, the framework includes a Finance Agent Identity Card<\/strong>: a document defining the agent\u2019s mission, accountable owner, operating limits, performance measures and required controls.<\/span><\/p>\n

Oversight that genuinely enables intervention<\/span><\/h2>\n

Governance must be incorporated into process design.<\/span><\/p>\n

Data quality, access rights, the verifiability of outputs, vendor dependencies and operational continuity are factors that Finance leaders must consider from the outset.<\/span><\/p>\n

Human oversight also requires specific conditions. Anyone reviewing an output must have sufficient expertise, information, time and authority to challenge it and intervene. Formal approval alone does not guarantee effective control.<\/span><\/p>\n

For every pilot, it should be clear which circumstances require a pause, who can stop the process and which conditions allow it to restart.<\/span><\/p>\n

This means applying disciplines familiar to Finance to new technological possibilities: defined responsibilities, evidence, control thresholds and exception management.<\/span><\/p>\n

Design the new operating model<\/span><\/h2>\n

Introducing AI changes how work is distributed.<\/span><\/p>\n

Some activities may be automated; others will be supported by technology. Some must remain under direct human control. Others may be eliminated because they no longer contribute to the objectives of the process.<\/span><\/p>\n

These choices require a new Finance Operating Model<\/strong>: a deliberate definition of how people, data, systems and agents work together.<\/span><\/p>\n

Who is responsible for the data? Who designs and supervises the agents? Who validates the outputs? How are responsibilities divided between the central function and business units? Where is the time freed up reinvested?<\/span><\/p>\n

For organisations operating across multiple locations, this also requires a balance between shared rules and operational autonomy. One possible approach is to define common standards for access, controls and data quality while allowing individual units to choose the applications most relevant to their needs.<\/span><\/p>\n

AI Wealth Machine | Free AI Income Guide<\/strong><\/a><\/span><\/p>\n

People at the centre of the transformation<\/span><\/h2>\n

Finance expertise remains essential. Alongside it, new responsibilities emerge in agent supervision, output validation and the management of data and AI controls.<\/span><\/p>\n

Teams will need a stronger understanding of data and artificial intelligence, together with judgement, communication and critical analysis skills.<\/span><\/p>\n

An output that looks convincing must also be verifiable in substance. This requires knowledge of the process, an understanding of its assumptions and the ability to recognise when a conclusion overlooks relevant aspects of the business context.<\/span><\/p>\n

Training must also preserve the organisation\u2019s accumulated experience. If some activities are automated, organisations need to plan how professionals will continue to develop the knowledge required to supervise them.<\/span><\/p>\n

A 90-day plan to move into action<\/strong><\/span><\/h2>\n

A useful roadmap connects a selected opportunity to a controlled pilot.<\/span><\/p>\n

The plan must establish a baseline, appoint accountable owners, identify dependencies and define measurable success criteria. Alongside productivity and speed, it must consider output quality, control effectiveness and adoption by people.<\/span><\/p>\n

Review points enable decisions based on evidence: is the project ready to launch? Should it continue? Can it be expanded?<\/span><\/p>\n

Modifying or stopping a pilot can also be a useful outcome. It prevents an interesting idea from becoming a costly commitment without sufficient evidence.<\/span><\/p>\n

An executive programme with EFFAS<\/strong><\/span><\/h2>\n

These principles underpin the EFFAS AI CFO Executive Workshop: Preparing Finance for the AI Era<\/strong>, which I developed and will personally deliver as its instructor.<\/span><\/p>\n

The programme guides participants through five connected modules, with practical cases and tools to take back to their organisations:<\/span><\/p>\n