AI is changing finance. Can CFOs show what it delivers?

European CFO Survey, Autumn 2026

Richard Muschamp
Dr. Christian Pfennig
Albert Riba Barea
Ram Krishna Sahu

Half of the CFOs asked cannot yet demonstrate AI’s impact in finance to their boards. Yet 42% of respondents report actively transforming transactional processes because of AI. The next challenge is to establish what has improved, what it costs to sustain and which uses warrant further investment. 

AI is changing finance faster than its value can be demonstrated 

Finance routinely asks other functions to show what an investment will deliver. On AI, it now finds itself on the other side of the table. In Deloitte’s Autumn 2026 European CFO Survey, 42% of respondents report actively transforming transactional processes such as accounts payable, receivables and invoice matching with AI. Four in ten report transforming management reporting and variance analysis; 27% say the same of financial planning and analysis model-building. Activity is not the same as demonstrated value. Yet half of the CFOs asked about outcomes say AI’s impact in their finance function is not yet measurable. For CFOs, the task is to establish what AI has improved, what those improvements cost to sustain, and which uses warrant further investment. 

This does not mean AI is not delivering value. It means adoption has moved faster than finance's ability to show what its adoption has changed. That changes the question CFOs face. The debate is no longer whether finance should use AI. It is which uses have earned the next investment and what it will take for the others to catch up. 

The responsibility increasingly sits with finance. In Deloitte's separate global Finance Trends 2027 research, 54% of finance leaders report leading cross-enterprise AI and technology capital-allocation decisions.i Business functions may identify and lead their own use cases, while technology teams provide the architecture, data and security needed to implement them. Finance can help set the investment standard across the organisation: how expected value and ongoing cost will be assessed, what evidence is required and when further funding should be considered. 

Skip to description

Figure 1: AI has moved into the core of European finance

Source: Deloitte European CFO Survey, Autumn 2026. 
Question: Which finance roles or activities is your organisation actively transforming because of AI - whether through redesign, automation, or structural change - not planning to, but doing right now?
Note: This question was not asked in Germany, Switzerland and the UK. (N=598). 

The conditions for value matter as much as the measure 

Over four in ten (41%) respondents have no formal framework for assessing returns on finance AI. Another 25% use the same financial measures they apply to other technology investments. Only around a third (31%) are moving towards outcome-based or longer-term measures of value. That leaves an important task for CFOs: decide which changes each use case is intended to produce and how those changes will be assessed alongside its costs. 

Measurement alone will not make a use case ready to scale. Asked to name their biggest obstacle, 37% of respondents cite data and technology infrastructure, 23% team culture and change readiness, and 15% leadership alignment. The next worthwhile investment may therefore be in data, integration or adoption rather than another application. Cost visibility becomes more important as those investments grow. Deloitte’s separate global Finance Trends 2027 research finds that 60% of finance leaders expect AI costs and complexity to rise substantially through 2027.ii

Skip to description

Figure 2: What CFOs identify as the biggest barriers to scaling AI

Source: Deloitte European CFO Survey, Autumn 2026.
Question asked: What is the single biggest constraint on your ability to scale AI adoption within the finance function?
Note: Note: This question was not asked in Germany, Switzerland and the UK. (N=744). *The response option ‘Labour law or works council obligations’ was not given in the Netherlands, Portugal, Spain and Sweden.

How use cases are selected affects what finance can learn from its AI investments. In the survey, 37% of respondents say their organisations select AI use cases within finance case by case, 31% do so through an enterprise-wide programme, 21% have no defined process and 12% follow a finance-led roadmap. A case-by-case approach can be appropriate for testing new applications. But CFOs should apply consistent measures of benefit and cost and look for data and integration work that can serve more than one use case. Otherwise, it becomes harder to judge which results deserve wider investment.

Skip to description

Figure 3: How finance decides where to deploy AI

Source: Deloitte European CFO Survey, Autumn 2026.
Question: How does your organisation currently decide where to deploy AI in the finance function? Approaches shown are not a maturity ranking; the survey does not link approach to results.
Note: This question was not asked in Austria, Germany, Switzerland and the UK. (N=605).

The same investment discipline matters beyond finance’s own AI use cases. Business units may choose and lead applications in their areas, while technology teams provide the infrastructure and implementation. Finance’s role is to help set the standard for the business case including what outcome is expected, how the full cost will be captured and what evidence would justify further funding. That gives CFOs a way to steer AI costs and value across the enterprise without taking ownership of every deployment. 

Those standards need to be set before a pilot becomes routine operations. Deloitte’s October 2026 global State of AI in the Enterprise cautions that organisations can accumulate successful pilots without changing how work is done. It also finds that visibility into AI’s total cost is better than visibility into its unit economics: 71% of surveyed organisations track total AI spend, but only 44% track per-unit inference cost.iii A pilot may show that an application performs a task without establishing the cost of running it reliably at scale. 

For a CFO, the question before scaling is therefore not simply whether a pilot works. It is whether the use case improves the process it serves, what it costs to run as usage grows and whether those benefits justify the continuing commitment. Evidence needs to be designed into that decision from the start. The organisation needs to agree on the baseline, intended improvement and way to attribute costs before deployment, otherwise, a promising pilot may reach the next funding decision without a reliable assessment of its value.

Fund the outcome, not the tool

Half of the European CFOs asked could not yet identify a measurable AI outcome in their finance function from those listed in the survey. The question asked what AI had demonstrably achieved that they could evidence to their boards today. This does not mean AI has delivered no value; it means those respondents could not yet point to one of the specified outcomes. 

Of the respondents asked, 26% reported more forward-looking analysis and 21% said finance business partners spend more time with the business. 17% reported better data quality or infrastructure; 15% cited a shorter close, 14% each cited more accurate forecasts or improved compliance reporting; and 13% reported lower finance operating costs.

Skip to description

Figure 4: Half of finance functions cannot yet evidence a single AI outcome

Source: Deloitte European CFO Survey, Autumn 2026.
Question asked: Which of the following has your finance function demonstrably achieved because of AI - changes you could evidence to your board today?
*Note: This question was not asked in Austria, Germany, Switzerland and the UK. (N=600).

For CFOs, the question is whether a use case has moved beyond deployment to a measurable improvement, and whether the benefit justifies the full cost of sustaining it. A shorter close, for example, may put reliable information in leaders’ hands sooner, reduce rework or free staff to address exceptions. Each benefit has to be assessed alongside the technology, data, controls and changes to working practices needed to maintain it.

Better data can be a valuable outcome, even before a reduction in finance operating costs appears, provided finance can show what has improved and how that improvement supports dependable use. The point is not to postpone AI until every benefit can be monetised. It is to define the result each investment is intended to deliver, then use the evidence to decide whether to expand it.

Deloitte’s separate global Finance Trends 2027 research reports that among the respondents who use an internally driven process that puts measurement at forefront, 27% of finance leaders use stage-gate funding for major technology investments and 25% require a quantified return on investment for approval.iv Those findings describe wider technology-investment practices, not what European CFOs said they do with AI. They nevertheless illustrate a practical discipline: commit enough to test a proposition, then decide whether to scale, redesign or stop. 

The CFO's investment test: scale, redesign or stop

As AI use cases move beyond initial deployment, the next funding decision needs a clear basis. Within finance, CFOs can decide what to expand, redesign or discontinue. Across the business, they can work with business and technology leaders to set the standard for investment decisions, while the function that owns each use case remains responsible for its delivery and deployment.  

  • Define the result. Specify the process or decision the use case should improve, such as the close, forecasting, a control or a reporting cycle, and establish its starting point.  
  • Establish the full commitment. Assess the technology alongside the data, integration, controls, process redesign and adoption needed to sustain the improvement. Data infrastructure and change readiness are the most-cited barriers to scaling finance AI in the survey; test what overcoming them will require.
  • Set the decision point. Agree what result would justify further funding before the use case builds momentum. Scale what demonstrates sustained value at a supportable cost. Redesign what shows promise but is held back by a specific constraint, funding that constraint rather than an indiscriminate rollout. Stop what a fair test does not justify, even if the technology works.   

Finance does not need to own every AI use case to shape how it earns investment. Its contribution is to make the expected benefit, full ongoing cost and evidence for further funding visible and comparable. That is how finance can move from approving AI expenditure to helping steer AI value across the organisation. 

Activity and investment discipline are developing together 
In the Autumn 2026 European CFO Survey, 67% of Dutch respondents report actively transforming transactional finance processes with AI and 57% say the same of standard reporting. Approaches to deciding where to use AI in the finance function also differ: 52% of respondents in France and 48% in Spain report using an enterprise-wide programme or a finance-led roadmap.

A cost objective still needs evidence of value 
Cost reduction is the main reason for investing in AI in the finance function for 51% of Italian respondents and 43% of Swedish respondents. In the Netherlands, 38% instead prioritise freeing capacity for growth and strategic decisions. Yet 67% of Swedish respondents have no formal framework for assessing AI returns. Stating the intended benefit is not the same as demonstrating it. Separate national CFO surveys make the cost challenge more concrete. In Germany, only 11% of respondents say they can link AI costs to corresponding value contributions.v In Switzerland, 40% report AI costs higher than originally budgeted, although 79% expect a positive net financial benefit within three years.vi The Swiss figures concern company-wide AI spending, not finance AI alone.  

The obstacle to scaling depends on the market In Denmark, 49% report no defined process for deciding where to use AI in finance. In Italy, 34% cite finance team culture and change readiness as the biggest barrier to scaling AI, compared with 63% in Finland who cite data and technology infrastructure. Among all the participating countries, Austria has the highest share of CFOs who cite insufficient confidence in the business case (22%) as the biggest barrier. These findings call for different responses. Where selection is undefined, establish how opportunities will be assessed. Where teams are not ready, invest in skills and adoption; where infrastructure is the constraint, strengthen the underlying data and technology. And where the business case lacks credibility, make the expected benefits and full costs clearer before committing to scale.  

For CFOs comparing peers, the useful question is not which country is furthest ahead. It is whether the proposed investment has a clear objective, a credible account of its full cost and evidence that addresses the constraint on scaling it. 

Methodology

Deloitte has conducted the European CFO Survey since 2015, giving voice twice a year to senior financial executives from across Europe. The data for the Autumn 2026 edition was collected in September and October 2026 via an online survey and reflects responses from 1,170 CFOs in 12 countries, including Austria, Denmark, Finland, France, Germany, Italy, Netherlands, Portugal, Spain, Sweden, Switzerland, the United Kingdom.

Did you find this useful?

Thanks for your feedback