ExCaMa

ExCaMa Perspective

AI is now part of business. Its value is still rarely demonstrated.

Many companies use AI and report improvements, yet its economic contribution is often difficult to establish. The finance function can help clarify costs and benefits and support effective implementation.

Zwei Personen und ein humanoider Roboter betrachten gemeinsam einen Bericht und einen Laptop in einem hellen Büro.

Artificial intelligence is being used by a growing number of companies. It helps handle customer enquiries, speeds up processes and opens up new possibilities for products and services. As its use grows, a question central to the financial management of any organisation becomes more pressing. What does AI actually contribute, and what does it take to deliver that contribution?

The finance function has an important role to play. It can assess the business case for individual applications, make costs transparent and put in place the conditions needed to demonstrate the value AI delivers.

Improvements are becoming visible

According to a representative Bitkom survey of 603 companies with at least 20 employees, 57 per cent of companies in Germany use AI in 2026, the first time a majority have done so. Two years earlier, the figure was 20 per cent.

Many companies using AI already report improvements. Some 66 per cent say their competitive position has strengthened, 48 per cent say they have significantly accelerated internal processes and 45 per cent see a measurable contribution to business performance.

These findings show that companies perceive benefits. To inform business decisions, however, those improvements need a closer look. Shorter processing times can free up capacity for other tasks. Whether that leads to higher revenue, lower costs or better quality depends on how the organisation uses that capacity.

Financial outcomes continue to vary

International evidence adds to this picture. In PwC’s 29th Global CEO Survey, which drew responses from 4,454 business leaders across 95 countries and territories, 56 per cent reported that AI had neither increased revenue nor reduced costs over the previous twelve months. Twelve per cent said they had achieved both improvements.

The surveys cover different groups and ask different questions, so their percentages cannot be compared directly. Taken together, however, they draw attention to how organisations can turn AI use into economic results they can demonstrate. The figures reflect respondents’ own assessments and do not replace an evaluation of individual use cases.

Bundesbank President Joachim Nagel also noted in his speech of 21 April 2026 that AI’s effects on the economy as a whole have so far been less apparent in aggregate statistics. His comparison with electrification illustrates the importance of organisational change. New technologies realise their potential in part as processes and structures evolve.

The finance function puts expectations on a sound footing

The Deloitte CFO Survey for autumn 2026, which polled 126 chief financial officers at German companies, offers a particularly useful perspective. Only 17 per cent already see AI making a positive contribution. Seventy per cent expect it to do so within the next two years.

Expectations are well ahead of what companies have been able to measure so far. At the same time, only eleven per cent of the finance chiefs surveyed can link AI costs to the value they generate.

This gives management a clear task. Expectations need to be turned into assumptions that can be tested. Which tasks should improve? How will we recognise that improvement? What costs will arise, and over what period should the investment pay for itself?

A framework of this kind allows organisations to review what they learn systematically and decide how to develop individual applications.

Account for the full cost

An assessment should include all costs associated with a use case. Alongside licences and usage fees, these may include infrastructure, data preparation, integration, training and staffing. The ongoing work of checking and refining outputs should also be taken into account.

Citing Bundesbank business surveys, Nagel explained that spending on generative AI in Germany was already comparable to spending on conventional digital investments. A large share went towards recurring costs, such as subscriptions and permanent IT staff.

An economic assessment should therefore capture both one-off and recurring costs. An application may be relatively inexpensive to introduce and require considerably more resources once it is widely used. Conversely, some initial spending may be essential to make later use possible.

Costs become clearer when they are assigned to a specific use case and assessed over an appropriate period.

Define the benefit for each use case

Every AI initiative needs a clear objective, a baseline and suitable measures. Depending on the task, these might include turnaround times, error rates, cost per transaction or the quality of an output.

It is useful to assess operational improvements separately from their economic effects. Completing a task more quickly saves time. To establish the financial benefit, the organisation also needs to determine how that time is used and what further costs are involved.

Improvements in quality should also remain visible. Greater reliability, better information or a quicker response to customer enquiries can all be valuable to an organisation. Their contribution should be described clearly, even if it cannot immediately be expressed in full in euros.

The assessment becomes more informative when results are regularly compared with the baseline and the original expectations.

Connect planning with implementation

Economic assessment should continue throughout the use of AI. Assumptions about costs and benefits belong in the planning stage. Reporting should show what actually happens. This provides a basis for deciding which applications to adjust, expand or discontinue.

The finance function helps bring this work together. Working with business teams and IT, it can develop assessment criteria and check whether the expected improvements are being realised in day-to-day work.

Finance and administrative processes can offer suitable starting points where workflows are clearly defined and results can be measured reliably. Here too, the specific task and the available data need to be assessed.

Clear responsibilities support reliable results

Demonstrating lasting value requires organisations to use AI outputs responsibly. This includes adequate data quality, clear responsibilities and transparent rules for review and approval.

Employees also need to understand the application and be able to assess its outputs in the context of their work. Training and changes to workflows should therefore be included in the planning from the outset.

Financial management and organisational development come together here. An application that works technically also needs a working environment in which it can be used effectively and developed further.

Managing AI for economic value

In our view, effective management of AI connects strategic objectives with specific use cases, transparent costs and an organisation that puts improvements into practice.

The finance function can provide direction. It makes assumptions explicit, helps set priorities and supports a clear distinction between actual results and expectations. This gives organisations a sound basis for developing AI where it can make a difference and showing what it contributes economically.

This is where ExCaMa Advisory works, bringing together strategy, financial management and organisational implementation.

Explore Advisory All Perspectives

Contact

Let’s explore the possibilities together.

Whether you are looking for strategic advice, support with international development or professional training, talk to us about your priorities and possible next steps.

Start a conversation