Evaluating the Business Value of AI Use Cases
Why Traditional ROI Falls Short
Not every AI use case requires an ROI that can be measured in immediate financial terms. But every use case needs a solid rationale explaining the strategic value it is intended to create and how that value can be assessed.
The Benefits of AI Are Not Always Reflected in Financial Results
Artificial intelligence is making an impact—but not always in ways that can be directly quantified in euros. In the McKinsey study “The State of AI in 2026,” 80 percent of respondents report higher individual productivity and 50 percent report better decision-making. By contrast, 37 percent see a positive contribution to their organization’s operating profit.
The data is based on self-assessments and relates to different levels of impact. Nevertheless, it highlights a fundamental problem: Improvements in day-to-day work are not automatically recognizable as business results and cannot easily be attributed to a single AI investment.
This raises a key question for CIOs and senior management: How can the business value of an AI use case be assessed if its benefits cannot be immediately measured in terms of a financial ROI?
The Focus on ROI When Using Artificial Intelligence Falls Short
Board members and executive management rightly expect a solid basis for decision-making regarding AI investments. The KPMG CEO Outlook 2026 shows just how high expectations have become: 77 percent of German CEOs surveyed view AI as a top investment priority. 94 percent expect a return on their AI investments within the next three years at the latest, or report that they have already achieved one.
For clearly defined efficiency initiatives, a financial assessment is useful. If an AI solution reduces processing times, prevents errors, or takes over manual tasks, the costs and benefits can often be quantified, at least approximately.
Other AI use cases, however, improve the quality of decision-making, ensure regulatory compliance, increase resilience, or lay the groundwork for new business models. Their economic impact may not become apparent until later, may be indirect, or may take the form of avoided losses. Forcing monetization too early, then, tends to produce false precision rather than decision-making certainty.
Conversely, “strategic relevance” must not be used as an excuse for vague goals. Even a non-financial value contribution must be described in concrete and verifiable terms.
The Business Value of AI: Understanding Value Case, Business Case, and ROI
Business value refers to the contribution an initiative makes toward a relevant business objective. This contribution can be financial—such as higher revenues or lower costs. However, it can also manifest as higher quality, a better customer experience, lower risks, faster decision-making, or greater strategic agility.
To systematically describe this concept, we use the term Value Case. This is not intended as a substitute for the business case, but rather as a clear combination of:
strategic goal,
targeted business results,
Mechanism of action of the AI solution,
appropriate supporting documentation and key performance indicators,
Responsibilities
and the resulting investment decision.
The business case expands on this impact logic to include costs, risks, feasibility, and alternatives. ROI is a financial metric within this assessment—it reflects the benefits that can be reliably monetized and attributed to the investment.
Not every relevant business value can be translated into an ROI right from the start. However, every AI use case should have a verifiable value proposition.
Strategic Relevance Requires a Concrete Result
Simply linking a use case to a strategic goal is not enough. Goals such as greater resilience, higher customer satisfaction, or greater innovation capacity are too abstract to assess the success of an individual AI use case. What matters is which specific business outcome is to be improved.
An example from our consulting practice illustrates what this translation might look like: An international manufacturer of electronic products must regularly review country-specific approvals, permitted radio frequencies, and electrical safety requirements.
The strategic goal is to ensure market access and delivery capability. The specific outcome: Relevant changes are to be identified earlier and more comprehensively, assigned to the affected products, and the necessary proof of compliance provided in a timely manner. If a change is overlooked or proof is provided too late, the goods cannot be imported into the target market in the event of a claim.
The value added therefore does not stem primarily from the time saved on searching. It lies, first and foremost, in ensuring a business-critical outcome.
What Specific Business Outcome Is This AI Use Case Intended to Improve?
Which key performance indicators are appropriate depends on the desired value contribution. Financial goals can be tracked using metrics such as revenue, costs, or profit. For operational goals, suitable metrics include lead time, error rate, quality, or capacity. When it comes to risk and resilience, relevant metrics may include control coverage, response time, or the number of changes processed on time.
In the regulatory example as well, business value is created on several levels:
Risk: Market access and supply capacity are better secured.
Operational Performance: More countries and sources can be reviewed in less time.
Capacity: According to an initial estimate, capacity equivalent to one full-time position could be freed up in the long term.
Growth: Faster country screening can speed up the selection of new target markets and identify market opportunities sooner.
These contributions should not be added together in isolation. The value case requires a primary business outcome. Additional effects complement the valuation but must not dilute the central value proposition.
From AI Solutions to Business Value
An AI model does not immediately generate business value. Impact is only realized once the solution is integrated into a relevant workflow, people actually use it, and its results meet defined quality standards.
In this example, the AI could identify new regulations and changes, presort documents, and suggest potential impacts on countries and products. Subject matter experts review the suggestions and decide on necessary actions.
Among other things, the following could be measured:
coverage of relevant countries and sources,
the time until the technical evaluation,
the accuracy of the assignment,
necessary manual corrections,
the number of documents provided on time
and the implementation of the revised workflow.
McKinsey reports that companies with a particularly high, self-reported AI value contribution are more likely to fundamentally redesign their workflows. In addition, they are about twice as likely to have established procedures for measuring impact. This is not causal proof of the success of a single use case. However, it underscores that process change and impact measurement go hand in hand.
The CIO Bridges the Gap Between Technology and Impact
The technical responsibility for achieving the desired business results and implementing the solution should lie with the respective business unit. However, the CIO plays a key role in ensuring that a hypothesis about potential benefits can be turned into a robust and scalable solution.
This includes the necessary data and integration foundations, a robust architecture, security, governance, and operations. It is equally important to look beyond individual use cases: Can data products, models, interfaces, or governance mechanisms be reused for other projects?
Business and technology must therefore jointly define goals, prerequisites, and evaluation criteria. The McKinsey Global Tech Agenda 2026 shows that 29 percent of respondents align their business and technology plans throughout the year. Among companies rated as particularly successful based on revenue and earnings growth, the figure is nearly half.
The technical go-live remains an important milestone. However, it does not yet prove business impact.
Evaluating AI Use Cases: Measuring Impact and Making Decisions
Before getting started, the initial situation, the desired outcome, the time frame, quality thresholds, and the testing schedule should be clearly defined. Equally important is the question of what decision will be made based on the results: Will the use case be adjusted, scaled, paused, or terminated?
In an early-stage test, the initial value may lie in verifying a critical assumption with a limited budget. It must then be clear which uncertainty is to be reduced and which investment decision this will enable.
Five questions can help with the evaluation:
What strategic goal does this AI use case support?
Which specific business outcome should change?
How is the AI solution supposed to contribute to this change?
What metrics and supporting documentation demonstrate progress and impact?
What decision is made if expectations are met or not met?
These questions do not lower the bar for AI investments. They make them more precise. Not every AI use case requires an ROI that can be measured immediately in financial terms. But every one needs a robust value case that links the strategic goal, the expected impact, the evidence, and the next decision.
Written by
Dr. Alexander Roth is Director of Data & AI at Arvato Systems. With over 15 years of experience, he and his team help companies maximize the value of data, automate business processes, and implement sustainable innovations to efficiently drive digital transformation.