How to Create AI Use Cases in the Public Sector Based on ROI.

Generative artificial intelligence has gone, in a very short time, from being a technological promise to becoming a tool with real impact on governments and public administrations. However, as interest in large language models (LLMs) and their capabilities grows, a key question is also being asked repeatedly by public officials and technical teams: where to begin so that AI generates real value and doesn't just remain another experiment?

The most common mistake is to directly search for “AI use cases,” as if there were a universal catalog applicable to any institution. This logic usually leads to pilot projects disconnected from strategy, difficult to justify budgetarily, and complex to scale. Experience shows that the effective path is different: stop looking for AI use cases and start designing business cases where AI is a means, not an end.

This article proposes a practical framework for building generative AI use cases based on return on investment (ROI), especially adapted to the public sector and organizations with a social mandate.

1. Change the right question: from “Where to use AI?” 1. “Where is value being lost today?”

When an organization begins by asking “What AI use cases can we implement?”, it often ends up replicating generic solutions: chatbots, automated summaries, or internal assistants with no clear impact. The result is a proliferation of initiatives that are difficult to prioritize and even more difficult to defend before oversight bodies.

The starting point should be the operational or strategic problem, not the technology:

• Where are the most administrative work hours being lost?
• Which processes generate the greatest delays or friction with citizens?
• Which tasks rely excessively on manual or unstructured knowledge?

Only after answering these questions does it make sense to evaluate whether generative AI is the right tool to intervene.

2. Define the expected impact before designing the solution

In the public sector, value is not measured solely in euros/dollars, but it must be explicit and measurable. A solid use case begins with a clear statement of the expected impact. Some examples of measurable public value include:

• Reduced response times for citizen procedures or requests.
• Decreased administrative errors or rework.
• Optimized use of specialized human resources.
• Improved traceability, transparency, or access to information.

Example of correct formulation: “Reduce the average response time to public information requests by 40% without increasing staff.” This clarity allows for defining metrics from the outset and prevents project evaluation from being subject to subjective perceptions.

3. ROI in Generative AI: A Pragmatic Approach

Although the public sector does not pursue direct economic benefit, ROI analysis remains a key tool for prioritizing and justifying decisions. The classic formula remains valid:

ROI = (Value generated – Total implementation cost) / Total implementation cost

What goes on each side of the equation?

• Value generated: savings in man-hours, reduction of outsourced costs, increased operational capacity, or risk mitigation.
• Costs: development, integration with existing systems, infrastructure, training, maintenance, and governance.

Not all value is directly monetizable. In these cases, it is advisable to complement the financial ROI with qualitative indicators, such as improved public trust, reduced legal risks, or increased institutional capacity.

4. Step-by-step method for building an ROI-based use case

A well-designed use case typically follows a clear sequence to reduce the likelihood of failed pilots and facilitate informed decision-making.

• Identify the priority challenge, aligned with strategic objectives.
• Define the key stakeholders (internal users, affected areas, legal and technology managers).
• Map the current process, identifying bottlenecks and points of friction.
• Formulate the use case in simple terms, describing what would change with AI.
• Estimate value and costs, even using conservative ranges.
• Evaluate feasibility, considering regulations, available data, and organizational maturity.
• Define success metrics and monitoring mechanisms. 

Examples applied to the public sector

1. Automation of responses to citizen inquiries
   1. Challenge: Saturation of service channels.
   2. Solution: Generic AI-based assistant for first-level response.
   3. Impact: Significant reduction in operational workload and response times.
   4. Value: Efficiency and improved citizen experience.

2. Generation of draft audit reports
   1. Challenge: Lengthy and highly manual processes.
   2. Solution: AI that structures and summarizes historical information.
   3. Impact: More reports in fewer times.
   4. Value: Greater control capacity with the same resources.

3. Simplification of regulatory language
   1. Challenge: Low comprehension of legal texts.
   2. Solution: Automatic translation into plain language.
   3. Impact: Fewer repetitive queries and greater access to information.
   4. Value: Difficult to monetize, but measurable in terms of satisfaction and use.

If you want to see more examples, we invite you to read our post: Seven Use Cases of Generative AI for Governments

Conclusion

The effective adoption of generative AI in the public sector does not depend on finding the “perfect use case,” but rather on building well-founded decisions based on real problems, clear metrics, and a coherent value narrative. When AI is integrated into concrete processes and evaluated using impact criteria, it ceases to be an abstract promise and becomes a real transformation tool. 

At Mottum Analytica, we help governments and public organizations identify, design, and deploy generative AI use cases with measurable impact, aligned with their mission, regulatory framework, and actual capabilities. If your organization is exploring AI and wants to move from interest to results, write to us at and let's discuss how to build use cases based on real value and ROI.