In many organizations, the AI strategy begins with a list of diverse use cases that are difficult to compare.
This episode of AI.impact demonstrates how to organize these ideas to establish clear priorities before any technical development begins.
A growing number of use cases that are difficult to organize
Let’s take the example of a large diversified conglomerate where discussions about AI emerged simultaneously across several business units. The teams quickly identified several opportunities: field assistance, legal support, sales decision-making support, email processing, and bid analysis.
In total, about ten use cases were identified. These cover very different scenarios, with varying impacts, constraints, and levels of maturity. While this diversity illustrates the potential of AI, it also makes decision-making complex. Comparing field support for technicians with a legal or sales tool amounts to comparing very different scenarios. These solutions do not share the same objectives, value propositions, or operational constraints.
This context comes at a pivotal moment, as generative AI begins to take off. Technologies are evolving rapidly, with capabilities improving week by week. But significant uncertainties remain, particularly regarding the quality of results and user adoption.
Plan before you start: the key role of methodology
Before any implementation takes place, the task here is to organize our thinking and align the teams.
An initial onboarding phase provides a shared framework for understanding AI, its capabilities, and its limitations. The goal is not to train experts, but to enable different departments to speak the same language.
The goal here is to review, elaborate on, and expand upon each existing use case. This process helps clarify their scope, objectives, and expected benefits.
The key element lies in the implementation of a common evaluation method that allows for the comparison of very different cases. Each case is analyzed along two main axes:
- the value generated (revenue, customer experience, employee experience, efficiency, risk management),
- speed of implementation (technical complexity, training requirements, change management)
Each criterion is rated on a common scale and then aggregated to provide an overview.
We can then plot the use cases on a visual representation that combines value and speed, with a third dimension: the affected population. This approach makes priorities immediately clear, helping to avoid decisions based solely on intuition.
Not every use case has to become a project
This analysis allows us to highlight a key point: not all of the identified use cases are intended to be launched.
Some have promising potential but require significant effort to implement. Others can be tested quickly but generate less value. Still others do not offer any significant differentiation from solutions already available on the market.
The process, therefore, is not merely about selecting the best cases, but also about eliminating those that are not relevant in the short term.
Two strategies then emerge:
- delve deeper into a specific use case to gain a better understanding
- explore several cases to understand the potential of AI in different contexts
This decision depends on the organization’s level of maturity and its objectives.
Conclusion
This framework helps structure the AI approach, prioritize business use cases, and provide guidance for proof-of-concept projects. An AI strategy cannot be guided solely by a solution. It must start with the use cases, be structured, evaluated, and prioritized.
So before you dive into AI projects, your ability to choose where to apply it remains the most critical factor!