The success of AI in any organization hinges on carefully choosing the areas where it can deliver the most value and have the highest chance of making it to production. Leaders must be deliberate in selecting use cases with a clear understanding of AI’s limitations and the specific areas within their business where it can make a meaningful impact.Â
By taking a pragmatic approach to use case selection, you can mitigate risks and position your organization to capitalize on emerging technologies.Â
In this blog, we’ll explore the phData approach to AI in the enterprise and our framework for identifying use cases with the highest probability of success.
What Are the Risks of Poor AI Use Case Selection?
Selecting use cases that aren’t the right fit for your organization can lead to wasted resources, missed opportunities, and potentially damaging outcomes for your business. An ill-conceived AI initiative can drain time and money without yielding any tangible benefits, or worse, can lead to operational disruptions and reputational damage.Â
On the flip side, failing to invest in AI or being overly cautious could result in falling behind competitors who are leveraging technology to enhance their operations and customer experiences.
Why Selecting AI Use Cases Is Difficult
Choosing the right use case is inherently challenging. AI’s capabilities are vast and varied, but its limitations can be unclear at times. Understanding what AI can realistically achieve, especially as new technologies like Generative AI (GenAI) expand the scope of possibilities, requires a solid grasp of the technology, specific business needs, and capabilities of internal IT teams.
Without this understanding, there’s a risk of pursuing AI projects that either don’t align with your strategic goals or aren’t technically feasible.
phData’s Approach to AI Use Case Selection
To guide the selection process, we recommend a structured approach that filters potential AI use cases through a funnel, narrowing options based on value, cost, speed of return, and risk. However, before this filtering begins, the first step is to gather and vet potential use cases. This initial stage focuses on identifying where AI can make a difference within the organization.Â
The following section outlines how to effectively source and assess potential use cases to ensure they align with organizational needs and objectives.
1. How to Collect AI Use Cases
We recommend collaborating across the organization to identify pain points and innovation opportunities and prioritizing insights from those closest to daily operations. Involving a diverse range of stakeholders helps ensure that various potential applications are considered. This feedback, even if not immediately solvable by AI, helps assess the organization’s needs and encourages thoughtful, strategic planning.
Once use cases are sourced, the next step is to collaborate with IT and AI subject matter experts to determine if they are solvable with AI.
2. Filtering by AI-Relevant Use Cases
The next critical step is determining if a use case is suitable for AI. AI excels in areas like pattern recognition, generating text and images, and automating routine tasks, but is far from a panacea. Traditional machine learning (ML) works well for tasks that involve patterns and have sufficient training data. Tasks such as image classification, trend forecasting, and operational optimization are good candidates for ML.
ML thrives when there’s a structured problem and enough historical data to learn from.
Generative AI (GenAI) brings a different strength, excelling in creative tasks like generating text, images, or even code. It’s very useful when the goal is to produce new content or personalize experiences. However, GenAI may struggle with tasks requiring precise, error-free outputs, making it less suitable for environments where precision is critical.
Analytics, by contrast, remains invaluable for deriving insights from data, especially when predictive models aren’t necessary. For tasks like understanding business performance or uncovering correlations, traditional analytics methods—such as statistical analysis and data visualization—are often more straightforward and cost-effective.
When evaluating use cases, it’s important to align the problem with the right technology. Determine whether the challenge involves generating new content, predicting trends, or analyzing past data before trying to solve the problem with AI.
3. Value
Once you have a list of potential AI use cases, the next step is to evaluate their value. This involves estimating the direct impact each use case could have on your business, whether through cost reductions, increased sales, or other measurable outcomes.Â
We encourage clients to opt for easier-to-measure AI applications first instead of amorphous initiatives that don’t provide a clear and measurable outcome.Â
For instance, we partnered with a global financial services company to implement an AI-powered chatbot that dramatically reduced the time it took to respond to contract inquiries. This solution cut response times from days to seconds, leading to over a 70% improvement in efficiency and saving the company more than $400K annually in labor costs.
By focusing on high-impact, measurable outcomes like these, businesses can see immediate returns on their AI investments.
4. Speed of Delivery
Estimating how long it will take to implement each use case and seeing returns is important. We recommend first focusing on achieving short-term wins to build momentum and demonstrate early success. This incremental approach helps avoid the risks associated with ambitious, long-term projects that might not deliver results for years.
Such projects should be reserved for when a mature technical team is in place and a use case that aligns with the organization’s long-term strategy.
Additionally, adopting an agile methodology can enhance both the speed of delivery and the quality of your AI solutions. By using iterative development cycles, you can quickly ship initial versions of your AI solutions, gather feedback, and make necessary adjustments. This approach accelerates the delivery of value and allows for refinement of the solution based on end-to-end interactions.
5. Risk
Evaluate the risks associated with each use case. AI projects can fail for various reasons, such as poor data quality, inadequate model performance, or scalability issues. Consider the potential negative impacts if the AI were to fail.    Â
Begin by identifying the types of risks specific to your use case. Ask questions like:
What’s the criticality of the task the AI is handling?
How tolerant is the system to errors?
The level of risk varies with the application—an AI used in medical diagnosis carries far greater risk in the event of failure than an AI sorting spam emails.
Before abandoning use cases based on worst-case scenarios, consider the mitigation strategies available. Can the risks be reduced via redundant systems or human-in-the-loop systems? However, if the risks are too high relative to the value, it’s best to modify or drop the use case.
6. Cost
Finally, the costs involved in developing and maintaining the AI solution must be considered. It’s essential to include both fixed expenses—such as initial development, infrastructure setup, and acquisition of specialized tools—and ongoing costs, including data labeling, model updates, and continuous support from engineers.
While some use cases may offer high value, they may also come with significant fixed and ongoing expenses.
The initial investments in building the tools, skills, and infrastructure can lead to greater efficiencies over time. A well-designed system can be leveraged not only to deliver the current AI use case but also to support the development of future solutions, improving the efficiency of delivery as your capabilities expand.
Without a well-designed AI strategy, there are also potential costs if the chosen AI techniques or technologies prove unsuitable, which might require additional training, retooling, or even pivoting to entirely different solutions. Be sure to account for potential unexpected costs and include a buffer in your budget and timeline projections.
Proof-of-concepts are an excellent way to demonstrate the value of a solution without the heavy investment.
Conclusion
Choosing the right AI use case is critical to ensuring your strategy delivers value to your business. Evaluating use cases through the lenses of value, cost, speed of return, and risk allows for more informed decisions that boost the chances of success.
A well-planned approach ensures that your AI investments meet immediate business needs and set the stage for future growth and innovation.
Ready to take the next step? Wherever you are in your AI journey, phData is here to help!Â
Free Generative AI Workshop
Looking for guidance on how to best leverage the newest AI & ML capabilities? Attend one of our free generative AI workshops for advice, best practices, and answers.
