Chapter - Framing AI Use Cases

Supplementary chapter prepared for the BWXT Data Science Workforce Training Pilot.

Outline in development. This chapter is scaffolded from the maturity-model Business domain (Identify AI Use Cases, ROI Analysis, AI Risk Identification). The BWXT-specific scenarios, value figures, and risk policies should be filled in with the program's subject-matter experts. The conceptual outline below is ready to teach from.

About this chapter

Before any data is touched, someone has to ask the harder questions: Is this even an AI problem? Is it worth doing? What could go wrong? This is the Tier 1 Business capability in the maturity model, and it is what keeps teams from building impressive models that solve the wrong problem. Every data scientist should be able to reason about it; senior tiers own it.

Is this an AI problem?

Not everything needs machine learning. AI fits when:

• There is a pattern in data that is hard to write explicit rules for (recognizing a defect in an image).
• There are enough labeled examples to learn from.
• A prediction would change a decision or an action.

It does not fit when a simple rule, a lookup, or a spreadsheet already works. Part of the skill is honestly recommending not to use AI.

Defining the problem and its value

A good use case has a one-sentence problem statement, a measurable goal, and an honest value estimate.

• Problem statement: what decision are we improving, for whom?
• Success metric: how will we know it worked (tied to the evaluation metrics chapter)?
• ROI: what does the current process cost (time, scrap, missed defects), and what would an improvement be worth? (SME input: BWXT cost and throughput figures.)

Identifying risks early

Every AI solution carries risk. The maturity model expects awareness of conceptual risks at Tier 1, deepening to mitigation at Tier 4:

• Failure cost: what happens when the model is wrong? (A missed weld defect is a safety issue.)
• Data and bias: is the training data representative of real operating conditions?
• Trust and adoption: will inspectors trust and use it?
• (SME input: BWXT safety, compliance, and governance requirements.)

Communicating with stakeholders

A use case has to be explained to non-technical decision makers: the problem, the proposed approach, the expected value, and the risks — without jargon. This connects to the Communication domain and is practiced throughout the capstone.

What this chapter will cover

• A repeatable checklist for vetting a proposed AI use case.
• Worked BWXT scenarios that do and do not warrant machine learning. (SME input.)
• A simple ROI and risk template. (SME input.)