3.1.1. Identifying High-Impact AI Use Cases
💡 First Principle: AI creates value where four conditions meet: a high-volume or high-value decision or task, a pattern that AI is good at (prediction, language, perception, generation), data that captures that pattern, and a clear link to a business outcome someone owns. Use cases found by scanning business functions for those conditions outperform use cases found by scanning AI capabilities for a home.
The exam guide names four business functions as typical sources of high-impact use cases. Each has recognizable patterns:
| Business function | Typical high-impact use cases | Capability mapped | Outcome metric |
|---|---|---|---|
| Customer operations | Intelligent routing, agent assist, self-service answers grounded in policy, automated case resolution by agents | Language understanding, RAG, agents | Average handle time, first-contact resolution, cost per contact, CSAT |
| Sales and marketing | Personalized content at scale, lead scoring, next-best-offer, proposal drafting | Generative AI, predictive ML | Conversion rate, revenue per customer, campaign cycle time |
| Research and development | Literature and patent summarization, design exploration, experiment prioritization | Generative AI, ML on experimental data | Time to insight, cycle time, number of viable candidates |
| Software development | Code generation and review, test generation, documentation, legacy code modernization | Generative AI for code | Developer throughput, lead time for changes, defect rate |
Other functions (finance, HR, legal, supply chain) follow the same logic: document-heavy work favors generative AI with retrieval; forecasting and scoring favor predictive ML; multi-step operational work across systems may favor agents.
Mapping capability to outcome is the discipline that turns a use case into a business case. Write the chain explicitly: capability (summarize claim documents) → task change (adjusters read summaries instead of full files) → operational metric (review time per claim falls 40%) → business outcome (claims cycle time falls from nine days to five; customer satisfaction rises; adjuster capacity increases without hiring). If any link is vague, the use case is not ready.
Useful sources of candidate use cases include process maps with known bottlenecks, high-volume manual work (especially reading, writing, and searching), customer pain points from feedback data, and decisions currently made with incomplete information. Business users close to the work usually know where the friction is; the strategist's job is to collect, frame, and screen their ideas consistently.
⚠️ Exam Trap: A use case is "high impact" because of its link to a measurable business outcome and its volume or value, not because it uses the most advanced AI. When asked which use case to pursue first, prefer the one with a clear outcome metric, available data, and a willing business owner over the one with the most impressive technology.
Reflection Question: A contact center handles 2 million calls a year, 35% of which concern order status. Map an AI capability to an operational metric and a business outcome for this use case, and name the business owner you would want accountable.
