1 / 7
Mapping where AI could insert in a process
In this lesson
Mapping where AI could insert in a process
Classify at least four of the six AI-insertion patterns (draft, summarize, classify, extract, generate, suggest) by their fit profile, distinguishing tasks suited to each pattern based on input structure, output…
You'll be able to
- Classify at least four of the six AI-insertion patterns (draft, summarize, classify, extract, generate, suggest) by their fit profile, distinguishing tasks suited to each pattern based on input structure, output determinism, and human judgment requirements [^1][^3].
- Apply the Mollick pedagogical-role framework to map where an AI can serve as mentor, tutor, coach, teammate, student, simulator, or tool within a business process, selecting the role that matches the process step's learning or delegation objective [^1].
- Evaluate a candidate process step for AI insertion using Anthropic's 4D criteria (Delegation, Description, Discernment, Diligence), determining whether the step's context can be described clearly, whether outputs require human discernment, and whether the task warrants delegation [^2].
- Create a process-insertion map for a multi-step workflow (such as data classification, anomaly detection, or content generation) that assigns specific AI patterns to appropriate steps, justifies each assignment with reference to AWS Bedrock use-case categories (monitoring, classification, generation), and identifies handoff points where human judgment remains essential [^3][^7].
- Explain why the same foundation model produces different outcomes depending on the assigned role or prompt frame, citing evidence that minimal perturbations in instruction or context can shift model behavior from aligned to misaligned outputs, and why this variability requires explicit role design rather than assuming a single "correct" AI behavior [^1][^4].