Pattern Selection Guide
How to characterize your use case, select a primary pattern, and identify complementary patterns before you build.
The Pattern Selection Process
Selecting the right design pattern(s) for your enterprise agent is a three-step process: understand your use case characteristics, identify the primary pattern, then identify any complementary patterns the use case requires.
Step 1: Characterize Your Use Case
Answer these questions about your use case before looking at patterns:
- How many steps does the task take? (1–2, 3–5, 5+)
- Are the steps sequential or can some happen in parallel?
- Does the agent need to discover information before it knows what to do?
- How high-stakes is the output? (conversational, significant, irreversible)
- Does the agent need memory of past interactions?
- Will a single agent handle this or do you need specialists?
- What's the latency requirement? (real-time, minutes, hours)
Step 2: Select the Primary Pattern
Based on the most prominent characteristic of your use case, identify the primary pattern. The primary pattern governs the agent's core reasoning or execution approach. Use the When to Use Each Pattern table as your reference.
Step 3: Add Complementary Patterns
Most use cases require 1–2 complementary patterns that address specific requirements the primary pattern doesn't cover. Tool Use is almost always required — it governs how the agent interacts with systems. Human-in-the-Loop is required whenever the agent takes consequential actions.
Documenting Your Pattern Selection
Include your pattern selection in the agent design document produced in Phase 2 of the Implementation Journey. Document which patterns you chose, why, and what use case characteristics drove the decision. This documentation becomes essential when the agent needs to be modified or debugged after launch.