When to Use Each Design Pattern

Match your use case characteristics to the right pattern before building. Every row below maps to a real, sourced Workato pattern.

Design Patterns Strategy
Data table
CharacteristicPrimary patternSecondary consideration
Fixed, pre-planned skill sequenceDeterministic FlowHuman review of step input/output
Each skill's output decides the next stepGuided Skill ChainingComplex branching may apply
Variable input schema per request typeDynamic Input BuilderDeterministic Flow for the surrounding steps
Query or aggregation can't be predeterminedQuery GenerationControlled Data Volume for large results
Centralized intake from multiple event sourcesEvent RouterGenie Handover for mid-conversation routing
Need domain data persisted across sessionsPersistent State & Data EnrichmentRecord System if acting as system of record
Combine multiple sources into one responseInformation SummarizationNever aggregate via knowledge-base search
Dataset too large for context windowControlled Data VolumePagination, pre-aggregation in the skill
Genie needs to act as a lightweight system of recordRecord SystemPersistent State for scoring/enrichment
External event should start/continue a conversationApp Event-TriggeredHuman-in-the-Loop before persisting results
Genie needs to be called from Claude/ChatGPT/custom UIGenie as MCP ServerTool Use discipline for exposed tools
Multiple independent items to process concurrentlyParallelizationRepeat for each + Assign Task to Genie
Task spans multiple specialist domainsMulti-Agent CollaborationOrchestrator-Subagent for hierarchical decomposition
Irreversible or high-value actionsHuman-in-the-LoopSkill Call Approvals vs. Business Approvals
Agent needs context across sessions/clientsMemory-AugmentedData retention policy required
Complex tool selection with many optionsTool UseAvoid One-to-one Mirroring / Overloaded tools

Source: Workato Docs: Genie design patterns — decision matrix · Agentic Design Cookbook V2.0, p. 9

Put these cookbook patterns to work

Get started