← Back to Workshop Home

4 Agentic Workflow Patterns

The building blocks for every AI automation. Click any node to explore — then decide which fits your team's processes.

⏱ Day 2 · Interactive Reference

Why This Matters for Leaders

When your tech team proposes an AI automation, they'll describe it using these 4 patterns. Understanding them helps you:

🎯

Ask Better Questions

"Is this a chaining or orchestration problem?" helps you assess complexity and timeline.

💰

Evaluate Cost vs Speed

Parallelization is faster but costs more tokens. Chaining is cheaper but slower. You decide the tradeoff.

🛡️

Set Governance Boundaries

Orchestration includes human-in-the-loop gates. You define where humans must approve.

📋

Identify Automation Candidates

Match your team's manual processes to the right pattern — that's your automation roadmap.

Pattern 1

Prompt Chaining

Sequential steps — the output of one becomes the input of the next. Like an assembly line where each station does one job well.

💡Click any node to explore that step, or press Auto-play to watch the full flow.
or click any node above

💪 Strengths

  • Easy to understand and debug
  • Each step testable independently
  • Preserves context across steps
  • Supports rollback and retry

🏦 Your Team's Use Cases

  • Invoice processing (PTP pipeline)
  • Monthly close report generation
  • Sequential approval workflows
  • Data extraction → validation → output
👔
Leader's lens: Chaining is your lowest-risk entry point. If your team has a process that's "do A, then B, then C" — that's a chain. Start here. Each step can be tested independently before connecting them.
Pattern 2

Parallelization

Multiple tasks run simultaneously on the same input. Like sending a document to 3 reviewers at once instead of waiting for each one sequentially.

💡Click any node to explore that step, or press Auto-play to watch 3 perspectives analyze simultaneously.
or click any node above

💪 Strengths

  • 3x faster than sequential
  • Multiple perspectives on same data
  • Each branch is independent
  • Aggregator resolves conflicts

🏦 Your Team's Use Cases

  • Merchant risk assessment (multi-perspective)
  • Due diligence checks (ID + Credit + Fraud)
  • Budget review from multiple departments
  • Vendor evaluation (cost + quality + risk)
👔
Leader's lens: Parallelization mirrors how your team already works — you send a proposal to Risk, Product, and Compliance simultaneously. The AI version does the same, but in seconds instead of days. The tradeoff: it uses more tokens (= higher cost) but delivers faster.
Pattern 3

Routing

Classify the input and direct it to the right processing path. Like a smart mailroom — each document type gets its own optimized workflow automatically.

💡Click any node to explore that step, or press Auto-play to watch a document get classified and routed.
or click any node above

💪 Strengths

  • Handles diverse input types automatically
  • Each path optimized for its type
  • Scales to many categories
  • Reduces manual triage effort

🏦 Your Team's Use Cases

  • Document classification (invoice / receipt / contract)
  • Support ticket triage by urgency
  • Payment type routing (card / QR / BNPL)
  • Regulatory filing by jurisdiction (MAS / BNM / OJK)
👔
Leader's lens: If your team spends time sorting and triaging before processing, that's a routing opportunity. The AI classifier replaces the "which pile does this go in?" decision — freeing your team to focus on the actual processing.
Pattern 4

Orchestration

A central orchestrator dynamically manages the workflow, spawning subtasks and coordinating human-in-the-loop decisions based on conditions.

💡Click any node to explore that step, or press Auto-play to watch a credit application flow through decision gates.
or click any node above

💪 Strengths

  • Handles complex conditional workflows
  • Spawns subtasks dynamically
  • Human-in-the-loop at decision gates
  • Adapts to unexpected situations

🏦 Your Team's Use Cases

  • Credit facility approvals (threshold-based)
  • Insurance claims processing
  • Complex compliance workflows
  • Any process with "if X then Y, else Z"
👔
Leader's lens: Orchestration is where you define the governance boundaries. The thresholds ($10K, $50K) are your decisions. AI handles the routine; humans handle the exceptions. This is the pattern for high-stakes processes where you need control.
Chain Parallel Route

Combined: PayLater Application

Most real workflows combine 2–3 patterns. This AnyCompany PayLater application uses routing → parallelization → chaining in one pipeline.

💡Click any node to explore, or press Auto-play to watch all 3 patterns work together.
or click any node above

🔗 Chaining steps

  • Risk Score → Decision → Notification
  • Sequential, context-preserving

⚡ Parallel steps

  • Identity + Credit + Fraud (simultaneous)
  • 3x faster than sequential

🔀 Routing step

  • Classify application type first
  • Different paths for new vs renewal

📐 Your next step

  • Which patterns fit YOUR process?
  • Fill in the Agent Design Canvas
👔
Leader's lens: Don't force everything into one pattern. Identify which pattern fits each part of your process, then combine them. When you design your agent in the afternoon exercise, label each step with its pattern — this becomes your brief to the tech team.

Quick Decision Guide

Match your process to the right pattern:

If your process looks like…Use this patternExample
Do A, then B, then C (sequential)ChainingInvoice: extract → validate → approve
Get 3 opinions on the same thingParallelizationRisk + Product + Compliance review
Sort first, then process differentlyRoutingClassify document type → right workflow
Complex with "if/then" decisionsOrchestrationCredit approval with threshold gates
Multiple of the above🧩 CombineRoute → parallel checks → chain to decision