Sequential Workflows

Introduction

A sequential workflow runs steps one after another. Simple, easy to reason about, easy to log. The starting point for most agent designs.

Beginner analogy: a checklist — do step 1, then 2, then 3.

Understanding the topic

Core concepts:

• Easy to implement, debug, eval.
• Total time = sum of step times.
• Best when each step depends on the previous result.
• Use as the inner loop inside more complex workflows.
• Most agents start sequential, add parallelism later.

Syntax reference

Visual workflow / architecture:

step1 ─► step2 ─► step3 ─► done

Real-world use

Most onboarding, KYC and form-filling agents are pure sequential workflows.

Best practices

• Start sequential; parallelise only when latency matters.
• Log timings per step.

Common mistakes

• Parallelising prematurely — debugging gets harder.

Hands-on exercise

Interview preparation — practice these questions:

• Q1. When pick sequential over parallel?
• Q2. What's the latency profile of a sequential workflow?
• Q3. Scenario: a 5-step sequential agent is too slow. How do you decide what to parallelise?