When to Use an AI Task versus a Code Task

Process Street gives you two powerful ways to automate logic and decisions inside your workflows: AI Tasks and Code Tasks.

Both can process data, make decisions, and generate outputs, but they work in very different ways. Understanding which one to use will help you design accurate and reliable workflows.

AI Tasks: Great for Reasoning and Language-Based Work

AI Tasks use large language models (LLMs) to understand, interpret, and generate text.

They’re best suited for subjective, flexible, or context-based work that requires reasoning rather than hard rules.

Use an AI Task when your workflow needs to:

• Summarize, extract, or interpret natural language
• Generate or classify text (emails, reports, descriptions, insights)
• Make a judgment call or pattern-based decision (e.g. sentiment, intent, risk rating)
• Fill gaps in unstructured data
• Automate repetitive writing or reading tasks

Because AI models are probabilistic, you may see slightly different outputs even with the same input. That’s expected.

Example use cases:

• Summarizing a customer feedback comment into a short insight
• Extracting key terms from a legal document
• Generating an onboarding email based on role and department
• Classifying a support ticket’s tone or urgency

A Code Task is a better fit if you need identical outputs every time.

Code Tasks: Great for Precision and Predictability

Code Tasks execute scripts in JavaScript using exact, deterministic logic. They’re best when accuracy must be 100% consistent every time a workflow runs.

Use a Code Task when your workflow needs to:

• Calculate, compare, or transform structured data
• Validate inputs or apply strict business rules
• Connect to APIs or external systems where output must not vary
• Handle numeric or boolean logic (“if this, then that”)
• Guarantee identical results for the same input

Example use cases:

• Checking if a due date is within the next 7 days
• Calculating tax or pricing logic
• Validating data before submission
• Sending data to an external system API

Choosing the Right Task Type

Design Tip

If the workflow’s success depends on exact, repeatable results, start with a Code Task.

If the workflow depends on understanding or generating text, start with an AI Task.

For complex workflows, combine them. For example, use an AI Task to interpret an answer and then a Code Task to validate or act on it.

Key Takeaways

• AI Tasks are probabilistic — perfect for flexible reasoning and natural language work.
• Code Tasks are deterministic — perfect for precise, rule-based logic.
• Use AI Tasks for interpretation, Code Tasks for enforcement.
• You can combine both to balance intelligence and control.

• Learn the 5 stages of process building in our Getting started guide
• Watch helpful videos and webinars on our YouTube channel
• Learn about Process AI
• Collect info with Forms
• Organize data with Data Sets
• Document information and policies with Pages
• Get inspired by our Blog