Choosing an AI Step

Pick the right AI Step for your task by comparing inputs, outputs, and tool access. Glow provides three dedicated AI Steps on the canvas plus Workflow Assistant for building automations.

At a glanceVerified in app · Oct 30, 2026

Use it to

Choose between Workflow Assistant, AI Prompt, AI Data Transform, and AI Agent based on inputs, output shape, and tool autonomy

Find it

Bottom dock → AI (AI Prompt, AI Agent) · Tools → Data → Change (AI Data Transform) · Bottom right → Assistant

Cost

Managed models bill Glow AI tokens per call (1 call + optional repair for AI Prompt, 1 per instruction for AI Data Transform, multi-turn for AI Agent)

Limits

AI Data Transform is capped at 3 positional instructions · AI Prompt makes a single answer call without tools or memory

Instead

Deterministic Steps (Conditions, Switch, Summarize, Change Fields) when rules and calculations are fixed

Use AI Prompt for an answer from supplied context. Use AI Agent when the Step needs to act through enabled tools.

Start here

Answer these in order and stop at the first yes.

Ask yourself	If yes
Am I trying to build the automation itself, rather than run something?	Workflow Assistant
Do I need up to three ordered values from one piece of text?	AI Data Transform
Does it have to act in another app, or decide its own next move?	AI Agent
Anything else	AI Prompt

AI Prompt

One question in, one answer out. Reference the upstream data in plain language and receive a single response.

• Best for: Classifying, summarizing, rewriting, translating, and extracting structured fields (Text, JSON, or validated Separate fields).
• Limits: Cannot call tools or act in external apps. Sees only the values you explicitly reference.

AI Agent

A goal in, and it works out the steps. State the objective and grant the tools it may use, subject to permissions, approvals, and runtime limits.

• Best for: Multi-step research, cross-app triage, calling subflows or MCP tools, and retaining conversation memory across runs.
• Limits: Execution path, latency, and cost vary between runs. Require tool approval or a downstream Human Review Step for consequential actions.

AI Data Transform

One text in, up to three ordered values out. Write one instruction per value and read them positionally as {{ N.result.0 }}, {{ N.result.1 }}, and {{ N.result.2 }}.

• Best for: Pulling up to three values from messy text concurrently without configuring a JSON schema.
• Limits: Capped at three instructions and positional text output. Find it under Tools → Data → Change. Use AI Prompt with Separate fields for four or more named fields.

Workflow Assistant

Describe the automation and it builds it. It places Steps, connects routes, and maps fields on the canvas rather than running inside an automation.

• Best for: Drafting a new automation, configuring field mappings, or explaining an inherited canvas while you work beside it.
• Limits: Produces a draft to review. Specify required inputs, destinations, rules, and approval points before testing.

The comparison, side by side

	AI Prompt	AI Agent	AI Data Transform
Model calls	One answer call; optional shape repair	Multiple calls within configured limits	One per instruction
Usage factors	Input, output and repair	Model, turns, tools and output	Input and output per instruction
Generated result	Can vary between runs	Answers and tool choices can vary	Values can vary between runs
Acts in your apps	No	Yes	No
Picks its model	No	Yes	No
Remembers	No	Optional	No
Main limit	One answer	Time and cost vary	Three instructions

Two habits worth having

Match autonomy to the task. Use a prompt for a generated answer, deterministic Steps for fixed rules, and an agent for tool selection. Supply the context each needs.

Test before you trust. Run the step on five awkward real cases: the email with no signature, the form where somebody typed “n/a”. Read the output rather than the green tick. See Try it before you trust it.

Verify it works

Run

Walk through the decision table above for your current automation step

You should see

You match the step to Workflow Assistant, AI Data Transform, AI Agent, or AI Prompt before configuring fields.

What’s Next?

• 👉 AI Prompt →: configure classification and summaries from supplied context.
• AI Agent: build multi-step autonomous logic with tools and memory.
• Models: keep the default or pick a Flash or Pro model once you know which step you need.