Make · Product

Make AI & AI Agents

AI steps inside deterministic workflows — plus tool-calling agents.

Make's AI surface has two halves: model providers as ordinary modules for classification, extraction and drafting, and AI Agents that decide which tools to call. Because both live on the same canvas, you can wrap an agent in deterministic guardrails instead of choosing between the two.

Explore Make AI →Full review

Disclosure: We may earn a commission if you purchase through links on this page. This never affects our reviews.

Features

🧠

Model modules

OpenAI, Anthropic and others as standard modules with normal mapping.

🤖

AI Agents

Tool-calling agents for triage and research where the path isn't fixed.

📑

Structured extraction

Turn unstructured email, PDFs and tickets into typed fields.

🚧

Guardrails

Keep money and compliance steps deterministic around the AI step.

Use cases

Classify inbound support tickets and route by intent and urgency
Extract invoice fields from PDFs into an accounting platform
Draft first-pass replies for a human to approve
Summarise long threads into CRM activity notes

Tutorials

Add AI triage to an existing support workflow safely

AI decides the label; deterministic modules decide the consequence.

Insert a model module after the ticket trigger and request a structured JSON label.
Validate the returned label against an allow-list before using it.
Route on the validated label with a router, not on raw model text.
Log every classification so you can audit accuracy later.
Escalate to a human path whenever confidence or validation fails.

FAQs

When should I use an agent instead of a scenario?+

Use deterministic scenarios for anything with money or compliance attached; use agents for judgement-heavy triage where a wrong turn is recoverable.

Do AI modules cost extra operations?+

Each AI module execution is an operation like any other, and model provider usage is billed separately by the provider.