Item: Make
Author: GILD

GILD verdict

Pick Make when business and operations teams need more logic than simple app-to-app automation, at a lower cost per step.

Settle the credit budget, the owners of shared scenarios, and the Enterprise controls you need before more teams build on it.

Make

Workflow Automation by Celonis

0 Mixed

Fit and limits

Who it fits

Workflows branch, loop over records, or reshape data. Routers, iterators, aggregators, and filters sit on the canvas, and routers and error handlers use no credits.
Business teams will build their own automations. Every plan includes unlimited users, and the Teams plan adds team roles and shared scenario templates.
Work runs across many SaaS apps. Make lists 3,500+ apps, plus HTTP modules and custom apps for any service with an API.
Data must stay in Europe. An organization can run in the EU1 or EU2 data zone, where most of Make's own AI processing also stays.

Where it falls short

Most module runs use a credit, trigger checks included, and routers, filters, and error handlers use none. A scenario polling every minute uses about 43,200 credits in a 30-day month before it processes any data (GILD calculation).
Audit logs, SSO, custom roles, 2FA enforcement, team credit limits, custom functions, and the on-premise agent are Enterprise only.
Make runs only as a hosted service. Some AI features, including Maia and OpenAI models in Make's AI Provider, always process in the US whatever the data zone.
The Make AI Agent (New) app has been in open beta since February 2, 2026, and Make's release notes through October 9, 2026 add features without a general release. Make states its functionality and pricing may change.

Size the credit tier on annual billing
Count module runs per scenario run, trigger included, and multiply by runs a month. For US buyers, Core starts at US$12 a month billed annually for 10,000 credits. Monthly billing costs about a third more than annual. Annual Pro and Teams plans allocate the year's credits up front. Core resets monthly even on annual billing.
Prefer webhooks to polling
A polling trigger uses a credit each time it checks, whether or not it finds data. Use instant triggers where the app offers them, and set schedules to the slowest interval the process allows.
Set error handling and owners before go-live
Store incomplete executions so Make retries rate-limit, connection, and timeout errors on a backoff schedule. Add error handlers to modules that write to systems of record. Assign an owner to each shared scenario. Saved versions last 60 days.
Decide where AI runs and who pays for tokens
Make's AI Provider bills tokens in credits on every plan, at a rate Make publishes for each model. A custom provider connection, on paid plans, bills operations only. Your provider bills the tokens, and its settings decide where the data is processed. Agent chat adds 1 credit for each operation of every tool it calls. Maia, Make's AI scenario builder, bills each reply from the same pool, and Make's own example reply used 140.04 credits.

Make publishes token rates per model but no typical agent run. Log credits per run for each agent for two weeks, and move agents over budget to a smaller model or your own provider.
Recheck Make's credit page at each renewal while Make AI Agent (New) stays in beta, and re-price agent scenarios whenever a model's tokens per credit change.
Maia is in public beta on paid plans; for the first month, count scenarios that non-engineers build with Maia and the credits its replies use.