How to use automations in Speak AI
What are automations?
Automations let you build multi-step workflows that run automatically without manual intervention. Each automation is built on a visual canvas, where you connect a trigger, optional filter conditions, and one or more action steps into a sequence. You can use automations to run AI analysis on every new recording, translate files automatically, upload files from external sources, or chain these actions together.
Opening the canvas
Go to Automations in the sidebar
Click New automation to start from scratch, or click an existing automation to edit it
The canvas opens at
/automations/:idand shows a vertical stack of step cards that you configure one by one
Trigger step
Every automation starts with a trigger that defines what fires the workflow. Built-in triggers include:
Media analyzed in folder: fires when a file finishes analysis in a selected folder
Recording received: fires when an embeddable recorder submission is processed
Catch inbound webhook: fires when an external system POSTs data to your automation's unique inbound URL. See Triggering an automation with an inbound webhook for the full setup guide.
Filter step
Insert a Filter step anywhere in the canvas to add conditions that must be met before subsequent steps run. You can filter on built-in media fields or on your own custom extracted fields, for example, only continue if a speaker sentiment field equals "negative" or a custom field value contains a specific keyword.
Action steps
After the trigger (and any filters) you can add one or more action steps. Available actions include:
Magic Prompt (AI analysis): runs an AI Chat prompt against the media and optionally writes the result to a custom field
Translate: generates a translation of the transcript into your chosen language
Speak Upload: downloads a file from a URL and uploads it to Speak AI for transcription and analysis. See Automatically upload and transcribe files with the Speak Upload action for the full setup guide.
Inserting variables with the {{ }} picker
In prompt fields and in the Speak Upload action's URL field, click the { } button to open a searchable picker of payload fields captured by the trigger. Selecting a field inserts a {"{{"}trigger.payload.field_name{"{{"}}"}} token that is resolved at runtime. This lets you pass dynamic values, such as a caller's name from a webhook payload, directly into your prompt or upload URL.
Schedule mode
Instead of firing on an event, you can set an automation to run on a recurring schedule. Enable schedule mode on the canvas, then choose the frequency and time. Scheduled automations run the configured action steps across files in the selected folder at the specified interval.
Run history
Each automation records a log of every run. From the automation's canvas page, open the Run history drawer to see which runs succeeded, which failed, and the per-step output for each. You can also trigger a test run against a specific file to verify your automation before enabling it.
Questions about setting up automations? Send us a message and we can help you design the right workflow.