Enabling the feature is a two-step procedure. Everything that decides whether reps see anything useful happens afterwards, in the OmniScript and its schema.
- Set up Einstein generative AI
Every suggestion is a model call, so generative AI has to be configured before Autofill has anything to ask. It is a prerequisite of the Setup page, not a step inside it.
- Flip the Setup toggle
From Setup, enter Einstein Autofill in the Quick Find box, select Einstein Autofill Setup, and turn on Einstein Autofill.
- Assign the full permission list
Reps need Omnistudio User, Prompt Template User and Access Einstein Autofill, plus Salesforce Voice or Live Message access. Access Einstein Autofill exists only for this feature, which is why it gets missed.
- Add the component to your OmniScript
Open the OmniScript that holds the form, add the Einstein Autofill Omniscript component, and configure the autofill elements from the properties panel.
- Write the schema and expose the form
Key the properties object by OmniScript field name, then create the Action Launcher action that opens the OmniScript. That action is how a rep reaches the form mid-conversation.
Gives Autofill a live call to read, for contact centres running voice through the console.
Gives Autofill a chat session to read, for teams whose intake happens in messaging.
- A property key that does not match the OmniScript field name produces no suggestion and no error. You find it by noticing an empty box.
- English (US) is the only supported language, so a pilot run in another locale is not a fair test.
- The output is generative, so plan the rollout as a review workflow rather than as an automation.