Use Predict Jobs

A predict job enables an active model to write predictive outputs to a data model object (DMO). Create a streaming or batch job in Model Builder and use the model's outputs in Salesforce workflows.

1. In AI Models (formerly Einstein Studio), click a model to view its details.
2. From the Integrations tab, create a new predict job.
3. In Predict Job Builder, select the data space and DMO where you want to store the outputs. Then click Next.
4. Map the model schema fields to your variables. Then click Next.
   

   Note The inputs defined here initiate updates to your prediction when data on these fields change. To activate a predict job, you must also map all fields.
   1. To add a DMO for your inputs, click the object field and select Add an object.
   2. On Available Objects, select the object and its associated direct attributes.
   3. To add related attributes, select a default path to begin mapping your fields. Then click Select {"default path name"}.
   4. On Added Objects, review your objects.
   5. To remove an object that you added, click Remove.
   6. To see your mapped fields, click Back to Mapping.
5. To choose the predict job update type, select a streaming or batch job. Then click Next.
   

   Note

   Streaming jobs don't support externally stored data sources such as zero copy or Bring Your Own Lake (BYOL). Only batch jobs are supported for these object types. You must keep this in mind when choosing a data source.

   

   Note Streaming predictions automatically update when input data changes. This keeps your predictions up to date. Batch predictions must be manually refreshed.
6. Define which fields initiate prediction updates, and click Next.
7. Review your predict job settings.
8. If you make changes, update your settings. Then click Next.
9. Add a job name and description, and then click Create.
   After the job is created, you can find it in Model Details.

   

   Note

   A new DMO is created to store predictions for streaming jobs.