Salesforce introduced its AI layer, dubbed Einstein, toptechtrends.com/2016/09/18/salesforce-turns-its-attention-to-artificial-intelligence-with-salesforce-einstein/”>in 2016. More recently, at the Salesforce toptechtrends.com/2023/05/10/salesforce-generative-ai-trust/”>World Tour event in NYC in May, all the company talked about was generative AI and toptechtrends.com/2022/09/20/salesforce-genie-completely-transforms-the-way-data-moves-on-the-platform/”>Data Cloud, its in-house data lake. Today, it announced the next step in that journey with the release of Einstein Studio and the ability to bring your own model (BYOM).
“We are launching ‘bring your own model’, which allows our customers to bring their proprietary data into Data Cloud to build and train their model,” Rahul Auradkar, EVP & GM of unified data services and Einstein told TechCrunch. When you bring your external model and mix it with the Salesforce data in Data Cloud, Auradkar says that it’s a powerful combination.
The solution is really aimed at folks who have fairly sophisticated data teams and have been building models in other places like SageMaker. These companies want to put those models they’ve already built and made a significant investment in, to work in other contexts. That’s what Einstein Studio enables them to do.
Einstein Studio is a management console that lives in Data Cloud and enables customers to import an existing model with zero ETL. That means the customer should be able to import the data without having to go through the painful exercise of extracting, transforming and loading it. That’s a big deal for data teams and should make the solution more attractive because of that.
For starters, it will support Amazon SageMaker out of the box, but Salesforce is also working on a pilot with Google Vertex AI with plans in the works to support Databricks, Snowflake and others down the road.
Image Credits: Salesforce
While Einstein comes with a number of predictive models like which customers are most likely to churn, this solution lets customers design customized predictive models to predict things like which products are most likely to need maintenance or making product recommendations based on a customer’s interest.
It can also work with LLMs to generate content like sending an automated email when the product is ready for maintenance before it breaks. Salesforce wants to reduce hallucinations, where the model makes stuff up when it doesn’t have a definitive answer, by connecting to a graph database based on data inside Salesforce. So the LLM can see all of the data related to a particular customer, giving the model the information it needs to write a more accurate email based on the information in the customer record.
Once you import the model, you can put into to workflows inside Salesforce and generate insights or trigger actions like creating an email, all while taking advantage of work your data team has already done.
Einstein Studio with the ability to connect to Amazon SageMaker and bring your own model is available in GA starting today.
About The Author
