Salesforce launched Einstein Studio, which allows customers to bring their large language models (LLMs) from other services such as AWS' Amazon SageMaker, Google Cloud Vertex AI and other services via Salesforce Data Cloud.
With Einstein Studio, enterprises can bring their own LLMs to Salesforce Data Cloud without moving data. Salesforce's Einstein Studio, which is generally available, can use custom LLMs along with Einstein GPT's LLMs and remove the need for Extract Transform and Load (ETL).
The theme of model choice has become a key topic among enterprise buyers. AWS has also hit the bring your own model theme in recent days.
Salesforce Einstein Studio has a control panel to manage AI models, zero-ETL integration, connections to Salesforce data and no-code tools.
Here's what Einstein Studio means for customers, according to Constellation Research analysts Liz Miller and Andy Thurai.
Bring your own model approaches. Miller said:
"Einstein Studio is intended to be the point of model centralization and governance for a Salesforce customer. The goal is to be a center point and NOT the whole or only model. Larger enterprise customers were already developing models in everything from Sagemaker to Vertex. This move is in line with Salesforce's more partnership-forward and ecosystem friendly approach. Similar to how Data Cloud invites those lakehouse connections across Snowflake and Databricks, BigQuery and beyond, Einstein Studio is issuing an invitation for all AI models to find a spot in the Salesforce ecosystem."
What's in it for Salesforce? Thurai said:
"This move by Salesforce is rather brilliant. The problem with providing your own set of tools to create AI models is not easy. You have to constantly update the platform almost daily. For the true data scientists who are used to other tools, Salesforce provides an option such as AWS SageMaker and Google Vertex as well as other AI model training platforms to train their own models or use existing ones and fine-tune with custom data. The familiarity of the platform combined with the power of custom data makes this unique."
Model sprawl. Miller said:
"Most businesses are drowning in models that are under development and even more that have never made it out of development. Einstein Studio is Salesforce attempting to unlock AI investments faster by providing a fast, cost effective and readily available on-ramp for AI into business use cases."
Zero-ETL. "Zero-ETL means there will be no data movement. Data movement has been a major issue when it comes to model training," said Thurai. "Data needs to be moved around, managed, maintained, cataloged, secured and governed. Keeping the data where it is and creating the models where it is convenient is somewhat unique. Many vendors want you to create the model in their platform so you can be more tethered to them."
What's in it for customers? Miller said the bring-your-own-model approach from Salesforce means it's a vendor that's committed to moving with the market. Salesforce is moving to "deliver usable AI tools and solutions that bring AI models out of the development cycles and into view for many average business users."
"Right now, business users in sales, marketing and service are being promised a lot in this age of AI. But there is very limited runway to prove that these tools investments will yield real, profitable meaningful CX results," said Miller. "Salesforce is building customers glide paths to AI success."
More on Salesforce:
