Salesforce Inc. is expanding its generative artificial intelligence footprint today with new capabilities for its Einstein GPT general-purpose AI assistant that allow it to work with its Flow workflow automation suite.

The company is also tightening integration between its Data Cloud “lakehouse” and Flow. The integration of Einstein GPT and Flow makes it possible for users to can create and modify automated workflows using conversational instructions.

“You can give it a prompt such as ‘I want to create a guided workflow for a new customer or create a rule to follow up with an email to customers who haven’t responded in five days,’” said John Kucera, senior vice president of product management at Salesforce.

Users can also describe a formula they want and Einstein GPT will build it automatically. Or they can describe a function they need to locate and have Einstein GPT insert the correct flow automatically instead of requiring them to search for subflows and invocable actions, the company said.

Bring your own data

Perhaps most importantly, Einstein GPT can be trained with a company’s own data from its existing customer relationship management system and other sources. “We don’t expect customers to have to train this to get started fast,” Kucera said. The capability will be available on an early adopter basis with the winter 2024 release, which should be available in early fall.

“We want people to be comfortable that the AI works, so we want them to be able to check our work and that the output is as expected,” Kucera said.

Data Cloud for Flow links the company’s workflow automation with structured and unstructured data stored in a lakehouse, which is a cross between a data lake and a data warehouse. Customers can combine data from Salesforce’s databases and other sources to harmonize customer views and apply low-code tools to create automated processes and workflows.

“You’ve been able to use Flow with the Salesforce database data and now you can do it with Data Cloud data,” Kucera said. “This is the next generation of our [customer data platform]. In the past, you could get data out but you had to write a whole lot of code. We’re now bringing the richness of Data Cloud to non-technical users so they can automate based on triggers in data from Data Cloud and other places.”

For example, real-time data can now be used for dynamic pricing and financial service companies can automate fraud detection by flagging suspicious transactions based on patterns detected in transactional data, Salesforce.com said.

Kucera called the Data Cloud “our fastest-growing organic development. “People use [the Salesforce database] literally a trillion times a month,” he said. “We’re now adding those same capabilities to the lakehouse.”

Both features will be available as a pilot in the winter 2024 release. Kucera said Data Cloud integration will probably be offered as part of a standard Salesforce subscription. Einstein GPT for Flow may carry an additional charge, but details have not been determined.

Photo: Mark Albertson