Salesforce makes its Einstein Copilot generative AI assistant generally available – Business
Salesforce Inc. today announced that its customizable generative artificial intelligence conversational assistant Einstein Copilot is now generally available for every Salesforce application along with a few new capabilities that enhance its ability to execute actions automatically.
Einstein Copilot integrates with Salesforce and pulls data and metadata directly from enterprise corporate clouds, providing it the context to understand what users are doing so that it can “chat” with them. That allows users to ask questions and receive trustworthy answers grounded in their company data from within their cloud.
A fundamental component of the Copilot is its ability to take actions on behalf of users, or string together workflows. These Copilot Actions are pre-programmed capabilities and customizable workflows that companies can build that allow the AI to take actions such as using tools to complete tasks when asked such as sending messages, modifying data, updating contacts and other activities.
Jayesh Govindarajan, senior vice president of AI at Salesforce, told SiliconANGLE in an interview that the company has worked to build up Copilot’s capabilities with Actions since its beta availability release in February. He said the capability has been a big hit among Salesforce customers in particular.
“It comes out of the box with a set of actions,” Govindarajan said. “I have a Copilot engine if you will. How do I train it to be a good sales assistant? What data is needed? What actions need to be configured? Something that we baked in is the notion of a custom action. This is the ability of a customer to bring in any piece of code, whether it is their Salesforce workflow or an external API and then just register that as an action.”
In many cases, Govindarajan said, Copilot can learn to execute an action simply from clearly written documentation fed to it along with the code. From there it will know exactly when to call it and connect to the tools and even execute chains of events for the workflow.
“That’s been super-powerful, something customers use a lot,” he said. “We dogfooded much of this ourselves. We use our own software and that’s been quite interesting.”
Examples of new out-of-the-box Actions coming to Copilot affecting sellers include close plans that will help accelerate the time to close for sales reps and managers build personalized close plans for opportunities with step-by-step tactics. Copilot can provide forecast guidance when asked, “What deals are at risk?” to provide summarized lists of deals that might not hit a monthly or quarterly quota, including reasons why those deals were chosen. It can also automate follow-up emails to help facilitate deal progression based on prior calls and take some of the weight of remembering where a sales track was.
“At Salesforce we’ve worked to make work effortless and our teams working on AI this has been our goal to infuse AI into work to make it less of a chore,” Govindarajan said. “From that lens, I feel like Copilot is the coming together of these things. Because of what LLMs can do, we can do more than recommend an action but take that action on a customer’s behalf.”
Salesforce unveils Zero Copy Partner Network
Alongside the general availability of Einstein Copilot, Salesforce announced the Zero Copy Partner Network, which opens up a global ecosystem of providers for building secure data integrations for the Salesforce Data Cloud. That way, companies can access company data in AI systems as if it were native without needing to use application programming interfaces or pipelines.
Partnerships with Google LLC and Snowflake Inc. are generally available, and Amazon Web Services Inc. and Databricks Inc. are in pilot, expected to launch later this year to provide access to systems such as Amazon Redshift and Google BigQuery. Without Zero Copy, companies must rely on brittle data pipelines that must constantly move data and mirror databases to and from Salesforce to keep it up to date. With Zero Copy, data in enterprise clouds can be accessed where it lives and treated as if it’s natively available without the need for a custom-coded data pipeline.
Salesforce Data Cloud acts as a pool of a company’s metadata-enabled objects and unstructured data in one place, reducing information silos and data fragmentation for an enterprise for services and products such as Einstein Copilot. Data Cloud exists within the boundaries of Salesforce, but a considerable amount of enterprise data exists in enterprise clouds and needs to be copied over or moved to be used effectively.
“The challenge that many customers brought to us is that they don’t want to bring data over or maintain multiple iterations of that data across different systems to get value,” Tyler Carlson, senior vice president of strategic partnerships at Salesforce, told SiliconAngle. “So a core tenet of the Data Cloud platform from the beginning was being an open platform.”
Enabling Zero Copy from external sources means developers can query data directly or access the data through virtual files. It also means that when the data changes, it changes for every application “looking” in real time. There’s no need for a push notification a ping to happen or a replication request to update different databases.
Over the years Salesforce has built integrations into Data Cloud, and with the announcement today, it’s expanding its capabilities. They include new integrations for service providers with a data kit that will allow them to easily connect their data directly to Data Cloud, without needing to update their pipeline or data models.
“Many of our customers have very distributed data and lots of different data sets everywhere,” said Carson. “Bringing more and more partners and ISVs into this ecosystem means that more of that data can be mapped into Data Cloud and the Salesforce data model and show up effectively as native objects that you can take action on. You can use them in generative AI capabilities such as Einstein 1 Studio and reason about them without having to copy it first.”
As an example, an insurance company may want to ground generative AI with real-time weather data, which would be difficult based on traditional data pipelines since it would require API calls and analytics. Using Zero Copy, weather data from a partner could be accessed as if it were a native real-time data source within Data Cloud for the AI to create personalized emails for customers based on the weather in their city.