Salesforce Debuts Einstein Copilot Generative AI Assistant
Salesforce launched the Einstein Copilot generative AI conversational assistant at its Dreamforce conference this week. Einstein Copilot completes Salesforce’s resurrection of Einstein as a virtual assistant, integrating with its CRM platform and Salesforce Data Cloud to offer direct help through conversation across any connected platform. Salesforce also released the Einstein Copilot Studio for companies that want to customize their own generative AI assistants, enhanced with Einstein’s capabilities.
Salesforce kicked off its generative AI infusion with Einstein GPT in March, leveraging OpenAI’s large language models to compose emails and other content. Einstein Copilot rises to the next level of assistance by direct interaction with a user. The chatbot combines generative AI conversation with customer information from the integrated Salesforce Data Cloud to give personalized insights as well as automate sales, marketing, customer service, and other tasks.
The idea, evoked by the Copilot title, is to streamline workflows by proactively suggesting relevant actions and generating useful content. For example, Copilot can automatically prep for meetings by researching accounts, then summarize sales calls and highlight key moments before drafting follow-up emails based on the discussion. Einstein Copilot can also respond to service inquiries with consistent answers across channels, explain individual cases, and then fill out field work orders as needed.
“The reality is every company will undergo an AI transformation to increase productivity, drive efficiency, and deliver incredible customer and employee experiences,” Salesforce CEO Marc Benioff said. “With Einstein Copilot and Data Cloud we’re making it easy to create powerful AI assistants and infuse trusted AI into the flow of work across every job, business, and industry. In this new world, everyone can now be an Einstein.”
Alongside Copilot, Salesforce introduced Einstein Copilot Studio for customizing AI assistants. As Einstein Copilot builds on Einstein GPT, the Copilot Studio extends the Einstein Studio platform that Salesforce revealed in August. Einstein Studio enables companies to import and train their AI models on proprietary data within the Salesforce platform. Einstein Copilot Studio goes beyond by letting companies create custom AI assistants from scratch. That includes a prompt builder tool for creating branded conversational interactions without coding and a skills builder to automate some customer service, write code, design websites, and carry out other enterprise needs. Einstein Copilot Studio also helps export Einstien Copilot to Slack, WhatsApp, and other channels.
Copilot and Copilot Studio will operate within the Einstein Trust Layer, a secure architecture for generating accurate AI while preserving privacy and data security. It all passes through the Einstein Trust Layer, a moderation and control system that monitors and works to maintain business and regulatory compliance rules by stopping the models from holding onto any private or sensitive information. The Einstein Trust Layer watches for when a user might send sensitive data on an app and stops it from getting to the model, alerting the user that they might not want to submit that information. Einstein also watches for and filters out toxic and discriminatory language coming from both ends.
Einstein Copilot and Copilot Studio are set to arrive as a pilot in the fall. Salesforce envisions Copilot becoming a ubiquitous assistant for knowledge workers, predicting needs and efficiently handling repetitive tasks. The company calls it “the new interface for the digital HQ.” Salesforce’s approach with Copilot and Studio is to make every aspect of a customer interaction incorporate generative AI as a native element as opposed to simply adding a generative AI chatbot to the existing platform. The company is keen to continue broadening its generative AI options. Salesforce now hosts a range of AI models and has begun investing $500 million in related startups. It also supports models built in-house by Salesforce for business automation and writing code.