It’s one thing to have a basic chatbot answering simple questions; it’s another entirely to have a generative AI-powered platform that performs actions.
Salesforce is now expanding the availability of its Einstein Copilot, making it generally accessible. Alongside this rollout, the company is enhancing the platform with new Einstein Copilot Actions, designed to boost sales productivity through generative AI. Initially previewed at Salesforce's Dreamforce 2023 conference, Einstein Copilot transitioned to beta in February 2023, allowing more users to experience the technology.
A key focus of Einstein Copilot is its ability to connect with an organization's broader data ecosystem, going beyond just Salesforce's built-in data. As part of the general availability launch, Salesforce introduced its Zero Copy Partner Network, which supports various vendor technologies that utilize the open-source Apache Iceberg table format for data lakes.
“From our product launch, we’ve learned that the more complete the context, the better Einstein Copilot performs,” noted Jayesh Govindarajan, SVP of Salesforce AI.
Einstein Copilot Actions: Supercharging Productivity
Einstein Copilot offers a conversational AI interface for users to query customer relationship management (CRM) data and other connected sources. In today’s digital landscape, a conversational interface is essential for generative AI tools. However, Salesforce distinguishes itself by providing rich context and actionable capabilities. With Einstein Copilot Actions, users can initiate entire workflows that streamline sales processes and drive deals to closure.
Copilot Actions empower users to invoke any action that Einstein Copilot can execute, both within the Salesforce platform and beyond. The system can also decompose complex tasks into a series of actionable steps, including workflows, API calls, and custom macros.
Govindarajan highlighted that Einstein Copilot can handle a wide range of tasks, from simple to complex. For example, a straightforward task could involve retrieving specific data, while a comprehensive request might involve identifying optimal sales opportunities for a given day and drafting an email for those prospects.
This higher-order task goes beyond simple requests; it requires the system to understand the user's context, the nature of sales opportunities, and what constitutes the best opportunity in terms of closure likelihood and value.
Reasoning Capabilities of Einstein Copilot
To effectively manage complex tasks, Einstein Copilot employs advanced AI techniques. Salesforce has invested in developing planners to enhance its reasoning capabilities. Techniques such as sequential planning help break down tasks into logical steps.
Moreover, Salesforce utilizes chain-of-thought and density-of-thought reasoning methods, where the AI system processes prompts step-by-step to derive optimal outcomes. For ambiguous tasks, Einstein Copilot employs a reactive planning technique that initiates follow-up questions to better define the task at hand.
Einstein Copilot Analytics: Measuring Performance
To ensure continuous improvement, Salesforce introduces Copilot Analytics, a tool for monitoring how organizations utilize Einstein Copilot. This feature tracks user interactions, including the execution of higher-order tasks, conversations, and the resulting actions taken. Metrics include which tasks were successful, the effectiveness of prompts, and areas needing enhancement. These insights enable organizations to tailor prompts and models for an optimized Copilot experience.
Looking ahead, Govindarajan shared that Salesforce aims to enhance Einstein Copilot by developing smaller, more efficient generative AI models. “As this technology evolves, we anticipate significant performance and cost efficiencies by refining our models," he noted. "We're currently testing these concepts in our labs with promising results."
Ranking
