This has certainly been the year of ‘GPT’, kicked off by Einstein GPT being unveiled – and with the Salesforce AI Cloud roadmap including no less than 16 capabilities rolling out in the next few months, we’re now entering the ‘summer of AI’. AI Cloud, powered by Einstein GPT, delivers auto-generated AI content, right within the Salesforce platform. Just shy of one month ago, GPT-related announcements focused on two core Salesforce ‘clouds’: Sales and Service.
Following a set of oversubscribed pilots, Sales GPT and Service GPT are much anticipated, from all levels of the organization. Here’s an overview of what’s become generally available – right now.
Service Replies (Service GPT)
Auto-generate personalized responses real-time data sources, informed by your CRM data and other sources.
Work Summaries (Service GPT)
Create wrap-up summaries of service cases and customer engagements based on case data and customer history.
Sales Emails (Sales GPT)
Auto-generate personalized emails for every customer interaction, informed by your CRM data, with options such as “Introduce me (cold)” and “Follow Up on Prior Engagement”. Auto-generate emails inside Sales Cloud or through Gmail and Outlook, with just a single click.
Einstein GPT Trust Layer
The Einstein GPT Trust Layer aims to resolve concerns over adopting generative AI. With GPT technology taking the world by storm, it’s no surprise that this seemingly magic technology is raising concerns about where data is retained when it’s sent to an LLM (large language model).
- Zero Data Retention: The Einstein GPT Trust Layer prevents customer data from being stored outside Salesforce, meaning prompts and responses are not stored by third-party LLM providers, or used to train their model.
- Feedback Store: Feedback loops enable past success to improve the quality of prompts over time as users interact with GenAI capabilities. The Einstein GPT Trust Layer collects feedback data, including if generated responses were helpful or not, and whether, for example, a service agent ultimately accepted, rejected, or modified that response.
- Encrypted Communications: Prompts sent to an LLM, along with the responses sent back to Salesforce, are encrypted.
- Data Access Checks: Governs prompts by restricting outputs to only the data allowed by the user’s permissions according to the org’s data access policies – such as profiles, permission sets, roles, and sharing rules.
- Audit Trail: Securely logs all prompts, outputs, interactions, and feedback data so teams can benefit from generative AI while meeting their compliance needs.
Summary
With 68% of workers agreeing that GenAI will help them better serve customers, and with desk workers estimating it will save them an average of five hours each week, the race to embrace the power of GPT – while doing so responsibly, with the ‘human in the loop’ at every stage – has truly heated up.
These capabilities are the first in the Salesforce AI Cloud roadmap. With Sales GPT, Salesforce have taken Buyer Assistant (Sales Bots) into beta testing – plus, Sales Assistant and Call Summaries are expected to be generally available in October 2023.
With Service GPT, Knowledge Articles are in pilot, with Search Answers and Mobile Work Briefings entering pilot very soon.
