Einstein GPT Announced at TrailblazerDX ’23

Einstein GPT was just announced at Trailblazer DX ’23. Yes, Salesforce has jumped on the generative AI train with “The world’s first generative AI for CRM.” Einstein GPT as presented in the keynote demo is a helpful assistant that composes emails, completes tasks, builds websites, and writes code for apps across the Customer 360 platform. The positioning makes sense and the opportunities for automating mundane tasks across the ecosystem do seem limitless, if Salesforce gets this right.


There was a strong emphasis on the fact that Einstein GPT learns from your CRM’s trusted customer data and has human controlled checks for accuracy. This is expected given the recent industry hype and related concerns about AIs going rogue and turning us all into paper clips. Salesforce is clearly taking a cautious approach to releasing generative AI functionality to the ecosystem. No release date was given. It sounded like we may initially start seeing functionality introduced with the Winter 2024 release–probably for select pilot customers.

Salesforce is building an open ecosystem around Einstein GPT, so you will be able to utilize models built with Salesforce’s private Einstein AI tools, integrate with OpenAI’s ChatGPT, or bring your own model from other platforms. The Salesforce and OpenAI partnership will enable out-of-the-box generative AI capabilities, while new $250 AI investment fund from Salesforce Ventures will focus on investing in new AI startups and already seems to have investments in You.com, Anthropic, Cohere, Hearth.AI.

The demos were impressive, but they were clearly a highly controlled set of use cases. It will be interesting to see how the variety of novel use cases develops over the next year.

The initial Einstein GPT services demonstrated covered a variety of functionality for Sales, Service, Marketing, Slack, Tableau, and Developers. I captured some screenshots and notes in the following sections.

Einstein GPT for Sales

The service seems quite good at generating text summaries of accounts, writing emails, and performing other basic tasks like scheduling meetings.

Create an account summary and update information to the account record
Compose an email

Einstein GPT for Service

There are always many routine and repetitive tasks involved in service interactions, so Einstein GPT should eventually have a field day with service use cases. The demos focused on what you’d expect such as creating personalized agent chat replies, case summaries, and knowledge articles based on existing customer and case information. I expect to see many more use cases for service in the near future!

Generate a Case Summary
Generate a knowledge article

Einstein GPT for Marketing

It was interesting to see Einstein GPT generating a landing page and submission form from text commands. It’s an example of a not so trivial task, one that might take a few hours, being automated and standardized. You can easily imagine composing emails, mobile, web, and display advertising campaigns in a similar fashion. This starts to show more of the real potential for time savings with more complex automations.

Given that this is happening within Lightning, I’m assuming that this is Sales Engagement (a.k.a. Pardot) and not Marketing Cloud. It would be great to start seeing Marketing Cloud more tightly integrated with the customer 360 ecosystem, so it can take advantage of Einstein GPT and platform features and functionality in general. I hear this is coming and would love to know when–I’ve been waiting for this since 2013!

Create a website landing page with text commands

Einstein GPT for Tableau

Einstein GPT also put on an impressive show in Tableau generating a chart and natural language explanation of the chart contents based on a simple question about product line performance in Europe.

Einstein GPT for Developers

Developers can generate basic code, like the contact controller class pictured below, using naturally language requests in their code comments.

Overall the demos showed lots of interesting use cases and I’m curious to know when this functionality will actually be released into the wild. I’m looking forward to taking Einstein GPT for a test drive–hopefully some day soon!