News / Artificial Intelligence / Einstein

Breaking News: Salesforce Einstein GPT Officially Announced

By Lucy Mazalon

Salesforce is launching Einstein GPT to deliver auto-generated AI content, right within the Salesforce platform, with use cases including Slack, sales, service, marketing, commerce, and app builders. This is the world’s most powerful, intelligent, generative CRM – making every employee more productive and every customer experience better.

Marc Benioff, and Salesforce more broadly, announced the teaser via Twitter on February 8, 2023, including a few cryptic clues on what to expect. Now, Einstein GPT is in the process of being demystified, with the official news announcement.

What Is Einstein GPT?

Einstein GPT combines public and private AI models with CRM data so that users can ask natural-language prompts (i.e. conversational questions) directly within Salesforce CRM. The result? AI-generated content is delivered that continuously adapts to changing customer information and needs – hugely time-saving for users.

What do we mean by public and private AI models? Einstein GPT infuses:

  • Salesforce’s private AI models: In other words, those developed by Salesforce, which are leveraged by customers using features from the Einstein technology layer.
  • Combining OpenAI’s ChatGPT: Thanks to the Salesforce and OpenAI partnership, providing out-of-the-box generative AI capabilities to companies through direct access to OpenAI’s enterprise-grade ChatGPT technology.
  • Bring your own model: As Einstein GPT is open and extensible, you can opt to use it alongside your own external model/s.

Data is being pumped into these models, in real-time, from Salesforce ‘clouds’, and the Salesforce Data Cloud (Genie) to ingest, harmonize, and unify all of a company’s customer data.

Einstein GPT for Service

Service agents are kept on their toes, with faster case resolution (while maintaining customer satisfaction) as their main goal.

Einstein GPT for Service can be used to:

  • Auto-generate personalized agent chat replies.
  • Auto-generate a natural language summary during case wrap-up.
  • Generate knowledge articles from past case notes.

Let’s see these in action.

Here, you can see a customer has sent a chat message to the agent asking a question about the temperature rating of one of the garments. They don’t even need to think of the answer, as the Einstein Replies component (powered by GPT) has generated a response – including the message being phrased as you’d expect from a human response. The response is contextual; in this case, the information has been surfaced from Commerce Cloud, identified by the clear-to-see tag. Agents have actions to either send the reply immediately or edit the message.

Furthermore, the customer asks another question, this time with the specifics of where they are going to use the garment. GPT generates another ready-to-use response, including information about the garment’s suitability for the location the customer is heading to, tapping into external climate data.

The Einstein Replies widget also weaves in next best actions, such as “add to cart”, to take further legwork out of a service agent’s day.

During case wrap-up, a natural language summary is automatically generated, so the agent doesn’t need to spend unnecessary time documenting the conversation.

In addition, Einstein asks the agent whether they would like to create a knowledge article based on the chat conversation, meaning that this information comes out of the case comments, and can be opened up to other agents and customers.

Einstein GPT for Sales

Getting sales teams to adopt Salesforce is a challenge many admins face. The friction is often caused by sales reps wanting to move fast to meet their quota, having to leave Salesforce to send customer communication or do prospecting research, and spending too much time finding information stored various parts of the CRM.

  • Auto-generate composing emails, scheduling meetings, and preparing for the next interaction.
  • Surfacing external news for prospect research.
  • Adding contacts not already in Salesforce.
  • Generating additional collaboration channels.

Let’s see this in action.

Say that a prospect account is transferred to a sales rep. They can open a sidebar with Einstein Assistant and start asking questions, such as: “write me an overview of [company name] and include recent news”. Einstein used a combination of CRM data and external data to generate the response, opened up a view of the Account record, and suggested next actions, for example: “update Account description”.

Within the news, Einstein identified that the prospect account is planning expansion to a new region (a positive signal for new opportunity to do business), and an action to explore this was created. Einstein generates a list of people involved in the expansion – one of who is already a contact in Salesforce, another from external data sources. The rep can create a contact record for the person not currently stored in Salesforce.

Another action allows the rep to ask Einstein to generate an email that’s personalized and relevant. This can be copied straight into the email builder. For those who are familiar with GPT’s capabilities, you will know that it can also adapt the tone of copy, such as making the email sound less formal.

One last thing. Let’s ask Einstein to invite the prospect to a private Slack channel where the rep can communicate with them in real time. Not only is the Slack channel automatically created, the link to join is hyperlinked in the email body, too.

You can see how Einstein GPT gives reps a competitive edge – to seek out prospects, compose, and amend email copy, and automatically add alternative ways to collaborate, all in a few clicks.

Einstein GPT for Marketing

Marketers have already been tapping into the power of GPT to write email and landing page copy, marketing plans, and advice for data breaches. Now, the power of GPT is brought into the Salesforce interface, so marketers can get those tedious tasks done even faster.

  • Dynamically generate personalized content to engage customers and prospects across email, mobile, web, and advertising.

Let’s see this in action.

The marketing manager needs to create a landing page for an upcoming event. The marketers ask Einstein GPT to create a landing page template and campaign messaging – which they then ask to be made shorter. Elements can be added to the landing page in one click.

Einstein GPT can even surface images that fit the description the marketer gives, plus titles and forms (including an action to amend the form fields).

These steps have resulted in an Experience site being generated. This can be launched from Salesforce directly, or opened in Experience Cloud to configure further.

Einstein GPT for Developers

ChatGPT quickly piqued the interest of technical Salesforce professionals, who got creative with the ways generative AI could enhance their work:

“It is staggering what ChatGPT can create with just a few prompts: Validation Rules, Apex Code, LWC, lyrics for a song, or even blog posts.” Ian Gotts, What Does ChatGPT AI Mean for the Salesforce Ecosystem?

Now, it’s confirmed that Einstein GPT will improve developer productivity with Salesforce Research’s proprietary Large Language Model by using an AI chat assistant to generate code and ask questions for languages like Apex.

Let’s see this in action.

In an Apex class (e.g. ContactController), we can instruct Einstein by starting to write a comment that describes what the method does. An automatically generated smart comment appears, suggesting how the comment should be completed – that’s some smart hinting!

We also want to limit the query to ten records, and have the query invocable by Lightning components. The method is generated by Einstein GPT, including the AuraEnabled decorator (so that it’s invocable), the fields requested, and a limit.

Another use case is test classes. Einstein GPT can automatically scaffold tests, as shown for the ContactController class in the image below. Again, describe using comments for what we want, then the user has the opportunity to review and tweak before deployment.

Einstein GPT for Slack

Slack is a collaboration platform that encourages speed. The ChatGPT app for Slack, built on the Slack platform, integrates ChatGPT’s powerful generative AI technology to provide:

  • Writing assistance directly in Slack.
  • Research tools, namely AI-powered customer insights in Slack like smart summaries of sales opportunities and background research on accounts.
  • Instant conversation summaries.

Let’s see this in action.

In the team meeting channel, you can ask Einstein GPT to identify top Opportunities, which takes not only CRM data into consideration, but also recent news about the economy.

In a private channel, for example, with prospects and customers, you can use Einstein GPT to answer their questions with additional context, faster.

Summary

Einstein GPT is the next generation of Einstein, Salesforce’s AI technology that currently delivers more than 200 billion AI-powered predictions every day across the Customer 360 (Salesforce’s portfolio of products). We’ve seen how Einstein GPT can enhance users’ work lives across the organization – from sales, service, and marketing teams, to app builders – and everyone who uses Slack.

Aside from the embedded technology announced today, Salesforce Ventures are launching a new Generative AI fund, with the intention to invest in high-potential startups, bolster the ecosystem, and spark the development of responsible, trusted, and generative AI.

Einstein GPT is certainly big news to land, and it will be interesting to watch how it’s adopted by this range of Salesforce personas.

The Author

Lucy Mazalon

Lucy is the Operations Director at Salesforce Ben. She is a 10x certified Marketing Champion and founder of The DRIP.

Comments:

    WH
    March 08, 2023 9:01 pm
    This raises questions about privacy. For instance I might be SalesForce's paying customer - and one assumes that the 'data pumping' includes SalesForce's paying customers' data within that 'pumping'. However, the end-user i.e. the customers who are customers of the SalesForce CRM user aren't being given a choice to whether their data should be used by Salesforce (for free) to boost Salesforce's machine-learning systems. Maybe there's some contractual thing that allows this... but it still raises queries about privacy that need to be clearly understood and dealt with if they are abusive of individuals' data or using it for profit. I'd also like to correct you and SalesForce, in that these systems are not Artifical Intelligence (AI), they are in fact machine-learning systems. The two are different, though obviously one title is catchier :-) Overall I consider this worrying. I can absolutely see benefits, if done correctly and respectfully... but that doesn't seem to be the case generally at the moment regards GPT engine use 'in the wild'. What this seems to be as a general rule is a manner to of having less human interaction, which - though the technology companies would have us believe otherwise - hasn't actually been proven to be beneficial to anybody except shareholders, who see larger returns due to less salaries. In terms of customer satisfaction - there is a downward curve on that graph that started in the mid 00s and has not risen since. With each generation of system that puts more technology between the customer and the person seeking help, reducing human interaction, the curve steepens a little more. Rushing to implement systems based on excitement for them rather than a genuine wish to make customers interactions better is a mistake imo. However, despite the negativity it may seem I am expressing, I can absolutely see genuine benefits to this also... i.e. apex code creation etc. It is the customer facing stuff that concerns me... as anybody whom has ever tried to get a straight answer out of Google or Microsoft will attest, when one of their systems develops a bug... as a customer it is almost impossible to get it solved in a manner that suits one's business needs. It is instead solved when it suits Google or Microsoft's business needs. That's not treating a customer with respect. I worry Salesforce may end up the same way (or even more that way) with the introduction of tools designed to remove more humans from human interactions.
    Petr
    March 09, 2023 10:04 am
    Hi, when and how could I test this feature? There is a lot of information about the feature, but I can´t find the information about HOW to use it or when and where it will be available. Thank you very much! Petr
    Guru
    May 10, 2023 2:44 am
    This is great article! Can you point the source of these GPT demo from Salesforce or these are something you tried and embedded? Curious to know more on that. Appreciate your response.

Leave a Reply