As Dreamforce kicks off today, Salesforce has expanded its generative AI offering with the announcement of a pilot this Fall of Einstein Copilot and Einstein Copilot Studio.
What’s the pitch for these? According to the formal release, Einstein Copilot is:
A new and trusted out-of-the-box conversational AI assistant built into the user experience of every Salesforce application. Einstein Copilot will drive productivity by assisting users within their flow of work, enabling them to ask questions in natural language, and receive relevant and trustworthy answers that are grounded in secure proprietary company data from Salesforce Data Cloud.
Meanwhile Einstein Copilot Studio is positioned as:
An easy new way for companies to customize Einstein Copilot with specific prompts, skills, and AI models to close sales deals faster, streamline customer service, auto-create websites based on personalized browsing history, or turn natural language prompts into code, among hundreds of other business tasks. Einstein Copilot Studio will also provide configurability to make Einstein Copilot available for use across other consumer-facing channels like websites to power real-time chat, Slack, WhatsApp, or SMS.
Expanding on the rationale for the new offerings, Clara Shih, CEO of Salesforce AI, says that it’s down to customers wanting more:
Our customers want the ability for their employees to come up with their own prompts and so we are launching Einstein Copilot, which is a conversational AI assistant for both companies and employees, as well as their customers, to securely and safely be able to access generative AI to do their jobs better, faster, more easily, to augment and amplify their their own abilities, their skills, their work, to be more efficient and be more productive. That's the mindset of Copilot.
It's going to be available to every Salesforce user across every cloud and it extends what we launched in March, which were specific features. Now it's going to be an open-ended assistant that employees and customers can simply chat with in natural language. But not only that, we've always taken a platform approach to everything we build because we have this incredible ecosystem where customers want to make things their own, and partners have tremendous ideas and creativity that they want to bring onto the platform. So we're also announcing Einstein Copilot Studio, which allows companies to do just that.
The builders are in
Einstein Copilot Studio will include three additional features. Prompt Builder is intended to enable users to build, test, and deploy generative AI prompts that match their company’s brand and communication style, but without the need for in-depth technical knowledge. Shih explains:
This is for customers who want to customize the prompt templates that have been included in Einstein GPT. So, for example, if you're using Service GPT chat reply recommendations and you want to make [them] more personalized to your brand, your brand voice, or if you want to make sure that it references specific database fields that you have in Data Cloud, now you can go to Prompt Builder and you can add those customizations. You can make it your own. You can test it, actually AB test any number of versions, and then roll it out to your entire service team. You can also use Prompt Builder to create your own new prompts, either to expose through the assistant or through buttons and Lightning web components. The power of our platform is to do this in a No Code way and democratize who is able to build an AI app.
Meanwhile Skills Builder is pitched as empowering companies to create custom AI-driven actions that complete specific tasks, such as, for example, competitive market analysis for meeting preparation. Shih says:
This means that Einstein Copilot is now no longer just accessing your data and answering questions on the data, but companies also have the ability to control and designate which workflows they want the Copilot to have access to. The great thing is customers already have so many of their business workflows or customer workflows on Salesforce. All they have to do in Skills Builder is go in and select which ones they want the Copilot to have access to.
So, for example, a simple workflow could be resetting a password, or initiating a return if you're a retailer. With one click, a company can make it possible for their employees and their customer service department, the commerce and sales department, or even the customers themselves to access that workflow, not through clicks and navigation, but just through an actual conversation.
Finally, Model Builder will allow users to select one of Salesforce's proprietary Large Language Models (LLMs) or integrate preferred predictive and generative partner AI models.
This all builds on Salesforce’s track record in AI that pre-dates the generative AI hype cycle of the past six months, according to Patrick Stokes, Salesforce SVP of Product Management:
We have ten years or so of innovation built on top of Einstein, largely for predictive intelligence and, in some cases, being able to go build those predictive models yourself on top of Salesforce. Just a few months ago, we announced something called Einstein GPT, which brought generative capabilities into Einstein and introduced the Einstein Trust Layer.
But the Einstein GPT capabilities were really kind of just our first foot in the door so to speak. What we did is we connected Large Language Models to data in Salesforce in some very discreet, very common ways. So in Sales Cloud, for example, the ability to generate a sales outbound email to a prospect, or in Service Cloud, the ability to generate a reply. We wanted to put our foot in the door there and see how well that would work and what kind of data we would need to connect into the prompt in order to really make it flexible enough to provide meaningful generation.
There are two important things that are new in today’s announcement, he concludes, beginning with the new conversational user interface:
Every Salesforce app now will have this incredible new conversational UI, right in the right rail of your app. This isn't just a button you click; this is a way to interact directly in a safe environment, grounded in all of your customer context and your page context with Einstein. We think that's really incredible. You can just ask questions. You can ask it to do things on your behalf.
The second is the platform itself with the ability to extend prompts and the ability to give Einstein new skills, he goes on:
In the context of Salesforce, for example, if you go to ChatGPT and say, 'Hey, can you tell me how to update a record inside of Salesforce?', it'll give you the right instructions on how to do it, but it can't actually do it for you. With Einstein Copilot, it can. It can take those actions for you. It gives us this new application paradigm for our Trailblazers to go and build all of this new capability and make it available to the everyday user, the millions and millions of users of Salesforce, right in the flow of their work. We think that is a huge leap forward from some of the more prescriptive GPT features that we rolled out just a few months ago.
It seems odd perhaps to be talking about expanded functionality only a few short months after the generative AI movement got underway in earnest, but such is the nature of the game.