Salesforce outlined their strategy around Generative AI at their TrailBlazer DX developer event, which includes new product and partnering and ecosystem strategy.
At their TrailBlazer DX event in San Francisco this week, Salesforce made a series of announcements, all around the theme of Generative AI, which is specifically focused around the generation of content. They announced a new AI offering, Einstein GPT, a bot which resembles ChatGPT, which came out last year, but which is targeted at Salesforce’s own CRM and its other clouds. They also announced that they are working with OpenAI, which created ChatGPT, to add ChatGPT to its Slack collaboration software. The OpenAI partnership is part of a planned open ecosystem around this, and it will be followed by others. A new $250 million AI investment fund is also part of this broader strategy.
Salesforce views generative AI as the logical next step in the evolution of AI.
“AI is fundamentally changing the world,” said Sanjina Parulekar, Senior Director of Product Marketing at Salesforce. “The door is open for the next level of innovation in the enterprise. We first announced Einstein in 2016. This was the first generation of AI for CRM. It made a once unrelatable topic like AI extremely relatable for every single business.”
The introduction of Salesforce’s Data Cloud was the next step in this journey.
“AI is only as good as the data used to train it,” Parulekar said. “We announced Data Cloud during Dreamforce in September 2022 and it is one of our most significant innovations in Salesforce history, which already processes 100 billion customer records every single day. Data Cloud is the fuel behind AI and the engine is Einstein.”
Einstein GPT is the next stage of this process.
“Customer engagement is the next frontier for generative AI,” said Clara Shih, EVP and GM of Service Cloud. “Generative AI is changing the world. It will allow all businesses to completely reimagine how they interact with their customers.”
Einstein GPT is the world’s first generative AI purpose-built for CRM, Shih stressed.
“It combines Salesforce’s proprietary models with vetted external generative AI, and is being inserted into every Salesforce cloud as well as Mulesoft, Tableau and Slack,” she said.
Shih also emphasized that this is an open and extensible system.
“It includes our initial launch partner, OpenAI, so customers can use their own AI model or choose one out of the box. One of the best things of this is that OpenAI is a customer of Salesforce CRM, and they are also a customer of Slack. One of the products we are announcing, GPT for Slack, was built by the OpenAI team on the Slack platform. They are the initial launch partner this week, but there will be others that follow.”
Shih noted that Salesforce is well aware that Generative AI is not without its risks. In a well-publicized recent incident, ChatGPT in Microsoft Bing produced a bot which alternated between depressed and monomaniacal tendencies, and made HAL from 2001 look rather mellow and benign.
“We are addressing this with a trust-first, ethical design approach,” Shih said. “A human is also designed in the loop to have the final say.”
Five of these Einstein GPT offerings were announced at launch.
“Einstein GPT for Service lets service agents autodraft relevant tailored responses to reduce the average handle time and get the answer right the first time,” Shih said. “It also turns agent case notes into drafts of new knowledge articles, something that typically takes 10 days now.”
The other four new products announced were Einstein GPT for Sales, Einstein GPT for Marketing, Einstein GPT for Slack and Einstein GPT for Developers.
Jayesh Govindarajon, Salesforce’s SVP of Engineering, Einstein and Bots, provided more detail into how this all works.
“Generative context is contextual, and personalized, and made truthful from accurate data and having an expert human in the loop,” he said. “Edits and refinements go back into reinforcing the model, so that the more you use it, the better it gets.” He also noted that customers will be able to train AI with their own data.
“It can also help you create code by training on your code base,” said Stephan Chandler-Garcia, Lead Developer Advocate at Salesforce.
The final part of the announcement is a new Generative AI investment fund from Salesforce Ventures.
“It is $250 million to nurture the next generation of AI startups,” Shih said.
So when will all this be available to customers.
“It is still very early days, especially with our trust first, ethical design approach.” Shih indicated. “We have pilots across every single cloud, but GA isn’t imminent.”