Salesforce GenAI, data tools likely major Dreamforce themes
We read the tea leaves of Salesforce's summer product releases, executive interviews and Winter '25 release notes to figure out what's coming at Dreamforce.
Say hello to Agentforce, which appears to be one of the big -- if not the biggest -- splashes to be unveiled at Dreamforce next month.
Salesforce has confirmed that Agentforce will be connected to Salesforce's Einstein 1 AI platform, although final details have yet to be locked down.
From the looks of product releases and what Salesforce executives have hinted at this summer, Agentforce appears to be a gathering of sales and service generative AI bots sharing common customer data, underpinned by Salesforce Data Cloud.
CEO Marc Benioff has teased Agentforce features and demo videos on X (formerly called Twitter) since July 5, including Einstein Sales Agent, with promises of painting a fuller picture at the company's Dreamforce user conference Sept. 15-17. Einstein Sales Agent, a generative AI sales bot released Thursday, has features to onboard salespeople as well as generative AI coaching as it listens in on live calls or digital conversations.
Einstein Sales Agent follows Service Agent, released last month, and a number of other significant releases that together point to some sort unified platform of generative AI bots working together across teams, starting with sales and service. Generative AI agents like Salesforce's promise to bring a massive improvements for customer chatbots and agent assist tools because they fuse natural language processing and complex data sets instantaneously, said Liz Miller, principal analyst at Constellation Research.
"With these types of bots we're allowing people to converse in the way that [humans] converse, but we're also allowing the AI to have a bidirectional, real-time conversation that feels natural and organic," Miller said. "It doesn't feel like you're talking to a bot. You're like, 'Oh, you kept up with a very complex situation.' In the old times -- a year ago -- you wouldn't have that."
Sometimes it can be difficult to drive adoption of sales tech, especially among seasoned reps who already make their sales targets without the aid of AI. Because generative AI can handle administrative work such as updating records and managing initial, low-level customer contact, Salesforce Sales Agent may have better traction among this hard-to-please crowd, predicted Ketan Karkhanis, executive vice president and general manager of Salesforce Sales Cloud.
"Veteran salespeople will actually embrace this faster because now they can truly focus on what their true skills are, what their art is -- and that's the art of making the deal happen, not doing all this other work to get to the deal point," Karkhanis said.
Release notes for Winter '25 -- the next version to come out, scheduled to hit customers the weekends of Sept. 6, Oct. 5, and Oct. 12, depending on geography -- also offer some clues about new features to come. Users can now put "mini carts" for shopping on any web page; there is a newly launched Einstein for Developers documentation site; and the company has placed a new spotlight on Heroku, which will bring the ability to generate actions from External Services using Heroku apps through a metadata API, as well as more functionality for Heroku apps in general.
Generative AI in spotlight, Slack reborn
No release date information has been given about Agentforce. Salesforce typically takes six months to a year from its Dreamforce debuts to roll out big new initiatives, however, often with a handful of features arriving at a time in agile fashion.
Agentforce won't be the only tech that Salesforce likely will release or preview at Dreamforce. Einstein Copilots -- previewed last year -- feature generative AI to assist salespeople, marketers, customer service agents and others within a Salesforce user's enterprise. As Salesforce turns those previews into product releases, Dreamforce attendees can expect to see more features added.
Copilot for Slack is scheduled to be released around Dreamforce. When Salesforce acquired Slack in 2020, Benioff said that Slack "would become the Salesforce interface." While that hasn't exactly come true, Slack remains an important platform for teams of developers, marketers, customer service agents and sales representatives to work together to solve problems and create revenue.
In her first year as Slack CEO, Salesforce and Oracle veteran Denise Dresser has taken on the task of deepening the integration of Slack with Salesforce with the goal of increasing employee productivity though collaboration. She said generative AI tools such as Einstein Copilot for Slack will further progress toward achieving that.
Bringing the power of structured and unstructured data [to Salesforce and Slack] -- which we are doing -- is incredibly powerful.
Denise DresserCEO, Slack
"Bringing the power of structured and unstructured data [to Salesforce and Slack] -- which we are doing -- is incredibly powerful," Dresser said.
Salesforce also partnered with Workday to create AI-powered employee service agents for joint users, which tap data both in Salesforce and Workday that could be live as soon as the end of this year. There may be more information at Dreamforce on how those will integrate with Data Cloud and Agentforce, and what features they could provide such as employee onboarding tools and sales commission tracking and payments.
Data Cloud evolves
In interviews, Salesforce execs in the last three months have emphasized the point that AI works more effectively with more -- and higher-quality -- data.
Salesforce has been making the push to make better use of its users' customer data for years. First, it launched a customer data platform called Customer 360 Audiences in 2020 and rebranded it "Salesforce CDP" in 2021. In 2022 it was upgraded to Genie to span more than marketing use cases and gave it a "magic rabbit" mascot. One year later, Salesforce apparently did away with Genie the Rabbit and rebranded Genie as Data Cloud.
Things are different this year for Data Cloud, which most likely will keep its name.
As generative AI goes live on the Salesforce platform, piece by piece, Salesforce makes the case for users to at least consider buying in to Data Cloud, though they might use different companies' data lakes to hold customer data. Data Cloud's pre-built data models for specific vertical industries and zero-copy functions that can perform operations on data at rest elsewhere may help developers and integrators get tools live faster than without Data Cloud.
"Everyone's tried using ChatGPT or Bing on their own time," said Clara Shih, Salesforce AI CEO. "But when it comes to using it in the enterprise, it's quite different because of the corpus of internet data that's used to train and pretrain these models. There's a lot of good content there, but there's a lot of biased, inaccurate and simply incorrect knowledge out there."
The Einstein Trust Layer -- which cordons off a user's data from public large language models they might employ -- has been acknowledged outside of Salesforce as well thought out. Salesforce users training their own models on their company's data -- and then tapping public generative AI large language models securely as needed for assistance -- is the way customers will use the tools for business benefit.
To accomplish all this, Salesforce released its own vector database in June, laying the groundwork for numerous new potential Data Cloud integrations, AI-powered agents and features to be unveiled at Dreamforce.
Heady technical concepts such as vector databases and the Einstein Trust Layer aren't necessarily top of mind for many Salesforce users and admins, said Rebecca Wettemann, founder of independent research firm Valoir. But they are crucial building blocks for what will come next at Dreamforce and beyond.
"A ton of people are turning on capabilities out of the box," Wettemann said. "So, for data science, legal, compliance and everybody else to be happy, Salesforce has to have the Trust Layer in place. They have to have an approach to unstructured data. Those are foundations for what's coming with AI in the future, not necessarily things that many customers are going to be adopting today."
Altogether, it means that Salesforce can track customer behavior on the web, in sales records, or through other interactions and tag it with metadata that can be understood -- and deployed -- throughout Salesforce, said Rahul Auradkar, executive vice president and general manager of Salesforce Data Cloud. Examples of deployments might include cross-selling and upselling initiatives, improving customer service or running analytics on unstructured data that might not have been possible before.
"Data Cloud is deeply embedded in the Einstein 1 platform," Auradkar said. "It means that you can take the structured data that you get with engagement events -- somebody touched your website, somebody touched your mobile app, your last few purchases in the last few seconds, how did you scroll my website -- you can capture that structured data, unify that with our metadata that we already have about [the customer], and bring that data to life."
Don Fluckinger is a senior news writer for TechTarget Editorial. He covers customer experience, digital experience management and end-user computing. Got a tip? Email him.
