For at least a decade, Salesforce CEO Marc Benioff has emphasized the importance of shoshin, the Japanese word for “beginner’s mind”—including in a 2015 Fortune interview in which he said companies must “constantly be creating the future” and “always be ready to start again.”
In today’s generative AI era, Salesforce is indeed starting again to create its future, with what Benioff calls the company’s “hard pivot” to Agentforce—a new AI agent platform that will make a splashy debut at the company’s annual Dreamforce conference this month. Agentforce lets users quickly build and deploy autonomous AI-powered agents—pieces of software that can make decisions and act on information—that run on top of Salesforce’s existing apps for businesses and automate customer service tasks. Agentforce is meant to be a significant step beyond AI chatbots, like OpenAI’s ChatGPT and Salesforce’s own Einstein GPT, which answer questions but cannot complete tasks that require multiple steps and decision-making.
In an exclusive interview with Fortune, a week before the company’s Q2 earnings report, Benioff once again invoked his “beginner’s mind.” He explained that the next wave of AI will be all about agents, and that as a result, Salesforce, 25 years after its founding, is, in a sense, starting all over again as an AI agent startup.
“It is about driving through the innovator’s dilemma,” Benioff said, alluding to Harvard professor Clayton Christensen’s 1997 book, The Innovator’s Dilemma: When New Technologies Cause Great Firms to Fail. “The only way you’re going to do that is with shoshin, a beginner’s mind,” he explained. “You have to start at the beginning and think it all the way through.”
Benioff’s comments were also a response to a recent unflattering blog post by venture capital firm Andreessen Horowitz that cast doubt on Salesforce’s ability to keep its edge in the age of generative AI. The blog post, titled “‘Death of a Salesforce’: Why AI Will Transform the Next Generation of Sales Tech,” said, “AI will so fundamentally reimagine the core system of record and the sales workflows that no incumbent is safe.”
There’s no doubt that Salesforce is the most powerful maker of sales software, known in the tech industry as customer relationship management (or CRM). Today, the company has $38 billion in annual revenue and 75,000 employees. Benioff, Salesforce’s cofounder, who ended up putting legacy corporate software providers out of business, is now Zen about Salesforce itself being potentially disrupted in the same way that Oracle’s Siebel, the CRM leader of the late ’90s and early 2000s, was disrupted by Salesforce.
“This is the next big transformation,” he said. “The question is, what is the future of CRM? What is the future for all of our businesses and customers and companies?” The answer, he says, has always been the promise of AI.
But now, Salesforce must prove it can win a fierce AI race against competitors ranging from Microsoft and SAP to HubSpot and ServiceNow. After a rough first quarter, Salesforce delivered strong second-quarter earnings that Benioff used to talk up Agentforce’s planned debut, mentioning it a whopping 39 times during his earnings call.
But critics like RBC Capital analyst Rishi Jaluria say the company still must walk the walk when it comes to AI. He said the company talks a big game, as it did during its recent earnings call, but has yet to deliver.
Benioff, however, shrugs off the critics, pointing out that Salesforce has hundreds of thousands of customers and the few dozen companies already testing Agentforce, including Wiley, Open Table, and Fossil, “are all like, ‘This is the biggest transformation that we’ve seen in CRM.’”
Salesforce, he added, has “done a lot of things strategically to get ready for this next wave—and there’s no question we’re going through one of the biggest, most exciting waves in the history of technology.”
Salesforce’s acquisition strategy continues with AI
It was a decade ago, at a time of groundbreaking research in machine learning, that Benioff says he had his first “existential moment” around AI.
“I was like, ‘Oh, fuck, AI is going to take over the world,’” he said. “I need to hit the accelerator pedal—and Salesforce was about 10 times smaller.”
Since then, his strategy to speed up AI development has been similar to the company’s overall strategy for mergers and acquisitions, which has famously included spending tens of billions to buy companies including data software company Tableau, workplace messaging service Slack, and marketing technology firm ExactTarget. “What I did was I started buying a bunch of companies and buying a bunch of talent,” Benioff recalled. For example, in 2016 Salesforce acquired AI startup MetaMind, founded by Richard Socher, who became the company’s chief scientist and helped launch Einstein, which was Salesforce’s core AI product in the pre-ChatGPT days (Socher left in 2020 to found AI-powered search engine You.com).
By 2021, Salesforce had already developed large language models like CodeGen, which helped the company quickly release generative AI products after OpenAI’s ChatGPT premiered in 2022. Benioff described ChatGPT’s debut as a “wow” moment “where everybody needs to rethink what users look like in a future that is informed by generative AI, not just predictive AI.”
In March 2023, Salesforce introduced an LLM-based chatbot called Einstein GPT, which soon expanded to Einstein Copilot, adding AI-powered search and an AI assistant embedded across Salesforce’s applications. However, by May 2024, its shares had plunged 16%, missing revenue expectations for the first time since 2006. Analysts were unconvinced that AI would lift Salesforce’s profitability anytime soon.
Salesforce is steadily investing in hot AI startups, including Cohere, Together AI, and Anthropic. But Benioff insists it does not intend to acquire teams developing LLMs, as Microsoft did with Inflection in March and Amazon with Adept in June. “Model teams are going to be duplicative for what we already have,” he said.
What Salesforce did not have, however, was a company to help make the best easy-to-build AI agents, which is why it bought AI agent startup Airkit in 2023. Its cofounder, Adam Evans, is now SVP of product for the Salesforce AI platform. Over the next five to 10 years, Agentforce is “really going to be a chance to completely transform our business,” Evans said.
Data is an important Salesforce advantage
Clara Shih, who has served as CEO of Salesforce AI since 2023, said Salesforce may be an incumbent, but it has an important advantage in AI that others don’t: data.
“When you look at what enterprises actually need to deploy AI successfully, you need data,” she said. “It’s a huge advantage to be the trusted guardian and steward of data.”
One longtime Salesforce customer agrees. Peter Burns, director of marketing and digital at Heathrow Airport, said the organization spent five years bringing together disparate products, services, apps, and websites to create a single customer experience. “Salesforce was very much at the heart of that, and as a result, we’ve seen some really good revenue and service improvements.” After that, it was fairly simple to “switch on” Salesforce’s generative AI offerings, he said, and the organization already uses Einstein Copilot to support contact center agents with automated responses and call summaries.
While Heathrow is currently testing Agentforce, implementing it won’t happen overnight, Burns said. But over the next year or two the organization plans to integrate some of the airport’s booking and service capabilities using the platform.
Pivoting the entire company to AI agents
Both Benioff and Shih claim the generative AI moment has sparked a cultural moment for Salesforce, particularly as Agentforce prepares for its full launch.
“It’s energized and electrified the entire company in a way that it’s hard to describe in words,” said Shih. “Despite our size and scale, just in the last 18 months since ChatGPT, it’s really made the company feel like a startup.” That has led to changes in all things Salesforce, she said, including marketing and customer support. The rapid pace of AI evolution means the entire sales organization now gets training every single week, rather than just a few times a year.
According to Benioff, the bottom line is that “we have to pivot the whole company to agents, just as we have to Data Cloud in the last two years,” he says, referring to Salesforce’s platform that brings customer data together from multiple sources across its applications, and that has been a company focus in recent years. That means big changes at Dreamforce, which will draw tens of thousands of Salesforce customers to San Francisco starting Sept. 17. “It’s not going to say, ‘Welcome to Dreamforce’—it’s going to be ‘Welcome to Agentforce,’” he said. “We’re going to put thousands of customers through workshops, getting their agents live at Dreamforce. We’re going to go on a world tour after that, called Agentforce World Tour, with a goal of putting tens of thousands of customers live by January of next year.”
But some analysts, like Jaluria, remain skeptical. Being the incumbent with control over so much data may give Salesforce a head start, he said, but it’s not enough to win the AI race. “Generative AI is going to change the way salespeople work, but unless the incumbents can use the head start, they have to actually reimagine the way they do things; it can be really difficult,” he explained.
It’s not that customers will start unplugging Salesforce and buy sales software from AI upstarts, he added. “How it starts is, ‘I’m keeping Salesforce, but I’m going to buy something from an AI native company because I can do things that I cannot with Salesforce AI,’” he said. “And then five years from now, we might have conversations about some new AI native CRM that is literally built on top of OpenAI.”
For his part, Benioff remains confident about Salesforce’s prospects for AI success. “In the beginner’s mind, I have every possibility,” he said. “In the expert’s mind, I ask you—am I an expert in CRM? I am. Have I been doing enterprise software for 40 years? I have.”
Benioff says he believes he and his team have a beginner’s mind—and though they may make some mistakes along the way, and there is more to do, they are ready for what is coming.
“Could there be another hot company coming along? I hope so, I mean, it’s the history of our industry, and God bless all of them,” he said. “But we’re also going to do our best as well—this isn’t our first rodeo, and we have a lot of assets to bring to bear, and I think for our customers, we’ll find out over the next six months.”
