AI agents are more than a productivity boost; they’re fundamentally reshaping customer interactions and business operations. And while there’s still work to do on trust and accuracy, the world is beginning a new tech era — one that might finally deliver on the promises seen in movies like Minority Report and Iron Man, according to Salesforce CEO Marc Benioff. Benioff, who recently spoke with Foundry’s “Today In Tech” host Keith Shaw on a variety of IT topics, differentiates AI agents from just generative AI (genAI), calling agents “the magical layer” that turns raw AI potential into real business value. He emphasized their ability to handle tasks with context, personality, and deep data awareness — something earlier AI tools lacked. Much of the Today in Tech conversation focused on AI agents, the software programs powered by AI that can make decisions, take actions, and interact with people or systems — often autonomously — to complete a business task or goal. “When a lot of people think of [AI] agents, they think of Jarvis in the Iron Man movies, and that’s what we want. We want a personal assistant who can go and do all of these things for us.” — Salesforce CEO Marc Benioff Benioff mentioned a scene from Minority Report where customers walk into a futuristic Gap store and a hologram greets them by name and recommends products based on what they’ve bought before. He said current AI technology isn’t that advanced yet, but it’s definitely heading in that direction. “I know what the Gap store is going to look like in the future,” Benioff said. “When a lot of people think of [AI] agents, they think of Jarvis in the Iron Man movies, and that’s what we want. We want a personal assistant who can go and do all of these things for us.” GenAI tools like ChatGPT, Grok, or Gemini “are impressive,” but they still make mistakes. With more data and context, they’ll keep getting better, he said. Salesforce’s take on genAI and AI agents Salesforce has deeply embedded AI into its CRM product ecosystem through a strategy it calls the “Einstein 1 Platform.” It also offers Agentforce, a platform that enables businesses to create and deploy autonomous AI agents across various functions such as sales, service, marketing, and commerce.   Salesforce has also integrated genAI into its platform through Einstein GPT and Einstein Copilot, which function as intelligent assistants within applications. They can draft emails or customer responses, summarize case histories, auto-fill CRM records, and answer internal questions by leveraging a company’s own knowledge and data. Using Salesforce clients such as Singapore Airlines as examples, Benioff said when an AI agent is layered atop corporate data, it creates a hybrid customer support service with both humans and AI, offering an overall better experience. A customer can say, “Book me a flight to New York, and make sure it’s on Delta,” and an AI agent will check the passenger’s calendar, search flights, and book it with their loyalty number. It will then email anitinerary, all without the person ever clicking on a single button. Benioff also pointed to Lennar Corp., a Florida-based homebuilder that ran a hackathon after Salesforce’s Dreamforce and developed five AI agent use cases to improve customer service and save money. And he noted how Disney uses AI agents to help employees understand its vast ecosystem and personalize customer experiences in real-time — even rerouting guests to other rides if one breaks down. “Disney is an amazing company but let me tell you a secret about Disney that a lot of people don’t really think about, which is it’s really hard for their employees to understand all their products,” Benioff said. “This idea that an employee has to understand all those products, to talk to the customer, to put together the perfect package, the perfect idea for that customer. That’s pretty complicated.” Pandora Jewelry, he said, uses Salesforce AI agents to recommend products based on a customer’s existing purchases, enhancing real-time retail interactions. Data security and trust remain issues While AI agents are powerful, issues like privacy, data security, and trust still need to be solved. Though many companies consider AI agents as still in the experimental stage that comes with risk, others are looking to quickly adopt the tech anyway in the hopes of creating new efficiencies. As the adoption of autonomous AI agents explodes, vulnerabilities that allow them to be gamed or even weaponized are already emerging. Salesforce’s AI agents, Benioiff said, operate within strict security and data-sharing rules, so they only access what each user is allowed to see. Because the genAI tools are built directly into the platform, they understand both data and metadata, ensuring accurate, secure, and compliant responses — essential for sensitive industries like banking or airlines. The accuracy of Salesforce’s AI tech, Benioff acknowledged, currently stands at about 85%. But while current agentic AI is not perfect, next-gen “multisensory” models could change that. Benioff admitted that early genAI tools like Microsoft’s Copilot were overhyped in the enterprise, but said newer tools like Cursor and Surf Rider are surpassing earlier models. He also praised open-source AI, cost-saving innovations like Mixture of Experts (MoE), and said he supports emerging alternatives like DeepSeek. Most current genAI tools based on large language models (LLMs) are little more than next-word predictors based on patterns in text.  However, in the same way humans can draw on past experiences, and everything around them – sight, sound and touch – next-gen AI will be able to consider metadata, past experiences and information shared with AI models or multi-sensory inputs. The rise of genAI in healthcare And the technology is expected to improve healthcare by enabling a more complete patient diagnosis through near-instantaneous access to a vast depth of history and diagnostic tests. Patients will also be able to use the tech to be their own best advocates. Last September, Benioff said he ruptured his Achilles tendon, and his doctor suggested surgery. Taking advice from GPT, Benioff instead told the doctor he was going with a self-regenerative approach inspired by Tony Robbins’ book Life Force. Six months later, he said, his Achilles is fully healed. “I’m walking around without a boot and it regenerated. My doctor…is like, whoa. How’d you do that?” Benioff said. “It’s not magic.” “Think about if the agent is really there to help guide you and make help you have those decisions, make the right steps and go forward,” he continued. “You know cancer patients who are dealing with complex treatments like chemotherapy and others, being able to have that  24/7 care, I think, is going to make a huge difference.” In rural areas with limited access to specialists, genAI could augment orthopedic surgeons and others, helping them make better, data-driven decisions about treatments like surgery or regenerative options. That said, the technology remains limited in its abilities. Until there is a “multi-sensory model,” which is the next generation of AI, it will continue to be good, “but not perfect,” Benioff said. GenAI and the workforce One subject that has had the IT industry on edge is in recent years involves workforce reduction brought about by AI. As the technology becomes more competent at automating tasks, the fear is that human employees will be replaced. In the interview, Shaw referenced an opinion column Benioff wrote for Wall Street Journal in which he said a Morgan Stanley report showed a 20% to 50% cost savings from AI developments, largely from reduced headcount, office space, and overhead. “I think that we have to deal with reality here,” Benioff said. “So, I think it would be a huge mistake for me to not directly address this issue head on and to really talk about what is happening. That is that, yes, this is doing some of the roles that are being done by human beings. So, we need to start to adjust. We need to look at what is really going on.” “I think that it’s all about education that we’re going to need to make sure people are well trained, well educated — they can reskill.” — Marc Benioff Companies, he said, have to invest in reskilling workers, rather than just “making promises” there won’t be layoffs. Salesforce, he said, is investing in reskilling, but he also stressed the importance of corporate responsibility for helping to advance public education. “You can’t just talk about it; you have to fund it,” he said. “Our biggest grantee is our San Francisco and Oakland public schools, because I think that it’s all about education that we’re going to need to make sure people are well trained, well educated, they can reskill,” he said. “And reskilling is another area that we put a huge amount of money into the last 20 years, and this is important. And yes, we’re all going to have to think about this.” Despite its size, Benioff sees Salesforce as a startup at heart that’s focused on emerging tech and driven by vision, values, and innovation. He recalled the early days — bootstrapping with angel investors like IDG founder Pat McGovern, and how industry events like IDG’s DEMO shaped the company’s beginnings. Notable products introduced at DEMO include Salesforce.com itself, TiVo, VMware, Evernote, E*Trade, WebEx, and Fusion-io. (Until last month, Foundry was owned by IDG.) “Salesforce started in a super adverse environment where funding was highly constrained. So, we had to raise all our money — you know, privately. No venture capitalists would even invest in the company, which was amazing,” Benioff said. “And that’s why people like Pat McGovern…[were] angels in technology investing.” ​