Salesforce CEO Marc Benioff Why AI Agents Don’t Mean End of White Collar Work

Benioff: I’m looking at all the features and asking: How do I become an Agent Enterprise?

Support is a great example. We currently have over a million conversations between our customers and agents, and at the same time, we have around one million conversations between humans and customers. This has only lasted for 6-9 months, and so far reduced support costs by 17%.

The second piece is for sale. We have so many leads that we can’t follow up all of them. Salespeople basically choose what leads them to call back. Thousands of leads, tens of thousands, and hundreds of thousands of leads have never been recalled. But in the world of agents, there is no excuse. You can follow up on all leads.

Fortune: What about those who worked in customer support when these efficiency increased? Are they moving into other roles or are they shrinking their workforce?

Benioff: I’m constantly moving people and rebuilding my company. There are many horrifying stories planted by executives saying we are getting big AI layoffs. But what we have about AI is not 100% accurate as it is built on a word model. Without the dataset, these models are 50% or 60% accurate. Adding a dataset gives you 90% accuracy, but not 100%. Therefore, the loop requires humans. Humans will not leave. We are augmented by these technologies. Productivity is improving.

FORTUNE: You said this year you’re not hiring only salespeople, supporting people, or supporting lawyers. What does this mean for young people trying to enter these fields?

Benioff: No negative effects were seen. I think there are people who are harming the spirit of business by saying things that may not be true. What I see is much smaller businesses and more mid-market companies. With everyone increasing and the ability to do more, there is a possibility that more jobs could potentially be employed.

I think there will be an explosion of small and medium-sized businesses because small and medium-sized businesses can do more. We definitely see this in our business.

Fortune: If people need to move to new positions to improve these efficiency, how do you manage the transition internally?

Benioff: I don’t think it’s very complicated. We do what we call Trailhead, which is made available to our customers and employees. All products can be trained. It is recommended that all employees do it, be certified, get badges and get trained. It gives them more mobility to the organization.

People need to be more flexible in their thinking. We encourage people to have a beginner mind. We tell them: in the heart of a beginner, you have all the possibilities. In the mind of an expert, you have some. Which is your choice? There are many options. There’s only an opportunity. You can see all the open jobs we post externally and internally.

Fortune: There are concerns that AI will lead to massive layoffs across the economy. Do you see it happening?

Benioff: I keep looking around, talking to the CEO and ask: What AI do they use for these big layoffs? I think AI will increase people, but I don’t know if it will necessarily replace them. Even the radiology department where AI can read scans are not 100% accurate. AI can read scans, but it may be wrong.

The reason is that much of this is still built on a word model. There may be more accurate future AI models, but that’s not where we are now. This is about humans and AI working together. I feel like I have a partner. But it doesn’t always make it right.

Fortune: Some people are worried that companies can use AI to create more custom-made software, potentially threatening traditional SaaS companies like Salesforce.

Benioff: It is always true that businesses can DIY, but it is true that only certain companies can do it. Small and medium-sized businesses don’t go DIY because their IT departments aren’t big. But when you’re talking to a company like Barclays With 15,000 engineers, our development team has always seen us as a potential competitor.

Apps are more dynamic and have the ability to dynamically generate apps. I see opportunities in the area, but we are not at that point yet. I can’t give me an example of someone dynamically building an enterprise app from AI. You can now define your app in English, which is exciting, but built on a platform like ours with the same framework requirements.

Fortune: What do you think about the current state of AI accuracy and capabilities?

Benioff: We need to be more realistic about what we have. There is this concept of intelligence that comes out of these tokens, but it is not so accurate. Users may help models write, summarise, edit stories and translate them into Spanish or Arabic. And they’re like, “Oh, this is cool.” But at the end of the day you still need to check.

All AI needs their own fact checkers, and those fact checkers are human, not AIS. This is because AI cannot check facts because it does not have that level of accuracy. Humans must stay in the loop.

Fortune: How do you see AI transforming different sectors?

Benioff: In healthcare, small towns don’t have enough doctors. There is a company called Artera, which has FDA certified for prostate cancer diagnosis and treatment planning. In our small town (where Benioff lives in Hawaii), there are no urologists or oncologists, so this can increase your ability. It’s not a replacement for having an expert, but it can help.

In education, children write their papers using tools like grammar. However, it is no excuse for teachers to not look at the history of grammar to ensure that their children are really learning. Education can be enhanced and healthcare can be enhanced.

But burger shops, pizza shops, supermarkets, dry cleaners, farmers markets – many of the small town’s cores probably won’t change that much. Probably no Robotaxis. For a while, no one will map our town. This vision of not only turning the switch over, but suddenly the car started driving itself, turns out that it wasn’t true. And I think that’s a good comparison for what’s happening more widely in the AI industry. This is about humans and AI working together, not human wholesale replacements.

luck: Please tell us more about how human-AI partnerships work.

Benioff: I think we are all expanded to our ability to do our job. It feels like you have a partner, right? So, for example, I sit every year to write a Salesforce business plan. This process is called V2MOM (vision, value, methods, obstacles, and measurement). I’ve always sat with Salesforce executives and now I work as a trinity with AI. And I’ll say, now, here’s my whole plan, give me a grade on it. And AI says “B Plus”, why isn’t it “A”? And then, “Well, you ruled out this, right? You’ve ruled out that.” I said, “I’ve ruled out that!” That’s right. In fact, it’s very good to find what you excluded. So it’s good at finding a gap in your consciousness. I was very impressed with it and it changed my mind a few times. But that’s what you’re working on the system. Andy thinks it’s a very powerful message for people and businesses. This helps employees become more productive and move faster. But it is not a human wholesale change. Or, if that opportunity exists, someone needs to explain it to me. Because as CEO of a company of 75,000 people, I don’t understand that.

Fortune: How will this change the way organizations are structured?

Benioff: This allows you to increase the control span and reduce the (administrator) layers. But people continue to talk about how robots are trying to replace the wholesale sector. Where are these robots? Because it’s not visible. Certainly there are some robots out there – this year we’ll do some walks in Dream Force, but it will be very constrained and somewhat restrictive, and that’s about the vision of what’s possible in the long run. But let’s talk about where we are now. And where we are now, all companies can become agents, but we need to keep people on a loop.