Salesforce's Benioff Says Vendors Have an Agentic AI Pricing Problem

ORLANDO, Fla. -- The vendor community has yet to fully solve the pricing problem for agentic AI. That's according to Salesforce CEO and Time magazine owner Marc Benioff, who spoke with Yvonne Genovese, executive vice president of business and technology insights at Gartner, during a keynote Tuesday at Gartner's IT Symposium/Xpo in Orlando, Fla.

While Salesforce has opted to present customized options to customers for agentic AI pricing, Benioff said the company is still working out the best approach for pricing -- along with every other vendor.

The concession was the only pointed critique that Benioff -- fresh off his company's blockbuster Dreamforce conference and days after retracting his statement about deploying National Guard troops to San Francisco -- offered regarding the company's push to make agentic AI an integral and dynamic feature of the Salesforce platform.

"About nine months ago, I made the decision that I wasn't just going to have an FAQ and I wasn't just going to have a bot, but I was going to unleash an 'agentic layer' or agents, and link them into the whole service platform," Benioff said.

The latest iteration is Agentforce 360 , a three-year development effort that the company unveiled at its Dreamforce event this year. The result, Benioff said, is a platform that in the space of a week returned 50,000 incoming sales calls, freeing up time for Salesforce's 20,000 salespeople.

Related:Make your own mandate: How CISOs can implement GenAI governance

"An agentic enterprise is definitely a company where humans and agents are working together and are optimized," Benioff said.

Headcounts, How to Pay for Agentic AI

However, the use of agentic AI at Salesforce comes with a price of its own -- Genovese pointed out that Salesforce recently laid off 4,000 customer service agents as the company relies more on Agentforce to answer customer calls.

"We have more employees than we've ever had right now," Benioff responded, in a nod to the company's current workforce of 70,000 -- but it's balanced "differently."

"I have a smaller percentage in service ... I tilted more to distribution, because there were things that I did that all of a sudden created more opportunity for me," he said.

In addition to affecting headcount, there's another issue with agentic AI -- pricing and ROI, Genovese explained. CIOs can't justify the cost of AI on productivity improvements alone and are tasked with demonstrating the ROI of AI investments to their CEOs, she said.

"One of the things that [CIOs] are struggling with is that they don't know how to pay for this thing that you're talking about. This is not off the shelf. You're not going to go and take your AI, point it at your data and say, 'Go forth and prosper,'" Genovese said.

Related:How Collectors and Verizon use AI: Billion-dollar plans and 1,000 models

Pricing agentic AI in a manner that's palatable to customers is indeed difficult, Benioff acknowledged. Customers differ on how they want to pay for agentic AI, including based on outcome, per user, per action, or by consumption, which has created challenges in pricing the technology, he said.

"So we introduced a new agentic enterprise license agreement, which is just a flexible model that gives us the ability to open it up, that the customer can choose their pricing," Benioff said. He added that the company is not at a place where it can dictate pricing but instead needs to "let customers have ultimate flexibility on pricing."

AI ROI Requires 'Extreme Partnership' Between Vendors, CIOs

Still, CIOs need more than just flexible pricing options, Genovese said, pressing Benioff on how they can demonstrate the ROI for AI.

"This is going to take extreme partnership," Benioff said. "And if you think anybody has the complete answer or any vendor gets on this stage and says, 'Oh, I know exactly what to do here in the world of this new agentic enterprise.' No, they've made a mistake."

Related:Salesforce shows off eVerse: Another small step to enterprise general intelligence?

Vendors will have to change the way they work with customers, working closely with them to determine how "highly customized pricing" can support their AI strategies, he said.

Benioff also challenged Gartner to step up in solving the agentic AI pricing problem, encouraging the research firm to guide CIOs on getting the right price.

"Obviously, Gartner is going to help the customer with the vendor selection, the pricing, the agreement," Benioff said. "This is an opportunity where Gartner can step in and elevate all of us."

• Preparing for the Autonomous Era in ITOps

  Tuesday, December 9, 2025 at 1 PM EST

• Your Enterprise Cyber Risk Assessment

  Thurs, Nov. 6, 2025 at 1pm EST

• IT Automation in 2026: It Isn't ALL About AI (Just Mostly)

  Oct 23rd, 2025 from 11am - 5pm ET | A Sponsored Virtual Event | Doors Open at 10:30am ET

• The Architecture of Trust

• Uncovering and Thwarting a Coordinated Identity-Based Attack at Enterprise Scale

• Top Airline Identity Challenges in 2025

• Modernizing MFA Enrollment for Stronger Security

• Just-in-Time Step-Ups: Balancing Security and User Experience at Authentication