Salesforce fesses up: Demand for AI agents is slow

Welcome to Runtime! Today: Why Salesforce is going to have to wait a lot longer than it expected for the agentic AI revolution, Nvidia, meanwhile, still can't make enough GPUs, and the latest enterprise moves.

(Was this email forwarded to you? Sign up here to get Runtime each week.)

Hold please

Pretty much every single enterprise software company went on an agentic AI push last year, but nobody puffs up an interesting but immature technology like Salesforce CEO Marc Benioff. It became clear on Wednesday that it will take a long time before customers adopt its Agentforce product.

Salesforce missed Wall Street's guidance for fourth-quarter revenue and lowered expectations for its new fiscal year, which sent its stock down 4% on Thursday. It also missed analyst expectations for revenue from its two most important products — Service Cloud and Sales Cloud — according to CNBC, which are only growing at 8% a year.

• Nevertheless, Benioff opened Wednesday's earnings conference call by saying "this was just the best quarter we've ever had," which is simply not true.
• Salesforce did generate more cash than expected (which is good) and Benioff also said that revenue from its Data Cloud product grew 120% to $900 million.
• However, that's less than half of the revenue generated by both Sales Cloud and Service Cloud's product lines.
• And most of that growth appeared to come from existing Salesforce customers, as Benioff said that 25% of new data coming into Data Cloud came from outside its network.

Benioff let outgoing CFO Amy Weaver drop the real bit of news from the call: Salesforce doesn't expect to see "meaningful" revenue from its Agentforce product until its 2027 fiscal year, which doesn't start until February 2026.

• "We are assuming a modest contribution to revenue in fiscal '26," Weaver said, according to a transcript posted by Seeking Alpha. "We expect the momentum to build throughout the year, driving a more meaningful contribution in fiscal '27."
• Undaunted after hearing that statement, Benioff closed the call by proclaiming that "this is going to be the absolute year of Agentforce," minutes after the company said that's actually going to happen next year.
• "Sales can't force Agentforce adoption," said D.A. Davidson analyst Gil Luria according to Yahoo Finance, and unfortunately for the company sales is the thing that Benioff and Salesforce do best.

It's become increasingly clear over the last six months that generative AI adoption is moving slower than vendors desperate for growth would like, and it's good to see Salesforce acknowledge reality, which is not something it has been particularly good at doing in the past. The fact is there just aren't a lot of companies like Liberty Mutual — which modernized its approach to data management and analysis long before ChatGPT dropped — that have their data prepared to deploy production apps.

• As a result, companies like Salesforce have two challenges in getting their customers to adopt generative AI tools; they have to walk customers through the process of preparing their data, and then they still have to sell them on the value of their tools when every enterprise software company in the world is selling more or less the same thing.
• "Without quality data, the tools just deliver junk. Feed them bad data, and they deliver more junk," wrote Mike Pastore on MarTech. "Now imagine AI agents with access to bad data and little human oversight. Does this sound fun? No, it does not."
• As Pastore pointed out, the fact that Salesforce is seeing strong growth for its Data Cloud product is a good sign that customers are starting to figure it out.
• But a lot can happen in a year or two, and as companies continue to move their data into data lakes with open storage formats that can be used alongside any vendor's tools, it becomes harder to see how Salesforce plans to deliver on Benioff's boasts.

Blackwell turns green

Nvidia remains the clear winner of the generative AI boom. It beat Wall Street's lofty expectations for revenue and profit during the last year, and signaled Wednesday that it's not slowing down as its next-generation Blackwell chip rolls out.

Nvidia recorded $39.3 billion in revenue during its fourth quarter, a 78% jump compared to last year and more revenue as the company recorded in its entire 2023 fiscal year, which ended right after ChatGPT emerged. And CEO Jensen Huang downplayed the brief freakout that investors had last month after DeepSeek launched its cheap reasoning model, telling CNBC that the computing power required to run reasoning models is "100 times more than what we used to do."

One potential problem for Nvidia, however, is that 90% of its revenue comes from data-center customers, who are designing their own AI chips and looking for alternatives from companies like AMD, Arm and Qualcomm. Those customers will all add the Blackwell chip to their data centers this year, for sure, but by the time the next generation comes along the competitive situation could look quite different.

Enterprise moves

Beth O’Callahan is the new chief administrative officer at NetApp, a newly created role that adds HR responsibilities to her legal duties.

Joel Chaplin and Dave Shuman are the new CIO and chief data officer, respectively, at Precisely, an internal promotion for both.

The Runtime roundup

Dell missed Wall Street estimates for revenue but beat the profit estimates and raised its profit guidance for its fiscal year.

Snowflake beat the Street on both revenue and profit, and its stock rose nearly five percent on Thursday.

IBM closed its $6.4 billion acquisition of HashiCorp after U.K. and U.S. regulators cleared the way.

Slack suffered a widespread, day-long outage on Wednesday after "a maintenance action in one of our database systems, which, combined with a latency defect in our caching system, caused an overload of heavy traffic to the database," it said in a report to customers.

On the plus side, Benioff confirmed that DOGE is using Slack for internal communication, which must have made the destruction of the federal government a little bit harder yesterday.

Thanks for reading — see you Saturday!