The market had already written off software companies this year. Between mid-January and mid-February, about $1 trillion was wiped off the value of SaaS stocks, and Salesforce, the company that invented cloud CRM, was among those hit. Then came the late-August report, and the stock surged about 23% in a single day.

Itay Margalit, who has headed Salesforce’s business operations in Israel since August 2021, is unfazed by the situation. “AI is not a replacement for software,” he says. “It is a force multiplier.”
Salesforce and the stock
Salesforce was founded in 1999 and is the world’s largest player in customer relationship management software. It also owns the Slack platform and invests in Anthropic, the developer of Claude.

In the second quarter of fiscal 2026 (May through July 2026), the company reported revenue of $11.35 billion, growth of 11%, and raised its annual revenue forecast to $46.1 billion to $46.4 billion. Part of the surge in profit came from a $2.6 billion gain on investments, including Anthropic shares, which stood at about $240.25, with a market capitalization of about $197 billion. That is still about 9% below the end of 2025, but about 64% above this year’s low of around $146.
“A chat system that only estimates probabilities is not ready for organizations”
According to Margalit, the innovation is that customers no longer have to enter a Salesforce screen. Language models operate probabilistically: They guess the most likely answer. Salesforce adds a deterministic layer to them, with fixed and accurate business information.

“For nearly three decades, we have built business processes and data in a way that makes it possible to see the customer’s status at any given time, with a single unified truth,” he says. “When that is combined with the probabilistic component, you get tremendous power. If you only have the probabilistic component, it is not robust, and certainly not ready for organizations.”
And customers’ data? “Absolutely not,” he replies, explaining Zero Copy: “The model accesses the data for exactly the fraction of a second it needs to perform the action, and afterward the information is forgotten.”
Koa: Salesforce and Nvidia’s exclusive model
One of the prominent announcements at the conference was Koa, Salesforce’s first inference model built for CRM work, based on Nvidia’s Nemotron model. “This is an agent’s ability to analyze a complex task, make decisions and carry out a series of actions to complete it,” Margalit says. And the model is available only to the company’s customers: “It is for Salesforce customers.”

Margalit adds: “Nine of the 10 leading companies in AI use Salesforce and Slack, including OpenAI and Anthropic.”
From 2,500 unfiltered leads to handling 100%
“These are not just agents that make decisions or recommendations. They also take action,” Margalit emphasizes. Siemens, whose CEO attended the conference, uses agents to screen incoming leads in 110 countries. Before Agentforce, it received about 2,500 unfiltered leads a month. Today, it handles 100% of them.

In Israel, he cites Assuta, which operates an Agentforce agent for patient service. The agent, “Assi,” responds on WhatsApp 24 hours a day. According to Assuta, in its first week it managed more than 3,000 conversations, and about 60% of inquiries were completed without the need for a human representative to intervene.
“Cut positions or grow? That is a management decision”
“Things that took me a full workday before the agentic era now take me several minutes,” Margalit says. “But no matter how many resources an organization has, managers will always say they need more. Ultimately, it is a management decision for each organization whether to use this to cut positions or to grow and expand into new areas.”
The Israeli market: High-tech and finance lead, public sector awakens
“There is very strong interest in the financial sector, and customers are already using agentic technologies and seeing value,” Margalit says. High-tech, both Israeli and global, is also involved. “The Israeli market is entrepreneurial by nature, with early adopters. I know almost no organization we work with that is not using, examining or preparing to begin using agentic AI.”

The surprising entry on the list is the public sector. “Even if it is a little behind, we have a customer that uses it across its entire organization. I am convinced it will happen, and it will happen quickly.” One example is the Real Estate Enforcement Authority at the Ministry of Finance, which is deploying Agentforce agents in its central work system across all districts.”
And will Salesforce Israel continue to grow, through new customers and existing customers expanding their use? “Absolutely.”
A strategic development center and a 10th acquisition
Israel is one of Salesforce’s three strategic development sites worldwide, alongside the U.S. and India, and it employs about 1,000 people in Tel Aviv and Nazareth. “Salesforce Israel leads or participates in a large share of our advanced development,” Margalit says, adding that the company is considering further expansion of the site in Israel.

More acquisitions? “We are always looking,” he says, “and history shows that we have already done so.” History repeated itself quickly. Between the interview and the article’s publication, it was reported that Salesforce was acquiring Israeli startup Loops, which develops an AI-based analytics platform, for about $30 million. This is the company’s 10th acquisition in Israel.
“I use AI to review my portfolio”
“Everyone would do well to try to learn about and understand the capabilities of agentic AI,” Margalit recommends to investors.

“I use it personally, and I know quite a few colleagues who use engines such as ChatGPT, Claude or Grok to review their portfolios: whether they are balanced, whether they are identifying opportunities correctly and whether they fit the level of risk they want to take.”
So is this the end of software?
Margalit quotes Salesforce CEO Marc Benioff: “This is not the end of software, but the end of the era in which software requires humans to do all the work.”
“A software company that wants not only to survive but to lead must understand that AI is an integral part of the process,” he concludes. “It should be viewed not as a replacement, but as a multiplier of the value that a software company can provide.”

And is there still room to grow? “Yes, a great deal. Until now, software companies sold tools that required humans to work hard. The next growth potential comes from software beginning to perform the work itself. This opens entire markets that were deterred by complex systems and allows software companies to charge based on business outcomes, not just subscriptions. The small chat systems will not replace these systems; they simply turn the organizational data that has been sitting unused until now into the business’s most powerful growth engine.”











