What is the story about?
In a wide-ranging interview with CNBC-TV18, Vala Afshar, Chief Digital Evangelist at Salesforce, made one thing clear: the window for businesses to decide on agentic AI is closing. With Salesforce down 31%, SAP down 28%, and ServiceNow down 35% year to date, the question of whether traditional enterprise software can survive the AI wave has never been more urgent. For Afshar, it is less a threat than a moment of reckoning.
He has been in technology for 30 years. He has lived through the cloud revolution, the mobile era, social, Internet of Things, and advanced analytics. None of it, he said, compares to what is happening now. "I have never experienced the velocity, the speed and direction which has led to the fastest growing products in my company's history." Since the last conversation with CNBC-TV18 in 2024, Salesforce has grown from roughly 20,000 employees to approximately 85,000, and its Agentforce platform now has over 25,000 customers globally. The growth in human capital and in agentic adoption, he argued, are not in contradiction. They are the same story.
Afshar invoked Darwin, but not in the way most do. Darwin, he noted, never said survival belonged to the strongest. "Darwin said if you had thick fur you're more likely to survive the winter. So it was about adaptability." What is happening now in enterprise technology, he called digital Darwinism. Size does not protect you. History does not protect you. The only thing that does is how fast you move.
The verdict for companies that are not moving: "You're either an agentic business or you're a dying one. You are on a path to irrelevancy if you don't take advantage of this technology." He was equally direct on scale. "When we talk about AI, specifically agentic and physical, it could be the impact 10 times bigger than the industrial revolution at a tenth of the time."
That is not a distant projection. At Salesforce, he said, it is already the present tense.
On what this transformation actually looks like inside a business, Afshar moved away from abstraction. He laid out a framework he calls the five Rs, each one following the last in sequence.
Redesign comes first. Before anything else, a company has to look honestly at what its people are doing and ask whether those tasks require human intelligence at all. "If there's complexity, certainly if there's deterministic work, repetitive work, it's really wasteful to have humans working on repetitive, highly deterministic, low value work." He put a number on just how much of current work fits that description. In sales, leaders spend only 28% of their time actually selling. The other 72% goes to planning, meeting notes, back office preparation. "Agents," he said, "are not taking jobs. They are taking the 72%."
Reskill follows. Once processes are redesigned, the people who were running them need to be equipped for something new. Then comes Redeploy. The people who are reskilled do not return to the same roles. They move into work that is more complex, more creative, and more often than not, better paid. "Very often it's a highly more compensated role. It's new, more creative, more complex work that could require strategic thinking."
Restructure comes fourth, covering how budgets, teams, and layers of management are reorganised once the work itself has changed. And the fifth R, Reclaim latent value, is where Afshar made his sharpest India point of the interview.
With agents handling Salesforce's customer support volume, the company found itself with both capacity and clarity it had not had before. It looked at where it was underserving its customers and found language gaps in growing markets. The result, announced at Agentforce World Tour Mumbai, was Agentforce Voice support for Hindi. "53% of this country uses that language," Afshar said. "We've been in business in India since 2001 and yet we didn't have support for the majority of the language that the men and women in this country use." Twenty-five years in the market. The fifth R is what it took to see it.
What it actually means for jobs
The internal Salesforce numbers are where the argument gets concrete. Agentforce resolved over 3.5 million customer service cases in 16 months at help.salesforce.com. The human team that had been doing that work, around 3,000 people, was not let go. "Those individuals became forward deployment engineers, solutions engineers, architects, sales professionals, programmers, professional service and technology training leaders." Some made the transition immediately. Most required reskilling first.
Today, 60,000 of Salesforce's 85,000 employees use Slack bot, the company's internal agent interface, every single day. The definition of work has shifted. "Hybrid work today is humans and AIs co-creating value at the speed of need." It is no longer a conversation about where you work. It is about who, or what, you work with. "There is no more single contributor at Salesforce. Everyone at my company is part of a team sport."
For anyone entering the workforce and trying to make sense of what this means for them, Afshar's answer was not reassurance. It was a standard. "What we're looking for is a high rate of learning and good judgement." That bar applies at every level. All 85,000 Salesforce employees, including senior executives who have run multi-billion dollar companies, are required to complete a mandatory programme to build an agent from scratch. "I tell you, it's nerve racking. I'm speaking from personal experience because it tests your technical chops."
Salesforce also runs 300 active internal agents, managed with the rigour of a professional sports roster, scoreboards, rankings, and performance metrics included. Agents that underperform get cut. Agents that deliver get more data, more rules, more responsibility. "Sometimes you fire agents, sometimes you promote agents." The language of talent management has extended, without apology, to the machines.
The commercial shift is just as structural. For most of its 27-year history, Salesforce ran on licence fees. That era is closing. "Until last couple of years it was licence-based monetisation strategy and now we're looking at agents that are consuming tokens and creating outcomes." To explain what that means in practice, Afshar drew a distinction between tokens that run a language model and tokens that execute real work, which he calls Agentforce Work Units. "I don't want the language model to write me a poem. I want it to close a case in our customer service centres."
The cultural implication of that shift runs deeper than pricing. When the model moves from selling software to delivering outcomes, the relationship between vendor and customer changes entirely. "When you go from a consumption-based, impact-based mindset, you naturally think about the marriage experience has to be better than the courtship experience." Which changes who you hire. Salesforce, he said, is no longer looking for mercenaries. "We're looking for missionaries." Asked whether the shift could ultimately be deflationary, he did not dismiss it. "It can be. I think if you don't have the right core values, if you're not guided by trust."
The question every investor and competitor is asking is what protects Salesforce from being replicated by the very AI models it is built on top of. Anthropic, OpenAI, and others are producing tools that can increasingly perform enterprise software functions. Afshar's answer came down to one word: determinism.
"Language models produce probabilistic outcomes. Determinism comes from a data foundation layer." A functioning agentic enterprise needs four systems running in concert: a context system built on a company's own data and metadata, a system of work covering sales, service, marketing, and ERP, an agentic layer, and an engagement layer. For Salesforce, that engagement layer is Slack.
The platform itself has already made a significant pivot. Headless Agentforce 360, announced weeks before the interview, was Salesforce declaring that users may never have to log in again. After 27 years of building dashboards and interfaces designed for humans, the platform is now being built for agents as the primary users, accessible through APIs and 60 MCP servers. "Agents don't need dashboards. Agents don't need login credentials." Afshar reached for the analogy of autonomous vehicles. Waymo still has a steering wheel. Tesla Cybercab does not. "We've removed massive amounts of steering wheels and gas and brake pedals out of our business."
His final word on the moat cut through the technical argument entirely. The real competitive advantage, he said, is not a data layer or an API architecture. It is people. "The question is not what's going to change in the next 5 to 10 years. The question is what's not going to change."
He has been in technology for 30 years. He has lived through the cloud revolution, the mobile era, social, Internet of Things, and advanced analytics. None of it, he said, compares to what is happening now. "I have never experienced the velocity, the speed and direction which has led to the fastest growing products in my company's history." Since the last conversation with CNBC-TV18 in 2024, Salesforce has grown from roughly 20,000 employees to approximately 85,000, and its Agentforce platform now has over 25,000 customers globally. The growth in human capital and in agentic adoption, he argued, are not in contradiction. They are the same story.
Afshar invoked Darwin, but not in the way most do. Darwin, he noted, never said survival belonged to the strongest. "Darwin said if you had thick fur you're more likely to survive the winter. So it was about adaptability." What is happening now in enterprise technology, he called digital Darwinism. Size does not protect you. History does not protect you. The only thing that does is how fast you move.
The verdict for companies that are not moving: "You're either an agentic business or you're a dying one. You are on a path to irrelevancy if you don't take advantage of this technology." He was equally direct on scale. "When we talk about AI, specifically agentic and physical, it could be the impact 10 times bigger than the industrial revolution at a tenth of the time."
That is not a distant projection. At Salesforce, he said, it is already the present tense.
On what this transformation actually looks like inside a business, Afshar moved away from abstraction. He laid out a framework he calls the five Rs, each one following the last in sequence.
Redesign comes first. Before anything else, a company has to look honestly at what its people are doing and ask whether those tasks require human intelligence at all. "If there's complexity, certainly if there's deterministic work, repetitive work, it's really wasteful to have humans working on repetitive, highly deterministic, low value work." He put a number on just how much of current work fits that description. In sales, leaders spend only 28% of their time actually selling. The other 72% goes to planning, meeting notes, back office preparation. "Agents," he said, "are not taking jobs. They are taking the 72%."
Reskill follows. Once processes are redesigned, the people who were running them need to be equipped for something new. Then comes Redeploy. The people who are reskilled do not return to the same roles. They move into work that is more complex, more creative, and more often than not, better paid. "Very often it's a highly more compensated role. It's new, more creative, more complex work that could require strategic thinking."
Restructure comes fourth, covering how budgets, teams, and layers of management are reorganised once the work itself has changed. And the fifth R, Reclaim latent value, is where Afshar made his sharpest India point of the interview.
With agents handling Salesforce's customer support volume, the company found itself with both capacity and clarity it had not had before. It looked at where it was underserving its customers and found language gaps in growing markets. The result, announced at Agentforce World Tour Mumbai, was Agentforce Voice support for Hindi. "53% of this country uses that language," Afshar said. "We've been in business in India since 2001 and yet we didn't have support for the majority of the language that the men and women in this country use." Twenty-five years in the market. The fifth R is what it took to see it.
What it actually means for jobs
The internal Salesforce numbers are where the argument gets concrete. Agentforce resolved over 3.5 million customer service cases in 16 months at help.salesforce.com. The human team that had been doing that work, around 3,000 people, was not let go. "Those individuals became forward deployment engineers, solutions engineers, architects, sales professionals, programmers, professional service and technology training leaders." Some made the transition immediately. Most required reskilling first.
Today, 60,000 of Salesforce's 85,000 employees use Slack bot, the company's internal agent interface, every single day. The definition of work has shifted. "Hybrid work today is humans and AIs co-creating value at the speed of need." It is no longer a conversation about where you work. It is about who, or what, you work with. "There is no more single contributor at Salesforce. Everyone at my company is part of a team sport."
For anyone entering the workforce and trying to make sense of what this means for them, Afshar's answer was not reassurance. It was a standard. "What we're looking for is a high rate of learning and good judgement." That bar applies at every level. All 85,000 Salesforce employees, including senior executives who have run multi-billion dollar companies, are required to complete a mandatory programme to build an agent from scratch. "I tell you, it's nerve racking. I'm speaking from personal experience because it tests your technical chops."
Salesforce also runs 300 active internal agents, managed with the rigour of a professional sports roster, scoreboards, rankings, and performance metrics included. Agents that underperform get cut. Agents that deliver get more data, more rules, more responsibility. "Sometimes you fire agents, sometimes you promote agents." The language of talent management has extended, without apology, to the machines.
The commercial shift is just as structural. For most of its 27-year history, Salesforce ran on licence fees. That era is closing. "Until last couple of years it was licence-based monetisation strategy and now we're looking at agents that are consuming tokens and creating outcomes." To explain what that means in practice, Afshar drew a distinction between tokens that run a language model and tokens that execute real work, which he calls Agentforce Work Units. "I don't want the language model to write me a poem. I want it to close a case in our customer service centres."
The cultural implication of that shift runs deeper than pricing. When the model moves from selling software to delivering outcomes, the relationship between vendor and customer changes entirely. "When you go from a consumption-based, impact-based mindset, you naturally think about the marriage experience has to be better than the courtship experience." Which changes who you hire. Salesforce, he said, is no longer looking for mercenaries. "We're looking for missionaries." Asked whether the shift could ultimately be deflationary, he did not dismiss it. "It can be. I think if you don't have the right core values, if you're not guided by trust."
The question every investor and competitor is asking is what protects Salesforce from being replicated by the very AI models it is built on top of. Anthropic, OpenAI, and others are producing tools that can increasingly perform enterprise software functions. Afshar's answer came down to one word: determinism.
"Language models produce probabilistic outcomes. Determinism comes from a data foundation layer." A functioning agentic enterprise needs four systems running in concert: a context system built on a company's own data and metadata, a system of work covering sales, service, marketing, and ERP, an agentic layer, and an engagement layer. For Salesforce, that engagement layer is Slack.
The platform itself has already made a significant pivot. Headless Agentforce 360, announced weeks before the interview, was Salesforce declaring that users may never have to log in again. After 27 years of building dashboards and interfaces designed for humans, the platform is now being built for agents as the primary users, accessible through APIs and 60 MCP servers. "Agents don't need dashboards. Agents don't need login credentials." Afshar reached for the analogy of autonomous vehicles. Waymo still has a steering wheel. Tesla Cybercab does not. "We've removed massive amounts of steering wheels and gas and brake pedals out of our business."
His final word on the moat cut through the technical argument entirely. The real competitive advantage, he said, is not a data layer or an API architecture. It is people. "The question is not what's going to change in the next 5 to 10 years. The question is what's not going to change."
