Salesforce says nearly 20,000 of its employees are now working in roles that did not exist 12 months ago as the software giant redesigns jobs around artificial intelligence.
Speaking during a “Radical Redesign of Work” boardroom briefing at Dreamforce, Salesforce President and Chief People Officer Nathalie Scardino described the adoption of AI agents as a people transformation rather than simply a technology upgrade.
Salesforce employs approximately 90,000 people across 50 countries, spanning six generations in the workforce. Scardino said the company has had to reconsider roles and organizational structures originally designed for a time when access to intelligence was limited.
“We designed jobs and org charts around that scarcity,” she said.
Salesforce is organizing the transformation around what it calls the four Rs: redesigning jobs, reskilling employees, redeploying talent into emerging roles and rebalancing human capacity with agentic technology.
The company says every job is being reconsidered as part of that process. All Salesforce employees have undergone AI training, while participation in quarterly leadership meetings has required completing 16 hours of hands-on instruction building agents in a sandbox environment.
The clearest example comes from customer support. As Agentforce began handling more interactions through Salesforce’s help website, work previously performed by support engineers started to change. Rather than simply eliminating those positions, Salesforce says it reskilled and redeployed employees into forward-deployed engineering roles that combine selling, implementation and technical problem-solving.
Scardino said nearly 20,000 Salesforce employees are now in roles that did not exist a year ago. The figure refers to existing workers moving into newly designed positions, not 20,000 additional hires.
One of the most significant emerging roles is what Salesforce calls the “agentic manager”: an employee responsible for overseeing both people and digital workers.
These managers may be tasked with onboarding agents, monitoring their performance, managing permissions and ensuring they use the correct policies, content and company knowledge. Salesforce has also added roles involving prompt engineering, multi-system agent orchestration and ethical AI design.
“This is the year of the manager,” Scardino said. “And that is the ability to manage humans and also agents.”
The shift is particularly visible inside Salesforce’s HR organization.
Andy Valenzuela, EVP of Global Rewards and COO of Employee Success, said Salesforce has consolidated approximately 150 processes across the employee lifecycle into 12 “agentic journeys.”
The company receives roughly four million employee inquiries each year. Its goal is for only 25,000 of those inquiries to require human intervention by the end of the year, with AI also helping employees resolve the remaining complex cases.
One automated workflow handles employment verification. An employee seeking a letter for a mortgage or another purpose can make the request through Slack, which guides the conversation, creates the case, produces the document and returns it to the employee. A process that once took days can now be completed in minutes.
Valenzuela said 30 percent of the employees who previously performed that work have already moved into new jobs since the beginning of the year.
Salesforce has also introduced an agentic promotion process that draws information from nearly 10 systems, helps managers prepare a case and challenges whether an employee is ready for advancement. Humans continue to make the final promotion decisions, but 99 percent of submissions during Salesforce’s most recent cycle were completed through the new experience.
MaryAnn Patel, Salesforce’s SVP of Product Management, said agents are also changing the role of sales representatives. Instead of personally completing every administrative and prospecting task, representatives can oversee agents that research accounts, recommend next steps, contact customers and schedule meetings.
“The rep, on the other hand, goes from being an individual contributor to really a manager of digital work,” she said.
Salesforce’s initial experiments were not universally successful. Patel recalled one agent deployment that performed poorly because the company had not connected the necessary data and different departments had conflicting ideas about what success looked like.
Sales, marketing, product, engineering and IT eventually developed shared metrics and redesigned the team around a common outcome. Within weeks, Patel said, the project had shifted from a failure to a successful internal case study.
The experience reinforced another principle behind Salesforce’s workforce strategy: “Agents don’t know silos.”
As AI works across departments and systems, Salesforce expects traditional job boundaries to become less rigid. Product employees may increasingly perform engineering work, engineers may develop products and salespeople may participate in implementation.
That shift is also changing how the company recruits. Scardino said Salesforce is increasingly hiring for “learning velocity”—a candidate’s ability to learn, unlearn and relearn—rather than relying solely on credentials or years of experience.
Candidates may be asked what they have built with AI or what they would build if given the opportunity. Salesforce also expects employees in every function to demonstrate a builder mindset, even if they are not in traditionally technical roles.
The company is bringing in nearly 2,000 new graduates and interns and pairing emerging talent with experienced employees through reverse mentorship. Salesforce has also created a workforce innovation team that breaks jobs down into individual skills, identifying cases where employees may need to acquire only one additional capability to qualify for several adjacent roles.
One question Salesforce has not yet answered is how employees should be compensated when agents allow them to produce more or manage substantially greater capacity.
Valenzuela said the company is experimenting with outcome-based compensation and considering whether agentic managers should be paid differently.
“I don’t think we have the perfect answer yet,” he said. “But we are absolutely looking into paying for outcomes.”
For Salesforce, the experiment suggests that deploying agents is only the beginning. Companies must also decide who manages them, how employees advance alongside them and what a job means when human and digital labour increasingly share the work.