SAN FRANCISCO — Artificial intelligence could generate new forms of work as it lowers the cost of performing complex tasks, according to a senior Salesforce executive, who argued that greater efficiency does not necessarily translate into reduced demand for human workers.
Jayesh Govindarajan, executive vice president for Salesforce AI and Agentforce, said the impact of AI on employment could follow the “Jevons paradox,” an economic principle under which improved efficiency and lower costs can drive greater consumption of a resource rather than reduce it.
“When the cost of a commodity goes down, more people want to use it. The demand actually goes up, not down,” Govindarajan said in an interview on the sidelines of Dreamforce 2026.
He cited the development of more efficient steam engines as an example. Although the technology reduced the amount of coal needed for a particular task, cheaper transportation encouraged more people to travel, ultimately increasing overall coal consumption.
“This is what we’re seeing with agentic systems too,” Govindarajan said. “As the cost of doing complex things has gone down… You can do things much more faster. Are you doing less of it though? You’re not. You’re doing more. You’re more busy than you were ever before.”
“People are needed more. Maybe in different kinds of roles and jobs, but I think that’s what we’re seeing. People need more agents, not less,” he added.
Govindarajan made the remarks when asked whether deployments of Salesforce’s Agentforce platform supported Nvidia chief executive Jensen Huang’s position that AI would expand work and create jobs instead of simply eliminating them.
While Govindarajan did not identify a specific Agentforce deployment that had directly produced new jobs, he said the technology was shifting bottlenecks and changing the nature of work performed by employees.
In software development, for example, he said AI now generates a “large percentage” of code, although human review remains necessary before it can be placed into production.
“It needs the right guardrails, which is really important to take into account. It needs to be reviewed before it goes into production,” Govindarajan said.
He said senior engineers remain responsible for determining what code enters production, how it is deployed and how it is improved. AI-assisted development has also allowed Salesforce teams to produce more code and spend more time understanding customers’ problems, he added.
“The senior engineers are doing a lot more than they ever were in our teams,” Govindarajan said. “They’re able to drive a lot more code into the system, but they’re still very much in the driver’s seat.”
For younger technology professionals, Govindarajan said AI tools have lowered the barriers to building products and services, but workers still need to develop the initiative to identify and solve problems.
“It’s never been easier to be a builder. It is such an amazing time to be a builder,” he said. “The tools and the barriers that existed before are no longer barriers because there are tools that overcome that.”
“My advice would be a high degree of personal agency. This desire to go solve a problem and go get something built out,” he added.
Govindarajan said AI agents are also progressing beyond simple request-and-response systems toward “long-horizon” tasks that can run over extended periods, retain information and coordinate multiple stages of work.
Salesforce is initially applying the capability to sales processes, where an agent could analyze a pipeline, identify prospects most likely to close, conduct research, initiate outreach and manage responses.
The company also announced multi-agent capabilities that would allow specialized AI agents to discover one another and transfer tasks based on their functions.
Govindarajan said these developments could enable companies to automate more complicated processes, although humans would continue to oversee areas such as accuracy, security, software review and production deployment.
