The workforce claim also deserves disciplined scrutiny. “Hours given back” is not a benefit until leaders decide how that capacity is redeployed and how it will be measured. CIOs should press business sponsors to define target outcomes in advance: cycle-time reduction, case volume handled per FTE, improved service levels, lower contractor spend, or slower backfill hiring. Without that linkage, agent deployments risk becoming a margin-protection story for vendors rather than a value-realization story for the enterprise.
Architecturally, Salesforce is positioning itself as an AI backbone of data, context, workflow, and semantic control. That may be compelling, but it raises a familiar management challenge: deeper dependence on one platform while the application estate remains heterogeneous. MCP access, headless application patterns, and Slack-centered development workflows may improve reach, yet they also increase integration, governance, and security complexity across boundaries.
- Clarify pricing exposure: what spend moves from licenses to consumption, and what guardrails exist?
- Define productivity capture: where will released capacity go, and who owns the benefit target?
- Test operating-model fit: does Slack-based human-agent collaboration strengthen delivery governance or bypass it?
SAN FRANCISCO — CEO Marc Benioff’s keynote at this week’s Dreamforce focused on a continuing Salesforce franchise and upcoming innovations. The company unveiled new agents, Slackbot, Agentforce Coworker and highlighted the Claudeforce announcement from several weeks ago. Claudeforce will be charged on a per-user, per-month basis. Regardless, there are lingering questions for CIOs.
Please stop talking about the SaaSpocalypse
Benioff, who’s also chair and co-founder of Salesforce, has argued that the SaaSpocalypse hasn’t touched the class of software the company makes. “AI is not going to eliminate our software or impact our economic rents,” he said.
In August, Salesforce’s second-quarter earnings call included customer comments from David Friedberg, CEO and co-founder of Ohalo Genetics, an agricultural biotechnology company. Friedberg and Benioff shared with investors how the company originally tried another CRM vendor, experimented with vibe coding its own version and then turned to Salesforce.
“We realized pretty quickly there are certain, I would say, software applications that are maybe verticalized where you have to build custom workflows that make sense for your particular business,” Friedberg said during the call. “But then there’s platform software. And platform software, we realized pretty quickly, is really what we need to build our entire company around. And Salesforce is what we kind of decided to go with on CRM.”
It has been this author’s view that replacing major systems of record is unlikely — CIOs and their teams have too much capital invested in these systems to turn back now and be responsible for R&D, productization and maintenance of major application software.
With that said, Benioff noted two interesting things: First, customers are using more of Salesforce than they used prior to the introduction of AI agents. Studies of Microsoft Office found that users utilize only 20% of its capabilities. This kind of analogy could explain the result Benioff is touching on. Meanwhile, AI applications are increasingly using the Model Context Protocol (MCP) to access Salesforce data and trigger actions.
Worries persist around software rents
Many industry analysts, including this author, have worried that software rents will decrease as the number of users — especially occasional users — needing seats declines. Salesforce could offset the potential revenue loss by charging for MCP calls. If this becomes the case, will it create a predictability problem for CIOs without the ability to control expenses? It will be interesting to see how this lands.
Second, Benioff claimed in passing that the company hasn’t seen customers eliminating employee headcount due to the advent of agents.
Instead, “the hours are being given back to employees,” he said. “We are really unlocking capacity limits.”
But organizations need a concrete payoff with AI , which they could achieve by not replacing workers who leave or by reducing the number of entry-level workers they hire. Recent polls show workers are worried about AI becoming a job killer. In our research, Dresner Advisory Services has found that achieving operational excellence is the top strategic business outcome driving most organizations.
Indeed, according to Gregory Daco, chief economist at EY-Parthenon, “productivity growth protects margins, not income. Margins hit a record 14.9% of GDP, while the labor share fell to 52.8%, the lowest level since the government started counting in 1947.” This implies hiring could indeed be slowing with the addition of AI capabilities.
Redefining the Salesforce stack
Benioff then took a bit of a swipe at large language model companies, saying models alone will not deliver what future customers want. He said companies will need a supporting platform including headless applications, work processes, data with trusted context, a semantic layer or shared business glossary (a recast mission for Tableau), and a customer 360 . This creates, in Benioff’s view, the digital backbone companies need for the AI era.
Building on this backbone, of course, are agents. At Dreamforce, Salesforce introduced several anthropomorphically named agents such as Albert, a personal co-worker; Piper, a lead qualification agent; and Casey, a help desk agent. With the expanded Anthropic partnership, the company says Claude is now for knowledge workers. Benioff said he believes Claudeforce will drive organizations to AI-based transformation — this is about more personal or managerial productivity.
Slack: An evolving agenda
Finally, Salesforce wants Slack to be where agent-based coding will happen, a shift that clearly matters to CIOs and their app development teams.
The Slack team argued that the collaboration platform will become the space where business and coding teammates will meet to build apps together. They are pushing for what they call a “multiplayer runtime.” Here, agents, business leaders and developers specify and code together. The Slack team said this change will democratize software building and drive coding collaboration that hasn’t occurred before.
While this is interesting, I did not explicitly hear about the kind of innovation with Slack that the authors of “HBR Guide to Generative AI for Teams” describe. This is a big opportunity, too.
When I asked, Joe Inzerillo, president of enterprise and AI technology at Salesforce, said, “It’s easy to think about AI as a solo act, something that just replaces the work we used to do ourselves — that’s the wrong picture. Think about it this way: humans for impact, agents for scale. And multiplayer work is where we see this as a big unlock. Agents take on the context gathering, the follow-through and the grunt work, so people can focus on what we’re actually great at: setting direction, making the hard calls and doing the creative and strategic thinking that machines can’t. When agents work in the open, right alongside people, AI stops being a shortcut for one person and becomes a force multiplier for the whole team.”
Parting Words
It seems clear from Dreamforce that we are at a moment of change for Salesforce and its customers — and the changes being enunciated will bleed into the CIO agenda.
As application software becomes headless and addressed by MCP calls, how will Salesforce charge and keep CIOs whole? Clearly, CIOs will assemble an application stack as in the past, but it will be heterogeneous. This means integration and data problems will persist. What will need proof is Slack’s role in coding; regardless, its role on the human side of AI is clear. This means enabling teams to function from in-person meetings to their prep and postmortems.
Enjoyed this article? Sign up for our newsletter to receive regular insights and stay connected.