Applications are not dying, and customer relationship management software continues to drive growth despite the rise of artificial intelligence. Salesforce CEO Marc Benioff stated that agents accessing the company’s system are causing usage to explode, countering fears that AI would render traditional software obsolete. The San Francisco-based vendor reported that agentic use of its applications surged sixfold during the second quarter of its 2027 fiscal year, covering the three months ended July 31, driven by Model Context Protocol and command-line interface calls. This data challenges the narrative of a “SaaSpocalypse” where AI models would eliminate the need for CRM platforms.
Salesforce has expanded its strategic partnership with Anthropic, the maker of the Claude model, to deepen the integration between the two platforms. The collaboration, named “Claudeforce,” provides pilot customers with access to a Salesforce in Claude plugin containing 37 prebuilt sales skills. This tool allows sellers and agents to reason over live revenue context and automate pipeline updates. The company has also added custom models to its platform and marketplaces like Hugging Face.
Related: Partners see Supermicro deal easing Cisco server delays
Leah Sand, senior vice president of the AI-first marketing and engagement practice at Perficient, explained that customers want to solve for asset variations at scale. Her team helps clients address execution gaps by determining when to use generative tools versus automation. Sand noted that while AI strategies like answer engine optimization and generative engine optimization are becoming more complex, the focus remains on closing execution gaps. The solution provider has leveraged alliances with Salesforce and startups like Gradial and Swantide to offer these capabilities.
Enterprise Spending Shifts to Application Layers
Benioff stated that he does not foresee enterprises setting up their own AI model expertise and maintenance operations. Instead, customers prefer consuming models through packaged software like Salesforce and Slack. The CEO argued that the highest level of value is at the application layer, where users should not have to worry about the underlying hardware infrastructure. He emphasized that customers focus on getting applications to work rather than understanding the specific semiconductors or data centers used. This perspective suggests a continued reliance on established platforms to manage the complexities of AI integration.
Related: IBM CEO Highlights Opportunities in Power Storage Quantum
Benioff Doesn’t See Customer Focus On Models Benioff said he does not foresee enterprises setting up their own AI model expertise and maintenance operations, instead consuming models through packaged software like Salesforce in the front office and Slack as a collaboration application. Customers tend not to know the semiconductors they use and which data center, router, switch, fiber, cable and interconnect they use, instead just wanting their applications to work. Salesforce customers have not asked him about something as specific as open-source models and frontier models, with room for a variety of model types in the market, the CEO said. “At some level, we too will get abstracted back to kind of how the way the world used to be,” Benioff said. “I believe the highest level of value is at the application layer. So, I couldn’t tell you from my phone all the way through Salesforce and Slack all the way down to what the hardware infrastructure is, and I shouldn’t have to worry about that.” He continued: “I don’t think that’s where customers’ heads are at. At least when they talk to me, that’s not what their heads are at.”
