Key Takeaways
- Salesforce (CRM) shares have tumbled approximately 30% in 2026 amid concerns that artificial intelligence will undermine traditional SaaS business models
- CEO Marc Benioff defends the company’s position, claiming Salesforce’s AI prospects have “never been greater”
- Agent Albert, a new artificial intelligence platform, is scheduled to debut before year-end
- The company’s Agentforce product now serves 23,000 customers out of 150,000 total, delivering up to 40% reductions in support tickets at select organizations
- The platform handled 2.4 billion Agentic Work Units during the most recent quarter, representing a 57% sequential increase
Salesforce (CRM) has endured significant turbulence throughout 2026. Shares have declined approximately 30% since January, pressured by mounting investor anxiety that artificial intelligence will fundamentally undermine the software-as-a-service framework the company pioneered.
The central anxiety revolves around a simple calculation. Salesforce generates revenue through per-seat subscriptions — billing organizations according to employee count. When AI enables companies to accomplish equivalent work with smaller teams, demand for additional seats naturally contracts.
CEO Marc Benioff rejects this narrative entirely. “People think we have our back against the wall when in fact the opportunity has never been greater,” he stated in a Wall Street Journal interview.
The challenges extend throughout the software industry. The iShares Expanded Tech-Software Sector ETF (IGV) has dropped 20% during the identical timeframe, with the most vulnerable SaaS companies experiencing declines approximately double Salesforce’s losses.
Salesforce’s AI transformation began earnestly in early 2023, when Benioff convened approximately 40 senior leaders for an intensive three-day planning retreat at Salesforce Tower to restructure the organization’s strategic roadmap around artificial intelligence. Weekend planning sessions continued for several months afterward.
This initiative yielded Agentforce, which launched in late 2024. The platform enables organizations to deploy autonomous AI agents for functions including customer service resolution, sales lead qualification, and IT support requests. Currently, 23,000 of the company’s 150,000 client base has adopted the technology.
Tangible outcomes are emerging. At Pearson, Agentforce agents manage order inquiries, process refunds, and resolve access credential issues — increasing the proportion of customer inquiries resolved without human intervention by 40%. PenFed Credit Union reduced IT support volume by 40% through an agent handling password resets and account access restoration.
Current Limitations of Agentforce
Implementation hasn’t been flawless. Pandora’s chief digital officer reported that Agentforce encounters difficulties with ambiguous or sophisticated customer inquiries — such as providing jewelry recommendations when customers mention “my wife likes dogs.” Intricate problems continue requiring human expertise.
Initial client feedback also highlighted extensive data preparation requirements before AI systems could function effectively. Salesforce addressed this by developing a data-integration infrastructure within its technology platform and acquiring specialized firms in data management and AI-powered sales.
Agent Albert Launch and Revenue Model Evolution
Before 2026 concludes, Salesforce intends to introduce Agent Albert, a sophisticated AI platform that analyzes user patterns and executes actions autonomously. Named after Einstein, the company’s mascot, the platform represents three years of internal engineering effort.
Regarding monetization, Salesforce transitioned away from exclusive seat-based pricing approximately twelve months ago. The current hybrid framework maintains seat licenses while adding usage-based charges for Agentforce actions. A newly introduced metric called Agentic Work Units (AWUs) quantifies platform activity: 2.4 billion AWUs were executed during the previous quarter, reflecting a 57% sequential growth rate.
Benioff further contends that organizations cannot simply build custom CRM solutions through rapid AI coding. The data security protocols, brand protection mechanisms, and regulatory compliance features Salesforce has developed over decades remain difficult to duplicate, he maintains — even leveraging Claude Code or OpenAI’s Codex.
Salesforce has committed over $300 million to Anthropic investments beginning in 2023. In February, when both organizations announced Claude Cowork integration with Salesforce applications, CRM shares surged 4%.
Stifel analysts observed that “CIOs and CTOs prefer a unified platform that integrates agents, actions, data, and workflows” — a positioning Benioff emphasizes as discussions about SaaS viability intensify.
Data from Andreessen Horowitz indicates that AI-focused enterprise customers expanded their median Salesforce expenditures by 3% during the preceding three-month period.
