Salesforce’s AI reckoning: Even Salesforce execs don’t trust the bots
Correction (updated Dec 30): This article was updated to include a comment from Allen Tsai, Senior Director of Corporate Communications at Salesforce.
Salesforce (CRM) stock has largely missed out on the AI-driven market rally of 2025, weighed down by slowing growth and a bumpy rollout of its flagship agentic AI offerings.
Now, company executives are openly cautioning that large language models, at least in their current form, cannot be fully trusted to autonomously run enterprise workflows.
According to a recent report from The Information, Salesforce executives told enterprise customers that Agentforce, the company’s platform for deploying AI agents, performs better when it doesn’t rely exclusively on large language models.
Instead, Salesforce is developing a hybrid approach that combines LLMs with more deterministic, rules-based systems and traditional software logic, according to The Information’s Amir Efrati.
"While LLMs are amazing, they can’t run your business by themselves. Companies need to connect AI to accurate data, business logic, and governance to turn the raw intelligence that LLMs provide into trusted, predictable outcomes," Allen Tsai, Senior Director of Corporate Communications at Salesforce, told InvestorsObserver.
The goal is to reduce errors, hallucinations, and unpredictable behavior that can emerge when LLMs are used as fully autonomous decision-makers.
Salesforce has placed AI at the center of its long-term growth strategy, unveiling Agentforce in 2024 as a potential catalyst for revenue acceleration. However, investor enthusiasm has lagged.
Since the product’s debut, Salesforce shares have underperformed both AI-exposed peers and the broader market, even as enthusiasm around generative AI lifted much of the tech sector.
As Business Insider previously reported, adoption of Agentforce has fallen short of expectations. Fewer than half of customers with access to the product are paying for it at scale, and usage among active customers has remained light, according to people familiar with the matter.
The result has been a growing disconnect between the hype surrounding agentic AI and the slower, more cautious pace of real-world enterprise adoption, leaving Salesforce caught between bold promises and the practical limits of today’s AI technology.
CRM’s underwhelming performance
As Business Insider reported, Salesforce shares surged after the debut of Agentforce, reaching an all-time high in December of last year as investors initially welcomed the company’s push into agentic AI.
Since then, however, CRM stock has struggled. CNBC has described it as the worst-performing large-cap technology stock of 2025. Shares are down about 23% over the past 12 months, compared with a roughly 16% gain for the broader market.
Salesforce showed some signs of stabilization in its fiscal third quarter. The company said revenue tied to Agentforce had surpassed $500 million, reflecting early commercial traction.
Still, total revenue came in just shy of Wall Street expectations, even as earnings per share comfortably beat forecasts, reinforcing the view that cost controls are outpacing top-line momentum.
Taken together, the results have done little to dispel investor skepticism that Salesforce’s AI ambitions will translate into sustained growth anytime soon.