In the autumn of the previous year, Salesforce introduced a new artificial intelligence agent so astonishingly lifelike that it appeared less as a routine technological debut and more as an imaginative projection of what the future might hold. During a meticulously choreographed demonstration, the company presented how the luxury fashion retailer Saks Fifth Avenue had embraced Salesforce’s premier AI software suite, known as Agentforce, to design a digital persona named “Sophie.” This virtual assistant was endowed with a gently upbeat tone and an almost disarming patience—attributes that mirrored the temperament of an ideal human customer service representative. In the presentation, a Salesforce executive dialed into Saks’ customer hotline, only for Sophie to answer promptly, exuding friendliness and competence. She proceeded to recommend a sweater based on the executive’s historical purchase data and effortlessly discussed potential delivery alternatives. Through Sophie, Salesforce sought to embody the immense promise of AI-driven agents—versatile programs capable of learning, reasoning, and conversing in ways that rival human fluency. Few advocates have been as enthusiastic about such possibilities as Marc Benioff, Salesforce’s charismatic chief executive officer.
Shortly after unveiling Agentforce, Benioff penned an impassioned essay in the pages of *Time Magazine*—a publication he owns—where he heralded AI agents as the core of “a revolution that will fundamentally redefine how humans work, live, and connect with one another.” He proclaimed that Salesforce would stand at the forefront of this transformation and personally champion its rise. Around the same period, during a conversation with *Business Insider*, Benioff enthusiastically emphasized that Sophie was already operational on the Saks website, presenting the AI as a concrete realization of Salesforce’s capabilities. “This,” he explained, “is precisely what our customers can accomplish with Salesforce’s tools”—a veiled contrast intended to highlight the supposed limitations of rivals like Microsoft, which, as Benioff pointed out, “doesn’t have these examples, actually.” The optimism was infectious. In the weeks following Agentforce’s dazzling debut, Salesforce’s stock price skyrocketed by more than fifty percent, achieving an all-time peak by December.
Yet, only a year later, the euphoria appeared to have faded. While competing firms’ stocks continued to ascend, Salesforce’s share price sagged noticeably. Analysts, investors, clients, and even some company insiders began to wonder whether Benioff had wagered too heavily on a technology that had yet to deliver its promised returns. Public filings indicated that fewer than half of Salesforce’s 12,500 Agentforce customers were actually paying for the product. Moreover, less than two percent of all customers were conducting over fifty Agentforce-based interactions per week, according to individuals familiar with internal figures—a figure suggesting that widespread practical adoption remained a distant goal. A Salesforce spokesperson, however, cautioned that those statistics reflected only one interpretation of the data, arguing that they failed to portray the tool’s full usage or the multifaceted ways customers were engaging with it.
Sophie, once the symbol of AI’s near-magical possibilities, had also undergone a quiet transformation—one that diluted her original allure. When customers now call Saks’ hotline, they are greeted by a far more mechanical voice that can perform basic automated tasks before redirecting callers to human operators. This experience, observers noted, is scarcely distinguishable from the rudimentary interactive voice systems that banks and airlines have utilized since the 1980s. In August, Saks pulled focus toward a different collaboration, announcing a partnership with Amazon and AI specialist NLX to develop what it labeled a “new AI-powered virtual voice assistant named ‘Sophie.’” That phrasing alone hinted that the earlier Sophie had effectively been shelved or replaced.
Adam Evans, a senior executive vice president at Salesforce and the head of its Agentforce division, told *Business Insider* he could not comment on the specifics of the Saks deployment. Nonetheless, he insisted that substantial and measurable value was being generated for thousands of customers across numerous industries. Another corporate representative added that Agentforce had already powered over three billion interactions globally, enabling organizations as varied as Williams Sonoma, Singapore Airlines, and South American retail chain Falabella to uncover meaningful benefits as they scaled their use cases. Even multinational giant PepsiCo had joined the roster: its transformation officer, Athina Kanioura, reported that Agentforce was helping 1.5 million retail outlets manage restocking for PepsiCo products, with ambitions to quintuple that reach to five million stores by 2026.
Still, Salesforce’s growing pains were hardly unique. Across the enterprise technology sector, enthusiasm for generative AI has often outpaced verifiable productivity gains. A July report from the Massachusetts Institute of Technology revealed that, despite corporate expenditures of between $30 and $40 billion on generative AI initiatives, roughly ninety-five percent of surveyed organizations had yet to record any discernible return on investment. Similarly, research firm Gartner forecast that by the end of 2027, more than forty percent of projects reliant on agentic AI concepts would likely be terminated, hampered by spiraling expenses, ambiguous value propositions, or inadequate frameworks for managing risk.
Gil Luria, who leads technology research at D.A. Davidson, observed that Salesforce’s challenge did not stem from falling technologically behind its peers but from the strategic intensity of its gamble. “The company bet the farm,” he remarked, suggesting that Salesforce’s fixation on pioneering an unstable frontier came at the expense of its historically strong core operations. “They became so focused on their AI product while the rest of their business was decelerating sharply,” he said, concluding that Salesforce had devoted immense attention to the novelty of AI while neglecting the foundation that once secured its dominance. A company spokesperson countered, asserting that Agentforce was inseparable from Salesforce’s main business model and in fact enhanced every customer-facing application by making it “agentic,” or capable of adaptive intelligence.
Nevertheless, some insiders have noted that Salesforce occupies a more precarious position than competitors fortified by diversified revenue streams. “Salesforce is uniquely vulnerable because we don’t have a cloud business to fall back on,” one senior employee confessed. Firms like Amazon and Microsoft, with their massive and profitable cloud divisions, could afford far more patience in weathering AI’s uncertain economics.
According to internal strategic plans, Agentforce has been designated as the company’s top operational priority. The soon-to-be-finalized annual plan, due for completion by February, explicitly calls for Salesforce to “win the enterprise agent wars” as its foremost objective under the banner of creating “the agentic enterprise.” However, the market’s verdict has been less than encouraging: Salesforce’s stock has fallen more than twenty percent this year alone, while peers such as Oracle and Microsoft have enjoyed gains exceeding fifty and twenty-four percent, respectively. The company’s revenue growth rate has also dwindled into single digits—a first since going public. Adding further tension, an activist investor that had previously compelled Salesforce to undergo a sweeping cost-reduction campaign in 2022 recently increased its stake by nearly half, suggesting renewed scrutiny from the financial community.
Interviews with numerous current and former employees reveal deep internal strain. Teams continue to struggle under the weight of Benioff’s public commitments, scrambling to deliver a product capable of matching the grand claims showcased during demonstrations and at *Dreamforce*, Salesforce’s colossal annual conference. Insiders describe a persistent gap between showmanship and substance: software demos frequently highlight capabilities that are months or even years from tangible reality, and product roadmaps often shift before these promised features materialize. As one senior staff member admitted, “It’s incredibly hard—even for people on the inside—to know the difference between what we highlight in a demo, what’s on the future roadmap, and what actually exists in production.”
For many customers, deploying Agentforce has proven far more arduous than they anticipated. Kristi Valente, a Salesforce administrator with nearly two decades of experience, explained that the slick presentations made implementation appear both straightforward and cost-efficient. Her company, which purchased Agentforce in April, has yet to operationalize it. “If you are a normal business with normal administrators,” she warned, “you simply do not have the expertise to set this up.” Even after contracting external consultants, she found that few truly understood the software’s novel architecture. In her words, “It’s too new for anyone to be an expert.” Salesforce has since acknowledged such difficulties, noting that “a tremendous amount has been learned about the complexities of building agents.” Executives, including Adam Evans, stress that no two use cases are alike and that scaling AI agents involves far more than crafting clever prompts. Accordingly, Salesforce has introduced a “reasoning engine” designed to yield more predictable, reliable responses—in contrast to chatbots that may prioritize creativity at the expense of accuracy.
The confusion has sometimes extended to Salesforce’s own sales teams. Employees participating in mandatory Agentforce training sessions described the software as labor-intensive and less intuitive than advertised. In one exercise, staff were asked to construct a virtual concierge for an imaginary hotel chain capable of arranging dinner bookings and beach chair rentals. The setup process, a participant recalled, demanded tremendous effort, even for these relatively basic tasks. “Anyone who took that training knows Agentforce isn’t ready for primetime,” the employee said candidly. Salesforce’s official response emphasized the company’s proactive investment in hands-on training for its more than seventy-five thousand employees, asserting that mastery of large language models represented “a profound paradigm shift” requiring equally profound educational and managerial adaptation.
Industry analysts share this cautious tone. Kash Rangan, a top software analyst at Goldman Sachs, remarked after a recent technology conference that Wall Street remains skeptical of the lofty AI projections offered by software executives. He suggested that companies like Salesforce still have a long journey before achieving scalable impact. However, Rangan noted that Benioff, typically known for bold pronouncements, appeared somewhat more grounded and pragmatic during the conference, acknowledging the complexity of the task ahead.
Even so, Benioff’s trademark flair for showmanship persists. At the same Goldman event, he mused about Salesforce’s next frontier—a breathtaking target of $100 billion in annual revenue, almost triple its current earnings. The aspiration crescendoed the following month at *Dreamforce*, the company’s extravagant extravaganza that transforms San Francisco into an eclectically branded wonderland. Equal parts business symposium and carnival, the three-day spectacle drew roughly fifty thousand participants and featured a blend of celebrity appearances, concerts, merchandise, and self-congratulatory rituals. Slogans like “You have an agentic aura” adorned the venue, and actor Matthew McConaughey—on Salesforce’s payroll as its “creative director”—recited poetry with journalist Maria Shriver. Guests were serenaded by Metallica, adorned with golden hoodies as symbols of community achievement, and treated to surreal moments like Benioff kissing entrepreneur David Sacks on the cheek. The event, according to insiders, may cost the company upward of $100 million annually. Yet despite the carnivalesque ambience, seasoned employees described Dreamforce as “a marketing and sales event, not an engineering one,” a carefully orchestrated tool to generate hype and strengthen bonds with high-profile clients.
During his keynote, Benioff once again demonstrated his charismatic ability to galvanize loyalty. He mingled freely with celebrities such as will.i.am and tech luminaries including Michael Dell, weaving personal anecdotes into corporate messaging. He acknowledged the company’s imperfections in rolling out Agentforce but swiftly redirected attention to its record-setting pace, calling it “the fastest-growing product in our history.” Later, at an investor day held alongside Dreamforce—the first such gathering in three years—Benioff and his leadership team presented updated projections, targeting $60 billion in revenue by 2030 and a return to double-digit growth. Some analysts emerged reassured, suggesting that while generative AI adoption had been slower than promised, Salesforce remained a formidable contender, not a laggard.
The overarching question, however, lingers: how long can Benioff’s relentless charisma and public optimism sustain investors’ patience as Salesforce fights in the so-called “agent wars” against resource-heavy competitors like Microsoft, Google, and Amazon? For now, some inside sales teams report a rekindled sense of momentum in the days following Dreamforce. “Hype needed to build,” one salesperson admitted, implying that the spectacle may have indeed bought Salesforce a little more time to transform its ambitious vision into the tangible progress the market is waiting to see.
Sourse: https://www.businessinsider.com/inside-salesforce-struggles-agentforce-flagship-ai-agent-wars-benioff-2025-11
