Salesforce’s Agentforce product maturity questioned as KeyBanc cites weak customer traction

Salesforce’s AI agent platform, Agentforce, is seeing weaker-than-expected customer traction, according to a recent KeyBanc Capital Markets investment research note, which attributed the slowdown in part to the product itself, stating that “Agentforce, as a product, just isn’t there” yet, following customer checks and a CIO survey.

“Our checks and customer conversations have not been strong, nor has the feedback been on Agentforce. What we can piece together in the disclosed numbers does not signal building momentum and, most recently, our CIO survey delivered another blow with Salesforce being a standout for the wrong reasons,” according to a Seeking Alpha that quoted a KeyBanc research note.

“We attend more Salesforce partner and customer events than any other company in our coverage, and feedback from those customers has been consistent in two ways: 1) customers’ data is not in order to do meaningful AI work; and 2) Agentforce, as a product, just isn’t there,” Seeking Alpha reported, quoting the KeyBanc note.

The KeyBanc note quoted by Seeking Alpha also pointed out that conversations with Salesforce “partners” indicate that Agentforce proof-of-concept deployments are only now starting to generate pipeline opportunities, while its CIO survey found more respondents expecting to deprioritize Salesforce within their IT budget than the other way around over the coming 12 months.

The findings in the research note stand in contrast to Salesforce’s sustained push to position Agentforce as its flagship enterprise AI platform. Since introducing the offering nearly two years back, the company has expanded it with new foundation models, integrations, deployment options, and pricing initiatives, most recently introducing its Headless 360 strategy to make Agentforce available beyond conventional CRM workflows through a more flexible consumption model.

Parts of that flexible consumption model and Agentforce pricing, which Salesforce is still evolving, have already come under scrutiny with industry analysts expressing concern that Headless 360’s monetization model could create budgeting headaches for CIOs by making AI spending less predictable and increasing pressure on IT leaders to demonstrate measurable business outcomes and return on investment before expanding deployments.

Concerns over product maturity

Those concerns around Agentforce’s evolving pricing model appear to be intersecting with the latest concerns about product maturity that KeyBanc analysts mention in their report.

“Three pricing model changes in roughly 18 months make procurement committees nervous, and if the commercial model keeps shifting, buyers question whether the product has stabilized either,” said Bhupendra Chopra, chief revenue officer at IT consulting firm Kanerika.

Salesforce’s latest consumption-based pricing model, Chopra pointed out, is a bigger concern: “It is harder to budget for than seat-based licensing. CIOs want a clearer line between spend and outcome. That line isn’t clear enough yet.”

Greyhound Research Chief Analyst Sanchit Vir Gogia, too, said that the pricing model is a fundamental issue for enterprises.

While Salesforce’s pricing model charges enterprises for AI activity, it leaves customers to determine whether those interactions ultimately translate into meaningful business outcomes, Gogia said.

That, according to Gaurav Parab, principal research analyst at NelsonHall, is slowing adoption because enterprise leaders, such as CIOs, are under pressure to evaluate TCO and expected ROI.

“Most enterprises first want confidence that AI deployments will generate measurable business outcomes before committing to broader rollouts,” he said.

Data Cloud and data readiness remain as challenges

Pricing, though, is just one piece of the equation.

The discussion about Agentforce adoption, analysts said, cannot be separated from Salesforce’s broader product strategy, which positions Agentforce alongside Data Cloud as the foundation for enterprise AI deployments.

“Data modernization has become one of the biggest determinants of adoption. Agentforce depends on trusted, unified enterprise data to generate reliable outcomes. For many organizations, preparing that data foundation through Data 360, integration, governance, and data quality initiatives represents a significant part of the implementation effort,” Purab said, backing the KeyBanc research note.

Chopra seconded that assessment, saying Data Cloud has effectively become a prerequisite for production-grade Agentforce: “We worked with a private equity fund administrator where the AI layer only became reliable once we had clean, structured data feeding into Salesforce consistently. Before that, even well-configured automation produced inconsistent outputs.”

In practice, that means many enterprises need to invest separately in Data Cloud, data integration, governance, and cleanup before Agentforce can be deployed reliably at scale, Chopra said.

Implementation challenges are slowing adoption

Product maturity aside, those implementation challenges, Purab pointed out, are also contributing to a slower pace of Agentforce adoption in enterprises than anticipated.

While interest in Agentforce continues to grow, enterprises, according to the analyst, are largely limiting deployments to targeted, high-value use cases while they establish trusted data foundations, integrate with existing systems, and demonstrate measurable business value.

More so because Data Cloud accelerates Agentforce once the foundation is coherent,  it cannot make incoherence disappear, Gogia pointed out.

Chopra, too, said his conversations with enterprise customers closely mirror Purab and KeyBanc’s findings: “The issue isn’t appetite. The issue is that their CRM data is fragmented, partially duplicated, and inconsistently structured. You can’t put an AI agent on top of that and expect reliable outputs.”

“Cleaning that up takes months. That work doesn’t show in a vendor’s deal count, which is why signed agreements and actual production deployments are two very different numbers right now.”

Timing issue or execution gap?

Despite all the challenges, though, Purab said that the slower pace of current adoption is not necessarily indicative of a long-term problem.

“I see it as a timing issue, but it has always been the case. Enterprise AI adoption has consistently followed the maturity of data, governance, and operating models. Salesforce will undoubtedly continue refining the product, pricing, and go-to-market approach, but the larger challenge lies in enterprise readiness,” Purab said.

“As organizations strengthen their data foundations and gain confidence in deploying AI responsibly, Agentforce adoption is likely to broaden significantly,” Purab added.

Chopra, in contrast, offered a more nuanced view: “While data readiness, which is a customer issue, will improve with time, Salesforce needs to fix its go-to-market strategy.”

“Three pricing changes in 18 months, a product that requires significant pre-investment before it delivers value, and implementation complexity that most mid-market buyers aren’t resourced for is definitely a positioning gap Salesforce needs to close,” Chopra said.

Regardless of whether the current slowdown proves temporary or structural, both analysts agreed that the next six to twelve months should provide a clearer picture of Agentforce’s trajectory.

For CIOs evaluating the platform, they said, the focus should be less on headline product announcements and more on tangible indicators of enterprise adoption and operational maturity.

“The key indicators will be an increase in enterprise-scale production deployments rather than pilots, broader adoption beyond customer service into other business functions, stronger customer references demonstrating measurable business outcomes, continued simplification of pricing and deployment models, and greater maturity around governance, security, and operating models for AI agents,” Purab said.

For Chopra, CIOs should go a step further by measuring how AI agents perform in production rather than simply tracking deployment numbers: “Containment rate in production — what percentage of agent interactions are resolved without human escalation — is the real performance signal, not token volume or deal count.”

Salesforce did not immediately respond to a request for comment.