Salesforce, Inc. (NYSE: CRM) heads into the final stretch of 2025 with a familiar split-screen narrative: strong financial execution and accelerating “agentic” AI adoption on one side, and fresh debate over how enterprises will trust, control, and ultimately pay for AI agents at scale on the other.
With U.S. markets closed for Christmas Day, investors are effectively marking time at the latest close: Salesforce shares ended Dec. 24 at about $265, up modestly on the session. [1] On Dec. 25, the day’s Salesforce conversation is being shaped less by price action and more by new reporting and analysis about how the company is evolving its Agentforce strategy—especially in regulated, high-stakes workflows where reliability matters as much as intelligence. [2]
Below is a comprehensive roundup of the key Salesforce news, forecasts, and analyst takes circulating as of Dec. 25, 2025, and what they could mean for 2026.
What’s driving today’s discussion: AI agents, but with tighter guardrails
One of the most talked-about threads on Dec. 25 is reporting that Salesforce executives have been signaling more caution about relying on “open-ended” large language model behavior for Agentforce, emphasizing more deterministic approaches for critical tasks. [3]
That narrative aligns with what Salesforce is now putting directly into product messaging:
- On its Agent Script page, Salesforce explicitly frames the goal as blending LLM creativity with deterministic control—promising to “eliminate the inherent randomness” of large language models so workflows follow the same steps every time. [4]
- Separately, Salesforce’s developer documentation for the Einstein Trust Layer describes protections such as grounding outputs in CRM data, masking sensitive data, audit trails, and “zero data retention” agreements with third-party model partners. [5]
Why this matters: the enterprise market has been enthusiastic about AI agents in principle, but real production deployments often bottleneck on governance—accuracy, auditability, permissions, and predictable behavior. The Dec. 25 debate is less “are agents useful?” and more “what form of AI is safe enough to run the business?”
Salesforce’s latest results still set the tone: raised outlook as AI adoption “picks up steam”
The most recent hard catalyst for Salesforce remains its FY26 Q3 earnings (reported Dec. 3), where the company raised its full-year forecast and highlighted rapid growth in Agentforce-related recurring revenue. [6]
Key points investors continue to cite:
- Salesforce raised FY26 revenue guidance to roughly $41.45B–$41.55B, up from the prior range reported by Reuters. [7]
- It also lifted adjusted EPS expectations to around $11.75–$11.77. [8]
- Management pointed to Agentforce and its data offerings (often referenced together as Agentforce + Data 360) as a central driver of the AI monetization story. [9]
Even amid strong guidance, Salesforce’s AI narrative hasn’t been purely celebratory in 2025. Earlier reporting also captured skepticism across the broader market about whether enterprise AI spending is translating cleanly into vendor revenue—an undercurrent that still informs the “prove it in production” focus now showing up in the Agentforce trust/controls conversation. [10]
Agentforce and Data 360: the metrics Wall Street keeps watching
Salesforce is trying to make Agentforce feel measurable, not mystical—so it has leaned heavily into operational and revenue indicators.
In its FY26 Q3 materials, Salesforce said:
- Agentforce & Data 360 ARR was nearly $1.4B, up 114% year over year. [11]
- Agentforce ARR surpassed $500M in Q3 and was up 330% year over year, according to Reuters’ recap of the company’s results and commentary. [12]
- Salesforce also highlighted deal activity and usage at scale (including “paid Agentforce deals” and very large token volumes), framing Agentforce as a major platform transition rather than a feature add-on. [13]
A notable nuance as 2026 approaches: the market is increasingly differentiating between “AI adoption” and “AI monetization.” Salesforce’s story is that adoption is real—and the monetization model is evolving to meet how CIOs actually buy.
The monetization pivot: from “per conversation” anxiety to predictable pricing
Another major theme in late-December coverage is pricing—and whether Salesforce can make AI agent economics feel predictable enough for procurement teams.
TechRadar reports Salesforce has been signaling that per-user (seat-based) pricing could become a “new AI norm,” citing customer preference for predictability over consumption-style models. [14]
This pricing conversation also links back to the Dec. 25 “trust and randomness” theme:
- If agents are constrained with deterministic scripting (and better observability), they’re easier to cost, govern, and roll out broadly.
- If agents behave too unpredictably, enterprises not only worry about mistakes—they worry about runaway cost, especially under usage-based pricing.
In other words, for 2026, Salesforce’s Agentforce challenge may be as much packaging and procurement as it is model quality.
Partnerships: OpenAI, Anthropic, and the push to be “LLM-agnostic” (but best-integrated)
Salesforce is also positioning itself as the control plane for enterprise AI agents—where customers can use leading models, but keep governance and data policy inside Salesforce.
Reuters reported in October that Salesforce expanded partnerships with OpenAI and Anthropic to integrate their models into the Agentforce platform, with an emphasis on enterprise-grade use cases and regulated industries. [15]
A key strategic message from that coverage: Salesforce wants customers to view Agentforce as the place where frontier models become enterprise tools—connected to CRM context, workflow permissions, and audit trails—rather than standalone chatbots. [16]
“Meet users where they work”: Agentforce Sales arrives in ChatGPT
Salesforce’s distribution strategy for agents is also evolving beyond its own UI.
On Dec. 17, Salesforce announced an Agentforce Sales app in ChatGPT, aiming to let sellers pull live Salesforce context into ChatGPT and take actions without constantly switching tabs—what Salesforce frames as reducing “toggle tax.” [17]
Salesforce said an open beta for eligible customers was available starting that day via the ChatGPT app directory, positioning this as a milestone in bringing Salesforce workflows into the tools users already inhabit. [18]
For SEO-minded readers tracking “Salesforce + ChatGPT integration,” the key takeaway is that Salesforce is not just competing on model quality—it is competing on workflow placement.
Customer adoption spotlight: Novartis and the U.S. Department of Transportation
Two customer announcements from mid-December are particularly relevant to the Dec. 25 narrative because they emphasize regulated, mission-critical use cases—exactly where trust, governance, and determinism matter.
Novartis: global rollout over five years
Salesforce said Novartis selected Agentforce Life Sciences for Customer Engagement, building on existing investments across Agentforce Health, Data 360 for Health & Life Sciences, MuleSoft, and Agentforce Marketing. [19]
Salesforce added that Novartis plans to roll out the platform globally over the next five years, aiming to unify engagement across marketing, sales, patient services, medical, market access, and other stakeholders. [20]
USDOT: agents for citizen support, incident mitigation, and grants
Salesforce also announced an expanded initiative with the U.S. Department of Transportation (USDOT), describing plans to deploy Agentforce to handle routine tasks, provide around-the-clock citizen support, and generate alerts with proposed mitigation strategies for incidents. [21]
USDOT’s rollout description includes use cases like complaint management, inspections and safety data exchange, and accelerating grant processing—positioning agents not as experiments, but as operational capacity. [22]
“Trust is the product”: why Salesforce is emphasizing deterministic controls now
The Dec. 25 reporting about declining trust in LLM behavior is noteworthy not because Salesforce is “abandoning” AI—Salesforce has explicitly denied it is backtracking in the reporting cited—but because it suggests the company is recalibrating what “agentic” means in enterprise settings. [23]
Seen through Salesforce’s own product language, the company is increasingly selling a formula like this:
LLMs for language and reasoning + deterministic scripting for policy + Trust Layer for governance + Data 360 for context. [24]
For 2026, this may become Salesforce’s differentiator against competitors who lead with model capability but offer thinner control layers.
Salesforce stock forecasts as of Dec. 25: what analysts and market trackers are saying
While price targets vary by firm and update cadence, the consensus across major tracking pages leans bullish into 2026:
- StockAnalysis shows a consensus “Buy” rating and an average price target around $324, implying roughly low-20% upside from current levels, with a wide range (roughly low $200s to ~$400+ depending on analyst). [25]
- MarketBeat’s summary view similarly posts an average forecast around $326.68 (about 23% upside) and notes a “Moderate Buy” type consensus framing. [26]
At the same time, not all analyst narratives are aligned:
- One Barron’s argument published Dec. 24 frames Salesforce as ready to “emerge an AI winner,” suggesting the market may be overly bearish given margin improvements and Agentforce traction. [27]
- Another Barron’s piece earlier in December, citing a CIO survey, offered a more cautious angle—suggesting Microsoft may capture a larger share of AI budget growth and that Salesforce’s AI usage penetration could lag expectations. [28]
Bottom line: the Street’s forecast isn’t just a number. It’s a referendum on whether Salesforce can convert Agentforce enthusiasm into durable, budget-line-item revenue.
Valuation snapshot on Dec. 25: where Salesforce sits going into 2026
Valuation trackers also reflect the market’s current stance: Salesforce is priced like a mature software leader with an AI option value—but not like a “pure AI high-flyer.”
Multiples.vc lists (as of Dec. 25, 2025):
- Market cap around $249B and enterprise value around $248B
- EV/Revenue around 6.6x and a P/E around the low-20s (methodology and data sourcing attributed to providers like FactSet/Morningstar on the page) [29]
Those levels can be interpreted two ways:
- Bulls see room for multiple expansion if Agentforce monetization accelerates.
- Bears see a fair valuation until Salesforce proves AI is more than incremental attach.
Risks still in focus: security, integrations, and the “new attack surface”
Any “current” Salesforce outlook also has to incorporate security and platform risk—especially as agents and integrations increase system reach.
Reuters reported in November that Salesforce was investigating “unusual activity” involving certain Gainsight-published applications that may have enabled unauthorized access to some customers’ Salesforce data—while Salesforce also said it had no indication the issue was due to a Salesforce platform vulnerability. [30]
Separately, Reuters reported in October on claims by cybercriminals about stolen Salesforce-related records obtained by targeting Salesforce customers—claims Reuters could not verify, and where Salesforce said its systems were not hacked. [31]
As AI agents expand access across systems, these kinds of integration-layer risks become part of the enterprise buying calculus—not just the model’s accuracy.
What to watch next for Salesforce in early 2026
As of Dec. 25, the highest-signal checkpoints for Salesforce’s 2026 story look like this:
- Agentforce monetization clarity
Investors will watch whether pricing becomes simpler (seat-based or hybrid) and whether attach rates move from pilots to enterprise-wide deployments. [32] - Proof that “deterministic + agentic” scales in production
Salesforce is now explicitly marketing deterministic controls (Agent Script) to reduce randomness—expect more focus on observability, auditability, and repeatable outcomes. [33] - Regulated-industry wins and referenceability
Novartis and USDOT are the kind of lighthouse accounts that can drive follow-on demand—if Salesforce can document measurable ROI and safe deployment patterns. [34] - Ecosystem leverage (OpenAI, Anthropic, ChatGPT workflow placement)
Salesforce’s bet is that the “best” AI model will change, but the “best” enterprise control plane will win. [35]
Salesforce enters 2026 with raised guidance, expanding partnerships, and visible enterprise adoption—but Dec. 25’s most important signal may be philosophical: the company appears to be redefining “agentic AI” for the enterprise as AI you can constrain, audit, and budget for, not just AI that can converse. [36]
References
1. stockanalysis.com, 2. the-decoder.com, 3. the-decoder.com, 4. www.salesforce.com, 5. developer.salesforce.com, 6. www.reuters.com, 7. www.reuters.com, 8. www.reuters.com, 9. www.reuters.com, 10. www.barrons.com, 11. www.reuters.com, 12. www.reuters.com, 13. www.salesforce.com, 14. www.techradar.com, 15. www.reuters.com, 16. www.reuters.com, 17. www.salesforce.com, 18. www.salesforce.com, 19. www.salesforce.com, 20. www.salesforce.com, 21. www.salesforce.com, 22. www.salesforce.com, 23. the-decoder.com, 24. www.salesforce.com, 25. stockanalysis.com, 26. www.marketbeat.com, 27. www.barrons.com, 28. www.barrons.com, 29. multiples.vc, 30. www.reuters.com, 31. www.reuters.com, 32. www.techradar.com, 33. www.salesforce.com, 34. www.salesforce.com, 35. www.reuters.com, 36. the-decoder.com
