The number hits first: 15,000+ enterprise customers. Zero disclosed Agentforce revenue figures.
That's the gap in Salesforce's Q2 earnings narrative, and it's the loudest silence in enterprise software right now.
Salesforce positioned Agentforce—its enterprise AI agent platform—as the centerpiece of its earnings story. Not a side experiment. Not a roadmap item. The main event. But here's what the earnings report didn't say: how many conversations those agents actually handled, how much revenue they generated, or whether the per-dialogue pricing model is gaining real traction.
Speed is the asset, but silence is the warning.
I've spent years tracking how enterprise software vendors hype AI capabilities during earnings calls. The pattern is consistent: when metrics are strong, they lead with numbers. When they're not, they lead with narratives. Salesforce led with Agentforce's "early success" and "customer interest"—phrases that translate to "we don't have hard numbers to share yet."
The Pricing Paradigm Shift Nobody's Talking About
Here's what matters most about Agentforce, and it's not the AI itself. It's the pricing model.
Salesforce is charging roughly $2 per conversation for Agentforce. That's a fundamental departure from the traditional SaaS subscription model, where customers pay per seat regardless of usage. This is usage-based pricing for AI agents—customers pay for what the agent actually does, not for the right to potentially use it.
The shift from "per-seat" to "per-outcome" pricing isn't just a Salesforce strategy—it's a potential redefinition of how the entire SaaS industry charges for value.
Think about what this means. A company with 1,000 customer service reps paying $150 per seat per month is paying $150,000 monthly. With Agentforce at $2 per conversation, that same company might pay only for the 50,000 conversations the AI handles—$100,000. Cheaper on paper. But if those conversations spike to 500,000? That's $1 million. The cost structure becomes variable, unpredictable, and directly tied to AI adoption.
The house didn't build this pricing model to be cheaper. It built it to capture upside.
The Real Competitive Battleground: Data, Not Models
Let me cut through the noise on the competitive landscape. Salesforce doesn't have the best AI models. It doesn't need them.
Salesforce's moat is its CRM data—the largest enterprise dataset of customer interactions, sales pipelines, and service histories on the planet. That's the fuel AI agents need to function effectively in enterprise environments.
Microsoft Copilot has Azure's compute cost advantages. ServiceNow has ITSM expertise. AI-native startups like Decagon and Sierra AI move faster with specialized focus. But none of them have 15,000+ enterprise customers already running their core business processes on the platform.
The battle isn't about who has the smartest model. It's about who can deploy AI agents that actually work within existing enterprise workflows, access the right data, and deliver measurable ROI. Salesforce's advantage is integration depth—Agentforce doesn't just answer questions; it can trigger workflows, update records, and execute actions within the CRM ecosystem.
Microsoft is the real threat. Copilot Studio lets companies build similar agents inside Dynamics 365, and Microsoft's Azure OpenAI infrastructure gives it a cost per inference that Salesforce can't match. But Dynamics 365's data footprint remains far smaller than Salesforce's. That gap matters.
The BPO Disruption Angle Nobody's Covering
Here's the contrarian angle that mainstream coverage is missing: Agentforce isn't just competing with Microsoft. It's competing with the entire business process outsourcing industry.
India and the Philippines host a combined BPO sector worth over $200 billion, much of it focused on customer service. Agentforce's "end-to-end customer service request handling" is a direct threat to entry-level customer service jobs in those markets. The $2-per-conversation price point isn't just competitive with human agents—it's an order of magnitude cheaper.
FOMO drove the bus; reality hit the brakes.
The reality is that AI agents still struggle with complex, multi-step customer issues that require judgment, empathy, and creative problem-solving. But for the 70-80% of customer service requests that are routine—password resets, order status checks, billing inquiries—Agentforce is already good enough. And it doesn't sleep, doesn't take breaks, and doesn't need health insurance.
Salesforce is walking a delicate line here. It's selling automation that could displace millions of jobs while simultaneously positioning itself as a responsible AI company. The tension is real, and it's not being discussed in earnings calls.
The Valuation Question: How Long Will Patience Last?
Salesforce's stock trades at roughly 50-60x earnings, a premium to traditional SaaS peers in the 30-40x range. That premium is an AI narrative bet. The market is paying for the possibility that Agentforce transforms Salesforce from a subscription software company into an AI platform that captures value based on outcomes.
But that premium requires evidence. Analysts expect Agentforce to contribute $500 million to $1 billion in ARR in fiscal 2025. If the next few quarters don't show meaningful progress toward that number, the multiple compression will be brutal.
The capital expenditure picture adds pressure. Scaling AI agents requires significant compute infrastructure. Salesforce's capex is expected to grow 30-50% this year, which will pressure free cash flow and margins. The unit economics work on paper—at $2 per conversation with inference costs estimated between $0.05 and $0.30, gross margins look healthy. But those estimates assume efficient model deployment, caching, and token optimization. Reality tends to be messier.
The Security Blind Spot
One dimension is conspicuously absent from the Agentforce conversation: security and accountability.
AI agents that autonomously execute customer service tasks, handle refunds, or make commitments on behalf of enterprises create new risk surfaces. What happens when an agent makes a wrong decision? Who's liable—Salesforce, the customer enterprise, or the underlying model provider?
The prompt injection attack vector is real. Malicious actors could potentially manipulate AI agents into unauthorized actions through carefully crafted inputs. Salesforce's Trust Cloud provides a compliance framework, but the autonomy of AI agents adds complexity that traditional software never had to address.
Financial services and healthcare customers—the ones with the deepest pockets and the strictest compliance requirements—will demand answers before scaling Agentforce deployments. SOC 2, ISO 27001, HIPAA, FedRAMP certifications matter. So do red team testing and adversarial attack resilience.
Gravity always wins, even in a vertical chain.
The Verdict: Watch the Metrics, Not the Narrative
Agentforce is the most significant enterprise AI commercialization experiment of 2024-2025. Its success or failure will shape not just Salesforce's valuation but the entire enterprise software industry's approach to AI monetization.
The signals to watch are specific:
Q3 earnings should show Agentforce ARR contribution, customer counts, and conversation volumes. Without those numbers, the "early success" narrative is just narrative. Customer success stories with quantified ROI will separate real traction from pilot fatigue. Microsoft Copilot's enterprise adoption data will reveal whether Salesforce's head start is sustainable or illusory.
The $2-per-conversation pricing model is the most important innovation here. If it works, expect every enterprise SaaS company to adopt usage-based AI pricing. If it fails, the industry will stick with per-seat subscriptions and AI features will remain bolt-on enhancements rather than core value drivers.
Salesforce is betting that enterprise AI agents are the next major computing platform shift. The data, the customer base, and the workflow integration all support that thesis. But the market's patience has limits, and the next two quarters will determine whether Agentforce is a genuine transformation or just another earnings call buzzword.
The question isn't whether AI agents will transform enterprise software. That's inevitable. The question is whether Salesforce can convert its data advantage and customer relationships into durable AI revenue before Microsoft's ecosystem and AI-native startups close the gap.
Watch the conversation counts. Watch the ARR contribution. Watch the margin trends.
The narrative is compelling. The evidence is still pending.
