Disclosure up front: I work at ServiceNow. I have friends I love at Salesforce — people I've worked alongside, competed against, and respected for years. This isn't a dig. What's happening right now in enterprise tech is happening to everyone, and nobody has this figured out. I'm writing this because the speed of evolution is genuinely shocking, and I think the analysis is worth having honestly rather than politely.
If you'd rather skip the treatise on the epistemological crisis of SaaS disintermediation and just get the seven-slide version — click below. Arrow keys. Two minutes. Done.
For everyone else — let's get into it.
Marc Benioff is one of the best storytellers in tech. "Headless 360, no browser required" sounds bold. But look at the pressure map and tell me this was proactive.
Microsoft has Copilot embedded everywhere Dynamics touches. The interface layer — already taken. HubSpot keeps moving up-market, one tier at a time, cheaper and cleaner. ServiceNow is making a serious agentic-platform play — not a feature bolt-on, a category claim. And then there are the MCP-native startups building CRM-shaped functionality without twenty-five years of Visualforce and Apex debt.
This isn't foresight. It's necessity, dressed as vision.
Here's the thing nobody at Dreamforce is going to say out loud: seat-based pricing is on borrowed time.
The math is simple. If your enterprise customer has 500 sellers on seats today, and agents compress that to 200 humans plus agents doing the work, the revenue line compresses with it. The entire SaaS business model — the thing that built Salesforce into a $300 billion company — starts breaking.
Watch for MCP-call metering, agent-action pricing, or an Agentforce credits scheme within eighteen months. This isn't just an architecture play. It's groundwork for a new revenue model that Salesforce needs but can't announce yet.
The real bet isn't the headless UI. The real bet is becoming the control plane for agents touching enterprise data.
That's a real need. Forty agents from fifteen vendors, all reaching into your system of record — someone has to handle observability, guardrails, orchestration, audit trails. Salesforce has a clean claim to govern agents touching its own data. But governing the whole agentic enterprise — Workday, Snowflake, SAP, Slack — is a much harder sell.
And everyone is running for this position. ServiceNow with AI Control Tower for agent governance and Workflow Data Fabric connecting structured, unstructured, and streaming data so agents can act on real-time context without moving or duplicating data. Microsoft with Copilot Studio and governance via the tenant. Databricks with orchestration rooted in the lakehouse. AI-native startups building pure orchestration with no legacy to defend. Salesforce's Agent Fabric is defensible at home and contested everywhere else.
I work at ServiceNow, so take this with the appropriate grain of salt — but I know what a serious platform play looks like when an incumbent decides the old moat isn't deep enough. ServiceNow's bet is that the control plane isn't about owning any single system of record — it's about being the workflow layer that orchestrates agents across every system of record. AI Control Tower governs the fleet. Workflow Data Fabric gives agents the cross-enterprise context to act with judgment, not just instructions. It's a different architectural thesis than what Salesforce is building with Agent Fabric, and the market will decide which approach wins.
And honestly, ServiceNow is navigating the same existential questions. What does "platform" mean when agents start doing the work that humans used to do inside it? Nobody gets to sit this one out.
What gives me the most pause is the ecosystem cost.
Salesforce built one of the most underrated moats in enterprise software: over 100,000 Trailhead-certified admins, a massive SI partner network, and an entire "clicks not code" culture that made big implementations sticky. A headless future quietly erodes all of it.
"Your skill is obsolete" doesn't land quietly. Not to morale, not to renewal motion, not to the partner relationships that make big deals close. The partner ecosystem is one of Salesforce's most durable competitive advantages, and this pivot actively corrodes it.
That said — credit where it's due.
Most incumbents protect the UI. They bolt AI into the existing interface, keep the browser central, protect seat pricing for another cycle. That's the SAP path — what S/4HANA did for years, watching relevance drain out of the category.
Benioff is willing to cannibalize the UI entirely. That's uncomfortable for the org, for partners, for the entire ecosystem that was built on Lightning and Visualforce. But it might be the move that keeps Salesforce relevant in 2030. The willingness to self-disrupt is rare at this scale, and it deserves recognition even from people like me who think the execution odds are a coin flip.
My verdict: 50/50.
Salesforce isn't going anywhere. Data gravity and compliance are real — you don't walk away from a system of record that touches every revenue function in the enterprise. The question is whether they dominate the agent era or drift into being just another layer.
The bet is bold. The pressure that forced it was real. And whether it works depends on winning a control-plane race that at least four other serious players are also running.
I've seen this movie before. Sometimes the incumbent rewrites the category. Sometimes they just rewrite the marketing. But what's different this time is that every incumbent is watching the same movie at the same time. Salesforce, ServiceNow, Microsoft, SAP — we're all staring at the same question: what does your platform become when AI agents do the work that humans used to do inside it?
Nobody has the answer yet. The companies that figure it out first won't just survive — they'll define what enterprise software looks like for the next decade. The ones that don't will become infrastructure nobody thinks about.
To my friends at Salesforce: this isn't a dig. It's respect. Benioff is swinging at the right pitch. Whether he connects is the only question that matters.
Transmission posted: 8:14:22 PM