Salesforce Just Made $800 Million Selling AI Agents. Here's What That Means for Your Buy-vs-Build Decision.

Salesforce reported $800 million in Agentforce ARR in Q4 FY26 — its fiscal year ending January 31, 2026. That's 29,000 deals closed in the first 15 months since launch, growing at 169% year-over-year, with deal volume up 50% quarter-over-quarter. Production accounts increased nearly 50% in a single quarter.

This is not a pilot program. This is a product category with real revenue.

The question isn't whether AI agents work. The numbers closed that debate. The question is where you get them — and whether buying off the shelf is the right call for your business.

What $800 Million in Agent Revenue Actually Tells You

The headline number understates the market. When Salesforce combines Agentforce with Data Cloud 360 and Informatica, the combined ARR exceeds $2.9 billion, up 200% year-over-year. Across all deployments, 3.2 trillion tokens have been processed — these agents aren't sitting on shelves.

The more revealing number is the customer mix: more than 60% of Q4 bookings came from existing customers expanding, not new logos. Enterprises aren't just buying agents once. They're going back for more.

The pattern is consistent with how enterprise technology diffuses. Enterprises pilot, then scale. Mid-market follows enterprise by 12 to 18 months. That math puts 2026 and 2027 squarely in the mid-market window.

Goldman Sachs, in its March 2026 survey of 10,000 small businesses, found that 76% say they use AI, but only 14% have fully integrated it into core operations. Seventy-three percent say they need more training and resources. The gap between "using AI" and "integrated AI" is enormous — and that gap is exactly where the buy-vs-build decision lives.

The Buy Case: Platform Agents

The platforms exist, they work, and they deploy fast. Salesforce Agentforce, Microsoft Copilot Studio, and ServiceNow are the dominant players. Microsoft's Wave 1 update in April 2026 brought custom MCP servers to general availability, with computer-use agents following in May. The ecosystem is maturing quickly.

The advantages of buying are real. Platform agents deploy in days rather than weeks. They come pre-built for standard workflows: support ticket resolution, sales qualification, data routing, internal knowledge bases, HR intake. Salesforce deployments have reported ticket volume reductions of up to 45% for customer support operations. You get predictable subscription costs that scale with headcount rather than unpredictable engineering costs.

The risks are equally real. Platform agents create vendor lock-in. Salesforce customization lives inside Salesforce's constraints. Your competitive operations end up looking structurally identical to your competitors' competitive operations — because they're running the same platform on the same infrastructure with the same default behaviors.

Platform agents are the right call when the workflow is industry-standard: support resolution, appointment scheduling, document intake, FAQ deflection. These are cost centers, not differentiators. Optimize them with whatever deploys fastest and runs most reliably.

The Build Case: Custom Agents

Building custom agents means working with AI SDKs, open-source frameworks like LangChain or CrewAI, or hiring a builder who specializes in agent infrastructure. The upfront cost is higher. The maintenance burden is real. But the economics and the strategic position look different over a 24-month horizon.

Unit economics favor custom builds at scale. Platform agents carry subscription pricing and per-seat markups. Custom deployments typically run $0.25 to $0.50 per interaction once infrastructure is in place — versus $3 to $6 for human-led equivalents, and often significantly more expensive than that per-interaction on enterprise platform tiers. The crossover point depends on volume, but mid-market businesses running thousands of agent interactions per month hit it faster than most expect.

The reskilling cost is worth understanding before dismissing custom builds: $15,000 to $30,000 per person to build internal AI capability. That's Salesforce's own internal benchmark. It's real money, but it's a one-time investment in capability that compounds — unlike a subscription that resets every renewal.

The strategic case for building is straightforward. Platform agents make you operationally efficient. Custom agents make you competitively distinct. Medvi built $400 million-plus in revenue on custom AI infrastructure, not platform agents. The leverage came from building around proprietary workflows that competitors couldn't replicate by buying the same software.

Custom agents are the right call when the workflow is your edge: a proprietary pricing model, a customer-facing product, a data-driven process that generates revenue in a way competitors can't match. If the workflow is a differentiator, commoditizing it with a vendor platform is a strategic error.

The Framework: Buy Commodity, Build Differentiation

Most businesses will do both. This is not a binary decision and framing it that way leads to the wrong outcome. The practical framework is simpler: buy platform agents for the 80% of workflows that are industry-standard; build custom agents for the 20% that generates revenue or creates competitive advantage.

Gartner projects that 40% of enterprise applications will embed task-specific agents by the end of 2026. The JPMorgan Chase Institute, in its April 2026 analysis of actual SMB payment data — not surveys — found that businesses are already shifting from sporadic AI tool purchases to consistent, diversified AI spending. They're splitting the portfolio in practice, even before they articulate the framework explicitly.

If your AI spending is concentrated in a single platform vendor, that's a signal worth examining. Either you're underinvesting in custom capability, or you're over-relying on a platform that can't differentiate you. Most businesses doing this well are spending on both — standard platform agents for operational efficiency and custom infrastructure for competitive advantage.

The benchmark question is: which of your workflows, if an agent ran it better than any competitor's, would actually change your revenue trajectory? Build there. Buy everywhere else.

The Window

29,000 companies have already made their AI agent decision. The Goldman Sachs data says only 14% of SMBs have fully integrated AI into core operations — which means most of your competitors are still deciding. That gap is the window.

$800 million in Agentforce ARR is a market signal, not a buying recommendation. Salesforce built a $800 million business selling agents to enterprises. The mid-market version of that transaction is happening now, across multiple vendors and deployment models.

The concrete next step is an audit: catalog your workflows, flag the standard ones for platform agents, and flag the differentiated ones for custom builds. The former category will clear your plate. The latter will build your moat.

If you need help deciding which workflows to buy and which to build — and how to prioritize the sequencing — that's exactly what our AI audit covers in the first seven days. Start with the audit.