The Pilot Trap Is Real
Oppenheimer's Dreamforce survey found that 85% of Salesforce customers have tested or deployed Agentforce — and nearly half of them are still in pilot mode. Analyst Brian Schwartz flags production adoption and monetization as "uncertain" despite $1B+ in reported AI revenue and 29,000+ paid deals on paper.
That gap between paid and productive is not a product problem. It's a decision problem. Pilots don't stall because the technology stopped working. They stall because no one set terms for what "working" means.
Why Data Readiness Is the Hidden Blocker
The survey surfaces a specific structural issue: roughly 40% of data flowing through Agentforce is handled by third-party providers, not Salesforce Data Cloud. That means a significant share of deployments are running on data infrastructure that Salesforce doesn't control and hasn't optimized for agentic workflows.
For operators, this matters practically. If your CRM data is incomplete, siloed in a data warehouse, or enriched by a third party on a batch cadence, your Agentforce agent is reasoning on stale or partial inputs. The agent isn't underperforming — your data pipeline is. Fixing that is a 60-to-90-day project minimum, and it needs to be scoped now if you're targeting a Q4 production decision.
How to Make the Call Before Budget Season Closes
Use this as a forcing function. Before your next Q4 planning conversation, answer three questions about your current Agentforce pilot:
1. What was the success criteria when you started? If you can't name a specific metric — deflection rate, AE time saved, pipeline touched — the pilot was exploratory, not evaluative. That's fine, but label it honestly.
2. Where does your data actually live? Map the inputs your agent depends on against what's native in Salesforce Data Cloud versus pulled from external sources. If it's mostly external, add a data readiness workstream to the expansion plan.
3. What's the go/no-go date? Set one. Pilots that don't have a decision deadline become quarterly line items that survive budget cycles on inertia. Pick a date in the next six weeks, name the person who owns the call, and document the criteria.
The benchmark from Oppenheimer gives you external cover to ask harder internal questions. Half the market is in the same waiting room you're in. The operators who move to production first will set the internal benchmarks that everyone else chases next year.