Gearset has launched Cam, an AI agent designed to take Salesforce development tickets from request to pull request within a team's existing governed release process. The announcement comes as research shows that while AI coding assistants speed up individual developers, review bottlenecks and rework often prevent teams from increasing overall throughput.
In early customer pilots, Cam created up to 44% of a team's pull requests within a week of onboarding. The agent interprets change requests, asks clarifying questions, and builds Salesforce changes based on the context of a team's repository and specific org setup. Before submitting work for review, Cam validates that changes are deployable and comply with the organization's standards, documented conventions, and security controls.
Moving past the AI bottleneck
Kevin Boyle, CEO at Gearset, described a common pattern in development teams. "AI can write a change in seconds, but making sure it's right and getting it into a live business system without breaking anything is the hard part," he said. "Many development teams have quickly reached a level of AI maturity where every contributor generates code quickly with AI assistants, but the team as a whole hasn't increased throughput because the bottleneck moves to code review."
Research from Faros AI across tens of thousands of developers in 2025 and 2026 supports this finding. Individual coders produce work faster with AI, but review and rework consume the time saved before the team sees net gains. The challenge intensifies when organizations consider handing more autonomy to AI agents, where governance and security become non-negotiable requirements.
For teams working in Salesforce environments, the stakes are particularly high. Granting broad access to production orgs creates risk. Cam operates within the same testing, review, and delivery controls applied to every team member, producing a complete audit trail of what it built and why. One pilot customer saw 75% of Cam's pull requests successfully merged.
Governance as a feature, not an afterthought
Marilyn John, Sr. Revenue Systems Manager at Hudl, framed the tension many Salesforce teams face. "The rest of our company moves fast with general AI tools. We needed a way to move that fast inside Salesforce, without handing everyone read and write access to our production org," she said. "Cam gave us the speed with the permissions and the governance intact."
The launch expands Gearset's role from a platform that helps teams manage and release Salesforce changes to one that also creates them. The company's existing DevOps experience underpins the guardrails baked into Cam's operation. "Salesforce teams need AI they can trust - not just capable, but governed and secure by their own standards," Boyle said. "Cam is the only agent that delivers both."
Why this matters for IT and development teams
The gap between individual AI productivity and team-level throughput is a live problem for development organizations. Tools that generate code faster do not solve the downstream work of validation, compliance checking, and integration. Cam's design addresses that gap specifically for Salesforce teams by embedding governance into the agent's workflow rather than treating it as a separate step. For IT leaders evaluating AI for IT & Development, the pilot data offers a concrete benchmark: 44% of pull requests generated by an agent, with the majority surviving review and merge. The lesson extends beyond Salesforce - governed agentic workflows may determine whether AI delivers team-wide acceleration or simply shifts the bottleneck further down the pipeline. Developers looking to understand how agents fit into their workflow can explore the AI for Software Developers learning path.
Your membership also unlocks:
