The operations layer after the model call
Agentic AI gets useful when teams can see it, govern it, approve it, recover it, and connect it to engineering outcomes. The model call is the easy part — the operations layer around it is where AI-assisted delivery becomes safe enough for real teams. Jon Price builds that layer: the telemetry, cost accounting, approval gates, governance, and recovery paths that let engineering organizations adopt coding agents without giving up observability or control.
- Coding-agent observability and run telemetry — every autonomous run emits a trace of which agent acted, what it changed, and how long each step took, so teams can debug a bad change and trust the next one.
- Token, model, and cost visibility by repo or workflow — attributing spend to a repository, workflow, or team turns agentic AI from an unpredictable bill into a budget you can reason about and defend.
- Human approval gates for autonomous delivery — agents propose, and humans approve the changes that touch production, security boundaries, or customer data, so autonomy and accountability meet without stalling delivery.
- Repository intelligence and engineering governance — dashboards for ownership, test coverage, and dependency risk give agents and reviewers the same context, so automated work lands where it is actually safe.
- Failure handling, ownership, and rollback paths — when an agent gets it wrong, the system routes the failure to an owner, notifies the right channel, and offers a clean rollback, because recovery is what makes autonomy survivable in production.
An operating model, not a single tool
These pillars work as a system. Observability feeds cost visibility; cost and governance data inform which changes need a human gate; approval outcomes and failures feed back into repository intelligence. The result is a closed loop where agents can do more over time precisely because the organization can see and trust what they do — the same operating model behind the systems on the projects page and the tooling documented on the uses page.
Positioning
Jon builds agentic AI systems for engineering operations: the dashboards, approval workflows, telemetry, and platform primitives that make AI-assisted delivery safe enough for real teams. If you are hiring for engineering leadership in this space, the current-focus page has the short version.