Justin Smith is a founding product engineer at Resolve AI developing always-on background agents for production software. As AI accelerates coding and deployments, he focuses on the work that follows: monitoring releases, investigating anomalies, preserving operational knowledge, and reducing the burden on on-call engineers.
Smith has spent more than 15 years working on developer tools, monitoring, and observability, including roles at VMware and Splunk, where he was an architect of the Splunk Observability Suite. His interests in product design and frontend architecture shape his approach to operational AI: sophisticated infrastructure is useful only when engineers can understand it and incorporate it into their existing workflows.
- Production context determines useful automation. Running a query or opening a dashboard is easier than recognizing whether a metric looks abnormal for a particular service. Smith builds around evolving knowledge of dependencies, historical behavior, recent changes, and causal relationships. His writing on background agents argues that this environmental understanding is essential to effective production work.
- Operational work needs persistent ownership. His agents run in sandboxed cloud environments, respond to schedules, deployment events, and messages, and retain knowledge across tasks. They can investigate P99 latency drift, check service health and capacity, prepare on-call handoffs, or answer engineering questions without waiting for someone to initiate every task.
- Deployment checks should follow the changeset. Instead of applying identical monitoring to every release, an agent can inspect what changed, identify affected services, and assemble targeted checks for signals such as checkout latency, error rates, and Kafka pipelines. Smith treats this as an extension of existing CI/CD safeguards, with additional coverage for feature flags and infrastructure changes.
- Agents should work where engineers already work. Integrations with Slack and Microsoft Teams let agents respond when confident, ask privately for confirmation when uncertain, and adapt to conversational feedback. Adjustable autonomy and visible task histories keep people informed without requiring constant supervision.
At AI Engineer World’s Fair 2026, Smith connected these ideas to a practical engineering challenge: faster software creation increases the need for systems that can continuously understand, monitor, and support production.
