No. Plenty of work is better handled by a clean workflow, an integration, or a scheduled job. AI belongs where interpretation, drafting, classification, or judgment actually adds value.
No. I work with technical teams on AI engineering controls, DevOps, AWS, and infrastructure, and with small businesses on systems, automation, data, AI training, and practical AI integration.
A concrete weak point: unreliable agent changes, disconnected systems, duplicate entry, weak CI gates, or a process that depends on somebody remembering what to do.
As far outside the model as practical. Branch protections in source control, test requirements in CI, infrastructure rules in policy and code, business rules in workflows and permissions, failures in monitoring.
Yes. AWS infrastructure, Terraform, CI/CD, monitoring, security, data pipelines, automation, and platform engineering are still core parts of the work.
Yes. The work should live inside the repositories, systems, cloud accounts, and operating practices the team already uses instead of creating a separate AI process off to the side.
Both. A focused infrastructure, automation, AI-control, or business-systems problem may have a clear endpoint. Ongoing work makes sense when the systems and workflows continue to change.
When agents are producing meaningful code and the team is relying on review, prompts, documentation, or convention to keep changes safe. At that point the checking should be done by the system, not by the agent.
AI is good at generating work and helping people move faster. It should not be responsible for remembering every rule that keeps the business working. If a decision can be enforced mechanically, move it into the system.
Talk to Brian