DevOps Engineer in 2026: AI Writes the Terraform You Review

Kai's Monday starts with a Datadog Bits summary: weekend latency spike traced to a stuck Postgres query on the payments DB. The AI has already gathered the trace, identified the culprit query, and opened a draft PR with an index migration. Kai reviews the plan, runs the migration in staging, checks the new execution plan, and merges. 40 minutes, start to finish, for something that would have taken half a day in 2020 — and might not have been caught at all.

• Incident triage — Datadog Bits, PagerDuty AIOps, Vercel Agent summarize and root-cause.
• IaC drafting — AI writes Terraform, Pulumi, Kubernetes manifests.
• CI/CD — GitLab Duo and GitHub Copilot for pipelines generate and debug YAML.
• Cost optimization — FinOps AI surfaces idle resources and right-sizing opportunities.
• Observability — auto-instrumentation with OpenTelemetry + AI-generated dashboards.
• Chaos engineering — AI-driven fault injection tied to service graph.
• Policy as code — OPA and Terraform policies drafted from natural language.

• Vercel Agent — autonomous code review and incident investigation.
• Datadog Bits AI — observability with LLM-powered queries.
• PagerDuty AIOps — incident correlation and runbook automation.
• GitLab Duo / GitHub Copilot Workspace — AI-assisted pipelines.
• Harness AI — deployment and cloud cost optimization.
• Grafana / Honeycomb / Lightstep — observability platforms with AI assist.
• Crossplane / Terraform / Pulumi — infrastructure as code with AI.

Architectural decisions. Choosing between ECS, EKS, and serverless for a specific workload. Running a real incident where the AI does not know about the three-year-old compatibility issue between two services. Approving a Terraform apply that will drop a production database if the plan is wrong. Negotiating with security about what agents can and cannot do. Designing a rollout that degrades gracefully. And the most human skill: writing a postmortem that teaches, not blames.

• Linux — bash, systemd, networking, filesystems.
• Cloud — AWS or GCP deeply; Azure optional.
• Kubernetes — even if you do not use it daily, you will eventually.
• Terraform or Pulumi — IaC is table stakes.
• Observability — metrics, logs, traces, SLOs.
• Incident response and postmortem writing.

If you want to be a DevOps/platform engineer: In high school, install Linux on an old laptop. Learn bash and git. Break things. In college, CS or self-taught + certs work equally well; AWS Certified Solutions Architect and CKA (Certified Kubernetes Administrator) are the most-hired certs in 2026. Contribute to open-source tools like Terraform providers or kubectl plugins. Platform engineering is one of the fastest-growing specialties in 2026 because AI has made developers more productive — which means platforms need to scale faster, and platform engineers are the ones keeping up.