FAQ
Frequently asked questions
Clear answers about wallet credit, usage, subscriptions, and how Tycoon charges for work.
Can an AI really handle on-call?
With scope limits, yes. The AI DevOps Engineer handles first-line triage for the first 10 minutes: read alerts, correlate with recent changes, form a hypothesis, propose or execute a mitigation within its autonomy boundary. That boundary typically includes: rolling back the most recent deploy, scaling up a bottlenecked service, restarting a crashed worker. It excludes: touching the database directly, changing customer-visible behavior, triggering financial actions. For those, it pages a human with full context (logs, trace, timeline, proposed fix). Most founders report the AI handles 50-70% of night-time alerts without waking them.
Which cloud providers does it know?
First class: AWS, GCP, Cloudflare, Fly.io, Vercel, Netlify, Railway. Orchestration: Kubernetes (EKS, GKE), ECS, Cloud Run, App Runner. IaC: Terraform, Pulumi, AWS CDK, SST. CI/CD: GitHub Actions, CircleCI, Buildkite, GitLab CI. Observability: Datadog, Grafana Cloud, New Relic, Honeycomb, Sentry. If you run on Azure or less common platforms, the AI can work there but iteration is somewhat slower.
How does it avoid deploying broken code?
Layered defense. CI blocks on: tests, type check, lint, security scan. Deploy blocks on: container build failure, migration dry-run failure, health check failure. Canary blocks on: p99 latency regression > 10%, error rate regression > 0.5%, business metric regression > 5% (if configured). Auto-rollback if any health gate fails for 3 consecutive minutes. The AI DevOps Engineer treats the pipeline as the product and tightens the gates based on what actually catches issues. In practice this catches 90%+ of bad deploys before they affect customers.
What about security and secrets?
Secrets management goes through your existing store (GCP Secret Manager, AWS Secrets Manager, Doppler, Vault). The AI DevOps Engineer coordinates with the AI Security Engineer on rotation cadence (typically 90 days for API keys, 30 days for database credentials). Secrets never appear in Terraform state files, container images, or CI logs. Patch cadence: weekly minor patches via Renovate, monthly planned major upgrades, immediate out-of-band patch for any CVSS 7+ CVE with an active exploit. This is the hygiene a careful human engineer would run, executed weekly instead of quarterly.
When should I hire a human DevOps engineer?
Three cases. First: regulated environments (HIPAA, SOC 2 Type 2, PCI DSS) where a named human on the audit trail matters. Second: multi-region active-active architectures with complex consistency requirements. Third: organizations with 20+ engineers where the DevOps role becomes a platform-team lead, not just infra. Below that bar, most founders running Tycoon report the AI DevOps Engineer runs their infra more reliably than the freelancer or junior they previously hired, with better documentation and faster incident response.