The method
Each layer stacks on the one below. Foundation holds the cloud and identity. Delivery ships code. Orchestration brings every vendor into the same mesh. Intelligence watches and summarises.
Layer 1
The ground the company stands on. Cloud, identity, secrets, source control, DNS.
Layer 2
How code ships. CI/CD, release automation, mobile pipelines, failure notifications.
Layer 3
Every external tool brought into code. Vendor APIs, work management, calendars, payments.
Layer 4
Observability, analytics, AI insights, scheduled digests, cost visibility.
Where AI lives
Build time
AI reads the spec and emits Terraform, GitHub Actions, runbooks, KPI queries, vendor integration modules. The output is static code committed to the customer's repo.
Refactor time
When the spec changes (new vendor, new product line, new compliance requirement), AI recompiles affected modules and opens a PR. A human reviews and merges.
Upgrade time
When upstream tools change (Terraform provider bump, new Claude model, deprecated API), AI regenerates the glue. Same review gate.
Runtime, in fixed roles
Scheduled KPI digest, cron agent, insight job, code review pass, developer machine bootstrap, alert spike summariser. Fixed inputs and outputs, no improvisation on the hot path.
Runtime, agentic investigation
Triggered by the unpredictable (customer complaint, infra spike, unexplained metric). Agents read across database, logs, source, deploys, correlate, propose a solution, open a PR.
AI is the connective tissue. Code is the spine. Agents are the response to the unpredictable. The customer can unplug us tomorrow and the compiled output keeps running. The agents become useful precisely because the platform underneath them is cohesive.
Engage
OneOps leaves no proprietary runtime dependency. Every Terraform file, every Action, every prompt, every SQL view lives in your GitHub org. If we disappear, the compiled code keeps running.