CTO of One
[ ← CTO of One ]

Illustrative example, Fictional data, Not a client result

An agent scope sheet and triage drill for a billing service

Two artifacts a CTO can review before any agent touches production: what the triage agent may and may not do, and a drill showing where it hands over to a person. The company, services and incident are invented.

Build a question list for scope, evidence and written terms before starting live work. This preparation guide is not provider assurance.

Prepare your questions →

Before
Alerts page the CTO directly. The billing runbook lives in one person’s head, and agents use broad tokens from a shared account.
After
A scoped triage agent gathers context and drafts a first diagnosis. The on-call engineer decides, and only the escalation path reaches the CTO.

Inspect the sample

01 Input

The CTO of Quillhaven, a fictional SaaS company with 12 engineers, still takes every page for the billing service. The team has started opening agent-written pull requests, and nobody has agreed who approves them.

How it moves

  1. Scope the agent List the repos, logs and tools it may read, the commands it may run, and what it must never touch.
  2. Write the runbook Turn the CTO’s knowledge of billing failures into steps an engineer can follow at night.
  3. Run a drill Replay a past incident in staging and check that the agent stops and pages at the right point.
  4. Set the review rule Agent pull requests need tests and an approving engineer who understands the change.

02 Output

Triage agent scope sheet: billing service (fictional)
AreaAllowedNeverStop and page when
Code Read the billing and invoicing repositories Push to main or merge its own pull requests A fix touches payment-provider code
Logs and metrics Query the last 24 hours of billing logs and dashboards Export customer data outside the logging tool The error rate stays high after the runbook steps
Actions Restart the invoice worker in staging; open a draft pull request Change production configuration or secrets Any production change is needed
Data Use masked sample invoices for tests Read card numbers or bank details A test would need real customer records
People Page the on-call engineer with a summary Contact customers or the payment provider The on-call engineer has not acknowledged within 15 minutes

03 Review

Where a person decides

The on-call engineer approves every action beyond reading and drafting, and every agent pull request needs an engineer who understands the change. The CTO approves the scope sheet and any change to it.

Limits of this example

Fictional company, services and timings. A real scope sheet depends on your stack and your risks. A billing service touches payment records, which Build Plus lists; we confirm the level when we scope it, and rotation response terms are set in writing.

See whether this fits your work, and what it would take.

Is this a fit? → Ask about a Consult block