The not-ready call
Everything in this chapter converges on one moment: the go/no-go. Launch review, steering committee, the client's sponsor on the call, and you with the verdict. It is the most senior thing you do, and the AI era makes it harder in a specific way — the demo always works now. An AI builder produces something demoable long before it produces something done. The happy path gleams. The room, which has been burned by slow projects for twenty years, sees a working demo weeks early and wants to ship. The pressure to say "go" arrives earlier and better-dressed than it ever has.
"Not ready" delivered as a feeling loses that argument every time. You're one voice of caution against a gleaming demo and a calendar. "Not ready" delivered as evidence isn't an argument at all — it's a reading. Two launch-blocking checks are failing: C2, source verification on the compliance module, second rework attempt; C5, the load test, never run. Here's the record. Nobody negotiates with an acceptance record. They negotiate with your mood, your track record, your seniority — a record just sits there being true.
Ready is a computation
So the readiness gate is the smallest function in the chapter, and the one you'll run under the most pressure:
GO if every launch-blocking task in every workstream is green. Otherwise NO-GO, with the blockers listed by name.
Details that carry weight:
launch_blockingis assigned at spec time, like everything else load-bearing in this chapter. Deciding which tasks block launch while the room wants to ship is deciding under maximum pressure with maximum motivated reasoning. Decide early, in writing, when nobody's emotional about a date.- A non-blocking red doesn't stop a launch. The gate isn't perfectionism — a red on the nice-to-have reporting widget rides along in the notes. Ship, log it, fix it next sprint. A gate that blocks on everything gets routed around by Thursday, and then you have no gate.
- A NO-GO is a to-do list, not a verdict on anyone. The output names the exact tasks standing between the team and launch. That reframing is what makes the call survivable to deliver: you're not the person who killed the launch, you're the person holding the shortest possible list of what makes it launchable.
The Deloitte report from lesson two is what shipping through a failing blocking check looks like: the deliverable was demoable, the deadline was real, the one check that mattered had no pass on record, and the invoice for skipping it was A$97,587.11 plus the firm's name in every story about AI failure for a year. The no-go call costs a hard meeting. The go call, made wrong, compounds.
Two drills left: the gate, then the whole pipeline — records to roll-up to verdict. Finish them and you'll have the entire chapter in one runnable file: specs in, evidence through, and a status line you can defend to anyone.