Are We Ready? Building an AI Launch Go/No-Go Workflow (for Project Managers)

Every launch planning cycle ends with the same question. Are we ready.

It sounds like one question. It is really eleven answers from eleven places, each owned by somebody different and each reporting on its own schedule. Collecting all of that is the easy half. The hard half is knowing what to look into — and that part stays yours.

This session builds the small version of a launch readiness workflow, step by step, in the tools you already have. It does not decide anything. It gathers the pieces, tells you how old each one is, points you at what to open first, and keeps a record of what it read.

Session one of Beyond Chatbots, a series for project managers.

What you get

• A saved prompt instead of one you retype before every launch
• Your thresholds written down while you are still planning, not in the last week
• A rule that turns any yellow without a plan, an owner and a date into a red
• Three places to start each morning, in order, with the reason beside each one and the date its source was last touched
• The arithmetic in ordinary code, the model only on what somebody actually wrote — which is where most of a token bill goes
• A clock, its own identity, a gate with a person's name on it, five lines of record

Sources (checked 1 October 2026)

• Microsoft, Dynamics 365 go‑live checklist — eleven categories, 49 line items
• Washington State OCIO policy 121.10 — readiness thresholds are identified during the planning phase, and the Executive Sponsor makes the go‑live decision
• Fedora Go/No‑Go — a rule written in advance, with "no room for discretion in this determination"
• Ofgem, Project Nexus go‑live decision report — Amber means a credible mitigation plan exists; ready means no Reds
• Keil, Smith, Pawlowski & Jin, "The Pitfalls of Project Status Reporting," MIT Sloan Management Review, Spring 2014 — optimistic bias is far more common than pessimistic
• CSET, "AI Safety and Automation Bias" — operators over‑trust automated output, and explanations can increase that trust rather than reduce it
• METR, early‑2025 AI and experienced open‑source developer productivity — they expected to be about a quarter faster, were measured 19% slower, and still believed they had been faster
• NIST SP 800‑37 Rev 2 — the authorizing official is a senior official who cannot delegate accountability for the risk acceptance decision
• Model Context Protocol — the open standard, now with the Linux Foundation
• Microsoft Copilot Studio — messages and credits. These limits move; check them the week you build.

Next time

We build one of these live, inside the tools you already have. We fix the prompt that works beautifully for you and badly for everyone else. A real MCP server goes behind it, and we work out who owns that door before anybody opens it. And we move the arithmetic out of the model and into code.

Anacay · anacay.com Narration is an AI clone of my voice. The script, research and design are mine.