Persona Forge is a guided builder for an AI Persona delegation package. It walks you through the Three Questions of delegation in the methodology's own wording and has you score the persona's risk to set its tier. While you type, it writes the 19-section persona definition with its SOP, measurement blueprint and delegation checklist, then checks the sections against each other.
What it does
A readiness tab shows how many required sections each tier still lacks and what stands between the draft and a signature. The consistency checker catches sections that disagree, such as a $200K decision bound on a Tier 1 persona or a measure with no type, and it quotes the rule each warning applies. You can model a handoff to a second persona and get the register entry with the same clause in both packages.
For release testing, prompts for a frontier, a mid and a small model hand a generic agent only the SOP and two cases. You run them on your own models and record each run, and the page writes the vanilla-agent log. The whole set downloads as one zip, and the package prints with a sign-off page. Bring an edited package back in and the page shows what changed.
How to use it
Open the tool, pick one of the invented examples, and change anything. The examples run from Tier 1 to Tier 3, and supplier onboarding reaches Tier 3 through a cross-domain handoff. You can also start blank and work through the nine steps.
Drafts stay in your browser, and Export JSON moves one to another machine.
What it does not do
It never calls a model. The examples' numbers are illustrations, and their sign-offs read pending on purpose, since filling them would claim work nobody did. A package from elsewhere comes back with about 30 fields read and the rest kept as section text. The second persona gets a stub package only.
The four authority wording styles are under test in a registered lab study whose results aren't public yet, so the builder offers them and recommends none.
Where the numbers come from
Section names, tier rules, escalation tiers and measure types follow the AI-First Business Service Architecture templates and chapter notes. The numbers that come from that methodology are the $50K per-decision qualifier from the tier guide, the 4-hour, 1-hour and 15-minute escalation SLAs from Chapter 7, the 8, 14 and 19 required sections per tier, and Chapter 9's 12 test scenarios with a staged rollout at 10%, 50% and 100% for Tier 3. Every other number on the page belongs to the invented examples. The lab's research on what delegation costs and what review returns is the background to the questions it asks.
Open the tool
The page opens on a blank draft or your saved one. A self-check of 144 tests runs on every load and shows in the footer. It runs in your browser and sends nothing anywhere.
Open the tool Full screen in a new tab Open on the supplier onboarding example