Get started · Tool 2 of 3
The MVV Tracker
Twelve artifacts, four weeks, one team. Each artifact has a definition of done — check the parts as you build, and the tracker keeps score in your browser. Documents and resolve; no software required.
0 of 12 artifacts done
Progress is saved only in this browser. Start with your vital signs if you haven't — the weakest verbs tell you which week deserves extra care. Working offline or briefing your team? Download the MVV Field Guide (7-page PDF).
Week 1 — Purpose & promise
Why the team exists — the document every agent and newcomer reads first.
- Mission in one or two sentences
- Intent clause: the outcome + non-negotiables, written for the unanticipated case
- Scope (what we own) and boundaries (what we don't) listed explicitly
Framework: Component 1 · Kit file: CHARTER.md
Three services, each a published promise with governance attached at birth.
- Three services named, each deliverable in a sentence
- Each carries: default risk tier, automation ceiling, evidence level, review gate
- The team agrees it can actually deliver all three (cognitive-load check)
Framework: Component 3
One declared way in. Ambient demand is where work disappears.
- One channel declared the front door; others route into it
- Form captures: what, for whom, by when, category
- The hallway ask has a rule ("say yes to capturing it, not to doing it")
Framework: Component 4
Week 2 — The work
The governance checkpoint of first contact — the profile travels with the request for life.
- Classification questions written (type, urgency, risk, owner, evidence, approval)
- Governance profile fields defined (tier · PHI status · external-facing · human review)
- P0–P4 + Reject adopted, with the protection override in writing
Framework: Components 5–6 · Kit: workflow header
One real workflow made explicit — trigger to follow-up, leveled per step.
- Steps written from trigger to documentation
- Each step declares: who, automation level (L0–L4), evidence produced
- Review and approval are separate steps with different names on them
Framework: Component 7 · Kit: workflows/_TEMPLATE.md
Every chain of work terminates in a named human.
- The eight role types assigned to actual names
- AI agents listed as Operator/Contributor only, each with a human principal
- Escalation owner named and reachable
Framework: Component 8
Week 3 — Memory & control
What the team runs on, tiered by authority, with staleness owned by someone.
- Sources listed and tiered: governed / working / ambient
- Each governed source has a named steward and a review date
- The three tribal-knowledge items from your vital signs have a capture plan
Framework: Component 10
Not a human glancing at AI output — a designed control that stays alive.
- Gates placed by tier and reversibility (and ungated-on-purpose written down)
- Each gate names its active task (what the reviewer does, not just reads)
- Rejection counting started — a gate that never rejects gets re-examined
Framework: Component 14 · the six gate rules
If it isn't in the ledger, it didn't happen.
- Append-only log created with the entry format
- First real entries written (including one gate event)
- Rule adopted: no patient/confidential data in the ledger, ever
Framework: Components 15–16 · Kit file: LEDGER.md
Week 4 — The heartbeat
Work visible across levels, carrying its governance profile.
- Active work listed with owner, status, and due date
- Tier/gate status visible ("what high-risk work is currently ungated?" is answerable)
- Stale items surface automatically or by weekly review
Framework: Component 13
A brief you can read in the elevator; a meeting you could stand up through.
- Template set: Situation / Background / Assessment / Recommendation
- First brief produced and read
- One standing "challenge round": disagreements with AI output logged and praised
Framework: Component 12 · Kit: workflows/example-weekly-brief.md
The baseline — and the loop that makes everything else compound.
- Baseline snapshot taken (volume, speed, quality — whatever you honestly have)
- Governance health included: gate events, escalations, boundary blocks
- First monthly improvement loop scheduled before the month ends
Framework: Components 16, 18 · Kit: prompts/improvement-loop.md
When all twelve are done: run your first canvas rubric scoring to see the maturity picture, and put the improvement loop on the calendar — the loop is the part that pays.