Map where humans stay in control.
This workspace helps you draw your team's AI safety line. You name your team, pick the human roles that matter, set three boundaries, and export a document you can share. It takes a few minutes.
When you finish, you will be able to hand your team a one-page protocol that says exactly what AI may do, what a human must decide, and how you check that the line holds.
Where this fits: the FrictionIndex™ (free) shows where your week goes. This tool decides which of that work a human keeps. The True-Net ROI Canvas™ then prices the result — find it in the Library.
Step 1
Set your context
Every map belongs to one team, one date, and one owner. That is what makes it real.
Step 2
Select role archetypes
Every team has people who must stay in the loop. Pick the roles that describe them. Choose at least two.
Commander
Final accountability
The person who holds final decision authority. AI can prepare, but never owns the outcome.
AI leverage: AI drafts options; human decides
Guardian
Ethical oversight
Watches for harm, bias, and compliance violations. The conscience of the system.
AI leverage: AI flags risks; human evaluates
Connector
Stakeholder alignment
Builds trust across teams and ensures AI decisions are communicated transparently.
AI leverage: AI summarizes; human negotiates
Oracle
Deep expertise
Domain expert who validates AI outputs against real-world constraints.
AI leverage: AI retrieves; human interprets
Architect
System design
Designs the structure that keeps human judgment embedded in automated flows.
AI leverage: AI prototypes; human validates
Creator
Original insight
Generates novel ideas that AI cannot produce from training data alone.
AI leverage: AI iterates; human directs
Amplifier
Scale with judgment
Multiplies human reach without sacrificing quality through intelligent delegation.
AI leverage: AI executes; human audits
Step 3
Define HOZ boundaries
For each card, write three things: what AI may do, what a person must own, and how you check it. These three cards cover the failures we see most.
Worked example — Critical Decision Boundary
This is what a finished card looks like. Yours should be this concrete.
AI Scope
AI models three pricing scenarios and shows the revenue and churn impact of each.
Human Ownership
The pricing committee decides any change worth more than $250,000 a year, or any change that touches more than 5% of revenue.
Audit Rule
Finance logs every pricing decision with the AI recommendation, the human reason, and the result. The CFO reviews the log each quarter.
Relationship Boundary
Relationships with clients, teams and partners. AI can prepare the work. A person owns the connection.
Not counted yet: Still to add: AI Scope, Human Ownership, Audit Rule.
Critical Decision Boundary
The point where AI stops and a person decides. Name a number, a dollar amount, or a risk level.
Not counted yet: Still to add: AI Scope, Human Ownership, Audit Rule.
Conceptual Origin Boundary
AI can combine what already exists. People invent what does not. Review this line as AI improves.
Not counted yet: Still to add: AI Scope, Human Ownership, Audit Rule.
Step 4
Pass the compliance gate
Answer three questions about what already happens on your team today. Short answers are fine. Your answers go into the export.
Why these three questions?
Each one covers a different way AI work goes wrong: data leaving safely (Redaction), big calls made too fast (Human Review Threshold), and answers no one can explain (Explainable Feedback — the AI shows its reasoning, not just its answer). Miss one and your protocol has a blind spot.
Redaction
Who removes sensitive data today, and where is that written down?
Sensitive data means PII (personally identifiable information — names, emails, account numbers) plus financial or confidential records. Name the person and the document.
Example answer: Our support lead strips names and account numbers before anything goes into the AI tool. The steps live in the Support Runbook, page 4.
Write a short answer, then tick the box to confirm it is accurate.
Human Review Threshold
What threshold sends a decision to a human today, and who set it?
A human review threshold is the point where AI stops and a person must look. It is a cooling-off rule, not a technical request limit.
Example answer: Any AI-suggested vendor contract over $50,000 waits 24 hours for a human sign-off. Our COO set that limit in March.
Write a short answer, then tick the box to confirm it is accurate.
Explainable Feedback
Where can someone see the reasoning behind an AI recommendation today?
Explainable feedback means the AI shows its reasoning, not just its answer, somewhere a person can go and read it.
Example answer: Every AI code suggestion posts a "Why" comment on the pull request with the alternatives it considered.
Write a short answer, then tick the box to confirm it is accurate.
Step 5
Export your protocol
Your map is built below as you type. Copy it, or download it, and paste it where your team already reads things.
Map Completeness
A boundary card counts only when all three fields are written and the human-ownership line names a threshold — a number, a dollar amount, or a risk level.
Next Steps
- Share this map with your team lead
- Schedule a quarterly review in your calendar
- Link it from your AI usage policy
- Revisit when AI capabilities change
79 lines · 202 words · 4 sections · Markdown
Reflection
"If a free, basic browser sandbox can map your organizational safety line in a few minutes, what excuse do you have for letting automation complacency expose your company's data, culture, or liability?"
Want to go deeper?
Explore the Library at theCommons Academy — guides, frameworks, and templates to help you refine your HOZ boundaries and build intentional systems.
Explore the Library