The human gate

Maps one business process step by step and marks every step green (AI may draft), amber (AI may check, a person decides) or red (a person only), with the reason, then turns the map into a one-page AI use policy for that process. Use when an owner, manager or compliance lead wants to use AI safely in a process with legal, customer or chain-of-custody stakes, or needs an AI policy their auditor would accept.

Run it as a prompt

Paste this into any AI agent, or fetch it: curl -s https://uplift.page/api/v1/prompts/human-gate-map/raw

prompt.md

# The human gate

Show me exactly where AI may help in one of my processes and where a person must stay in charge, then write the policy that holds that line.

## Before you start

Ask me to describe one process from start to finish (how it starts, each step, what it produces, who signs off), then stop and wait. Rough bullets are fine. If I say **"use the sample"**, use the made-up process at the end of this prompt.

## Steps

1. **Number the steps,** splitting any that mix work and decisions.
2. **Color every step,** in a table with these columns: step, color, what AI may do, what a person must do, reason.
- **Green, AI drafts:** low risk and easy to check (summaries, first drafts, formatting).
- **Amber, AI checks:** AI compares, flags or suggests, and a person decides (mismatch checks, gap lists).
- **Red, a person only:** anything that signs, certifies, pays, commits the company, changes a record others rely on, or goes to a customer unreviewed.
3. **The gate list:** every red step, with who owns it and what they check.
4. **One-page AI use policy** for this process:
- purpose;
- what AI may do (green and amber);
- what it may never do (red);
- what data never goes into an AI tool (and the redaction rule);
- who reviews AI output;
- how mistakes are reported;
- review date.
Plain words, no legal boilerplate.
5. **One question for leadership:** the single decision this map can't make for them.

## Rules

- When unsure between two colors, choose the safer one and say why.
- Never put a signature, certification, payment or customer-facing send in green.
- This is a working policy draft. A manager adopts it, not the AI.

## Sample (made up, for practice)

Pickup to certificate: customer emails a pickup request and asset list → we quote → customer accepts → we schedule a truck → driver collects and signs the manifest → receiving counts and logs serials → data-bearing devices are wiped or shredded → QA verifies → we issue a certificate of destruction signed by the QA lead → account manager emails the certificate and settlement → finance invoices.

Install it as a skill

Agents that support the Agent Skills standard load it automatically when it applies.

download

curl -fsSL https://uplift.page/p/human-gate-map/SKILL.md --create-dirs -o .agents/skills/human-gate-map/SKILL.md

or from any MCP client

MCP server: https://uplift.page/mcp
Tool: pull_skill  {"slug": "human-gate-map"}

The full skill

Show me exactly where AI may help in one of my processes and where a person must stay in charge, then write the policy that holds that line.

Before you start

Ask me to describe one process from start to finish (how it starts, each step, what it produces, who signs off), then stop and wait. Rough bullets are fine. If I say "use the sample", use the made-up process at the end of this prompt.

Steps

Number the steps, splitting any that mix work and decisions.
Color every step, in a table with these columns: step, color, what AI may do, what a person must do, reason.
- Green, AI drafts: low risk and easy to check (summaries, first drafts, formatting).
- Amber, AI checks: AI compares, flags or suggests, and a person decides (mismatch checks, gap lists).
- Red, a person only: anything that signs, certifies, pays, commits the company, changes a record others rely on, or goes to a customer unreviewed.
The gate list: every red step, with who owns it and what they check.
One-page AI use policy for this process:
- purpose;
- what AI may do (green and amber);
- what it may never do (red);
- what data never goes into an AI tool (and the redaction rule);
- who reviews AI output;
- how mistakes are reported;
- review date. Plain words, no legal boilerplate.
One question for leadership: the single decision this map can't make for them.

Rules

When unsure between two colors, choose the safer one and say why.
Never put a signature, certification, payment or customer-facing send in green.
This is a working policy draft. A manager adopts it, not the AI.

Sample (made up, for practice)

Served from the uplift.page library and refreshed within 5 minutes of every update.