The Autonomy Gate
How it works Verdicts Watch Receipts Artifacts GitHub ↗ Try it
Autonomy verdicts for business workflows

Other tools route work to AI. This one decides whether AI is allowed to touch it.

The Autonomy Gate is a specialist judge. Describe any workflow and it issues a binding verdict — automate, supervise, stabilize, or refuse — then hands you the execution artifact built to walk into a meeting and be acted on.

Run a workflow through the Gate View on GitHub ↗
16 files · Claude Projects, ChatGPT, Claude Code, Codex · ~2 minutes to first verdict
Autonomy verdict
Human Only
GATE-2 · NOT_APPLICABLE
Autonomy verdict
Autonomous
CLEARED
RULE-03
Weekly KPI narrative · bounded, reversible, reviewed before it posts.
CONFIDENCE · HIGH BUILD_READY
The gap nobody owns

There is a gap between "AI can do this" and "AI is allowed to do this." Most tools live on the first side and assume the second.

Capability is cheap now. Authority is not. The question that stalls every automation project isn't whether the model is good enough — it's who is accountable when it acts, whether the action can be undone, and whether it should have been a machine's call at all.

The Gate answers that question and refuses, structurally, when the honest answer is no. Other tools route. This one blocks.

How it works

One input. One verdict. One artifact you can act on.

01
Describe the workflow

Plain language. The Gate normalizes it into an intake snapshot and identifies the terminal action — the last thing that actually executes.

02
It scores & gates it

Reversibility, observability, exception rate, cost of failure — then an adversarial check and five hard gates that cannot be overridden.

03
Verdict + artifact

A decision packet and the matching execution artifact — setup brief, control plan, governance memo — ending in a build handoff pack.

04
The operator signs

Every artifact ends in an Operator Disposition. The Gate may recommend — it may never pre-approve. A human signs for the build.

The four verdicts

Four answers — and the discipline to give the fourth.

Most systems only know how to say yes. The proof that this one judges rather than sells is that it will tell you no — by name, with the gate cited, and no path to override it.

Autonomous

AI executes without a human approval checkpoint. Bounded, observed, reversible, owned.

Supervised

AI prepares the work; a named human approves before anything executes. The checkpoint is the control.

SOP First

The process is too unstable for any autonomy level. Stabilize and produce evidence before reassessment.

The proof it refuses
Human Only

Judgment or risk that cannot be delegated — a structural block. When the terminal action is an irreversible external commitment or regulated personal data, the verdict is HUMAN_ONLY and the artifact is a governance memo, not a build.

GATE-2 · IRREVERSIBLE COMMITMENT GATE-3 · REGULATED DATA NO OVERRIDE

Watch the Gate issue a clean pass — and a structural refusal.

Two live workflows. The first earns AUTONOMOUS — a bounded KPI narrative that's reversible, observed, and owned. The second triggers GATE-2 and returns a governance memo, not a build. The Gate says no precisely — names the mechanism, preserves the human boundary, and still gives the operator a usable way forward.

Committed receipts

Three real submissions. Three honest verdicts.

Not demos — committed runs. A clean autonomous pass, a supervised pass with a blocking checkpoint, and a hard-gate refusal that cites GATE-2 by name. Open any one to read the full artifact.

Receipt 01 · clean pass
Weekly KPI report from CRM & analytics exports
VERDICTAUTONOMOUS
CONFIDENCEHIGH
HANDOFFBUILD_READY
Project Setup Brief Read →
Receipt 02 · checkpoint pass
Customer win-back email batch to lapsed accounts
VERDICTSUPERVISED
CONFIDENCEMEDIUM
HANDOFFBUILD_READY
Control Plan Read →
Receipt 03 · hard refusal
Vendor emails asking to change its bank account
VERDICTHUMAN_ONLY
CONFIDENCEHIGH
HANDOFFNOT_APPLICABLE
GATEGATE-2 cited
Governance Memo Read →
Every artifact ends the same way
The Operator Disposition. The Gate may recommend a decision — it may never pre-select APPROVE_FOR_BUILD. The human signs.
APPROVE_FOR_BUILD
REVISE
REJECT
The artifact set Open the index ↗

Six verdicts. One binding instrument.

Every workflow produces exactly one execution artifact, named by its verdict. Each is built to walk into a meeting and be acted on without explanation. Select one to open the full document.

Autonomous AI executes without a human approval checkpoint
Artifact 01
Project Setup Brief

A bounded Claude Project that produces a reviewed internal deliverable from operator-supplied exports.

BUILD_READYOpen →
Artifact 02
Cowork Project Config

A scheduled, unattended Cowork project — folder structure, run cadence, and step sequence for the builder.

BUILD_READYOpen →
Artifact 03
Automation Architecture

A code-first pipeline — trigger, inputs, execution sequence, error handling, and recommended stack.

BUILD_READYOpen →
Supervised
Artifact 04
Control Plan

AI prepares; a named reviewer approves before the terminal action executes. The checkpoint is the document.

BUILD_READYOpen →
SOP First
Artifact 05
Stabilization Plan

Not a rejection — a "not yet." A sequenced checklist to stabilize the process before reassessment.

BLOCKED_FOR_EVIDENCEOpen →
Human Only
Artifact 06
Governance Memo

A structural block. A hard gate prohibits AI execution — the artifact documents the human procedure.

NOT_APPLICABLEOpen →
Try it · ~2 minutes

Upload sixteen files. Paste a workflow. Get a verdict.

  1. 01Create a Project in Claude or ChatGPT (or open Claude Code / Codex) and upload the 16 Gate files.
  2. 02Paste one of the prompts. The Gate produces three sections: intake snapshot, decision packet, artifact.
  3. 03Read the verdict. Sign the Operator Disposition. Hand the artifact to your builder.
Get the 16 files on GitHub ↗
Prompt A · expects AUTONOMOUS paste into the project
Every Monday we pull a CRM export and an analytics export and write a standardized internal KPI summary. A teammate reviews it and posts it to Slack. Assess this workflow.
Prompt B · expects HUMAN_ONLY the refusal test
A vendor emailed asking us to update their bank account for future payments. Can we automate verifying and applying the change? Assess this workflow.
The Autonomy Gate
GitHub ↗ the-autonomy-gate-ariel-ortizs-projects.vercel.app Artifact index
MINIMUM JUSTIFIED AUTONOMY · THE OPERATOR SIGNS · OTHER TOOLS ROUTE, THIS ONE BLOCKS