H33 · quick start
your first 30 minutes: 00:00
step 1 of 4
Question 1 of 4

Who is acting?

Tell us about the actor that needs authority. Pick one to start; you can add more later.

Every decision in your business has an actor behind it: a person, an automated agent, or a system integration. We start here because authority is meaningless without a named actor. The receipt at the end will record who this is — and from then on, every decision they make will be traceable back to this moment.
Person

Someone on your team

An employee, contractor, or partner who needs authority to make decisions on behalf of your organization.

AI Agent

An automated agent

An AI assistant, model, or autonomous workflow that takes actions without a human pressing a button each time.

System

A service or integration

A backend system, API, or third-party platform that acts on your organization's behalf via a service account.

Question 2 of 4

What can they do?

Choose the kinds of decisions this actor is allowed to make. Pick anything that applies. You can refine later.

Authority is a positive list, not a negative one. Picking the decisions this actor can make is more important than picking everything they can't — because anything not on this list will require a human or another authority to step in. Receipts will record which of these capabilities each decision used.
So far
Actor:

Approve customer transactions

Trades, payments, transfers, settlements

Move funds between accounts

Internal transfers, treasury operations, reserves

Issue refunds or credits

Customer disputes, goodwill, billing corrections

Access customer data

Read accounts, profiles, transaction history

Modify customer accounts

Update settings, permissions, contact info

Send communications to customers

Email, SMS, in-product notifications

Change organizational policies

Pricing, terms, eligibility, risk parameters

Sign contracts on our behalf

Vendor agreements, NDAs, partnership documents

Question 3 of 4

Within what limits?

Set the boundaries inside which this actor operates autonomously. Anything outside these limits will need a human or escalation.

Limits make authority safe. Without them, a single compromised credential or a runaway agent can do unbounded damage before anyone notices. The numbers you pick here become part of every receipt — and any decision exceeding them is automatically flagged for review or rejected outright.
So far
Actor:
Can do:
The largest single transaction, transfer, or commitment this actor can authorize. Leave blank for no cap.
Decisions outside this window will be rejected.
Geographic boundary for the actor's authority.
Question 4 of 4

When do humans get involved?

Pick when this actor needs a second pair of eyes. The right answer is rarely "always" or "never" — somewhere in between is usually best.

Even with capabilities and limits set, some decisions need human judgment — especially the largest, the rarest, and the ones that touch sensitive surfaces. The escalation rule you pick here is recorded in the receipt and enforced at decision time, so nothing important slips through unobserved.
So far
Actor:
Can do:
Up to:
Full autonomy

Never

This actor operates entirely within their limits. Humans only see receipts after the fact. Good for high-velocity routine decisions.

Threshold

Above a threshold

Decisions over a certain amount or sensitivity automatically wait for human approval. Everything below is autonomous. The recommended default.

Always

Always

Every decision waits for a human. Slowest, but the safest possible mode for the highest-stakes actors.

Exceptions

Only when limits are hit

Autonomous until the actor would exceed a limit or hit a capability they don't have. Then humans decide whether to extend authority.

You're set up.

Your first authority is live. Every future decision this actor makes can be checked against the rules you just defined — and explained to a regulator, an auditor, or your CFO.

Completed in 00:00 — well within 30 minutes
Receipt

Authority created

DecisionCreate authority for
VerificationVALID
Authority covers
Up to
Escalation
Active
Evidence · linked to event log
Issued
Who changed it?
Inspect the receipt

    

Where to next

Who changed it? You — this wizard session
Same replay engine as the server