What is human-in-the-loop AI?

It is AI that completes the preparation but stops before the decision, handing consequential choices to a person. The assistant does the gathering, drafting, and checking; the human approves what actually happens.

Which actions need my approval?

Anything consequential — actions that affect other people, money, commitments, or things that are hard to undo. Guardian treats the ability to act as separate from the ability to know.

Can I give Guardian standing permission for routine things?

Yes. Authority can be granted for narrow, defined situations so routine work does not need repeated approval. Those grants stay visible and can be withdrawn at any time.

What happens after Guardian acts?

It verifies the result rather than assuming success, and reports what actually happened — including when something failed or is still pending.

Human-in-the-loop AI

Human-in-the-Loop AI: Guardian Assists, You Decide

A personal AI assistant that works for your life — not the other way around.

Human-in-the-loop AI does the preparation and stops at the decision. Guardian Daily works this way by design: it suggests options, prepares the work, waits for your approval on anything consequential, then verifies what actually happened. Authority is granted in narrow scopes, stays visible, and can be withdrawn at any time.

Start Guardian Daily See how Guardian works

Opens the Guardian Daily app. You may need to sign in or create an account.

The line

Guardian must know when intelligence should not become action

Capability is not permission. The most important judgment an assistant makes is recognising the moment a decision belongs to a person.

The loop

Suggest, prepare, approve, verify

Suggest

Guardian proposes an option, explains the reasoning, and names what it is uncertain about.

Prepare

It does the work up to the threshold: the draft, the booking details, the rescheduled plan — ready to review.

Approve

You decide, with the consequence visible. Nothing consequential proceeds on assumption.

Verify

After acting, Guardian confirms the real outcome and reports failures and waiting states honestly.

Approvals

An AI assistant with approvals you can actually see

Approval is only meaningful when it is legible. A confirmation dialog with no context is not consent.

What is being done, on whose behalf, and what it affects — before you approve.
Standing authority for routine work stays narrow, visible, and revocable.
Open loops and waiting states are surfaced rather than silently held.
Nothing about learning or improvement widens what Guardian may do.

Why it matters

Agency is the point, not the constraint

Automation that quietly absorbs decisions returns time and takes something more valuable. Guardian gives time back while leaving conscience, judgment, and responsibility with people.

Questions

Common questions

What is human-in-the-loop AI?: It is AI that completes the preparation but stops before the decision, handing consequential choices to a person. The assistant does the gathering, drafting, and checking; the human approves what actually happens.
Which actions need my approval?: Anything consequential — actions that affect other people, money, commitments, or things that are hard to undo. Guardian treats the ability to act as separate from the ability to know.
Can I give Guardian standing permission for routine things?: Yes. Authority can be granted for narrow, defined situations so routine work does not need repeated approval. Those grants stay visible and can be withdrawn at any time.
What happens after Guardian acts?: It verifies the result rather than assuming success, and reports what actually happened — including when something failed or is still pending.

Keep reading

Explore Guardian by need

Four short guides covering the questions people ask most before they start.

AI assists. You decide.

See what it feels like when preparation is automatic and decisions stay yours.

Start Guardian Daily See Guardian Daily