Skip to content
← Writing
Writing · October 2026 · 2 min read

Approving everything means catching nothing

Making a person approve every AI agent action feels safe. It is not. Reviewers tire, and the risky action slips through. Approve by risk instead.

Abhijit Chatterjee Chattopadhyay

The Logic Atelier logo above the words: Approving everything means catching nothing.

If a person approves everything your AI agent does, nobody is checking anything. It is the safest-sounding rule for agents, and the one most likely to let the bad action through.

Why it fails

Picture the reviewer on day one. They read every request carefully. By week three, there are two hundred a day, almost all fine, and they click approve.

AWS puts it plainly in its Trust in Depth (opens in a new tab) guidance: as agents take on more work, "human approvals pile up and review fatigue sets in. People asked to check everything stop catching the things that matter."

So you pay twice. The agent waits on people, and the people stop protecting you.

Limits belong outside the agent

The fix is not more trust in the agent. It is moving the limits out of it.

A prompt that says "never spend more than $100" is a request. A payment service that refuses anything over $100 is a limit. AWS makes the same point. Put the boundaries in the environment around the agent, because, as they put it, "it cannot turn off a control it does not run".

Check by rule, ask by risk

Sort every action your agent can take into one of three groups:

  1. Blocked by rule. Things it must never do. Remove the tool or the permission, so it cannot.
  2. Checked by rule. Things a system can verify: within budget, right customer, right format. Let it act, and let the check decide.
  3. Asked by risk. Costly, hard to undo, or genuinely ambiguous. These, and only these, go to a person.

In that picture, your reviewers now see ten requests a day instead of two hundred, and every one deserves their attention.

Watch what it did

Green dashboards are not proof. As AWS notes, "a healthy system is not the same as a successful task". Keep a trace of what the agent saw, what it called and what it decided, and read the ones that went wrong.

Do this today

  • Count how many approvals each reviewer handles per day, and how many they rejected last month.
  • If the rejection rate is near zero, that queue is a rubber stamp, not a control.
  • Move each approved-every-time action into "checked by rule".
  • Keep people for the actions that are costly or hard to undo.

Your next step

I write one practical note like this every day. Sign up just below to get them by email.

Drafted with AI help, edited and approved by me.

New articles by email

I will never sell your details, for any reason, and I never send spam.

How I use your details
  • I, Abhijit Chatterjee Chattopadhyay of The Logic Atelier, hold them, in my own newsletter system hosted on Fly.io. Emails go out through Resend.
  • Your name is for greeting you, your email is for sending new articles, and your phone number, if you give it, is so I can reach you directly about my work.
  • If you do not confirm your email within 7 days, I delete the sign-up. If you unsubscribe, I delete your details 30 days later.
  • If Google Analytics is running on your visit, it is told that a sign-up was made, and nothing you typed.
  • To see or delete what I hold about you at any time, email hello@thelogicatelier.com.
  • The full detail is on the privacy page.
NextCheap tokens, expensive bill
← cd ../writing

Approving everything means catching nothing

approve-everything.mdfront matter
title: Approving everything means catching nothing
date: 2026-10-19
reading_time: 2 min
author: Abhijit Chatterjee Chattopadhyay
approve_everything_cover_664e7aa3_30b16d38b6.pngcover
The Logic Atelier logo above the words: Approving everything means catching nothing.

If a person approves everything your AI agent does, nobody is checking anything. It is the safest-sounding rule for agents, and the one most likely to let the bad action through.

Why it fails

Picture the reviewer on day one. They read every request carefully. By week three, there are two hundred a day, almost all fine, and they click approve.

AWS puts it plainly in its Trust in Depth (opens in a new tab) guidance: as agents take on more work, "human approvals pile up and review fatigue sets in. People asked to check everything stop catching the things that matter."

So you pay twice. The agent waits on people, and the people stop protecting you.

Limits belong outside the agent

The fix is not more trust in the agent. It is moving the limits out of it.

A prompt that says "never spend more than $100" is a request. A payment service that refuses anything over $100 is a limit. AWS makes the same point. Put the boundaries in the environment around the agent, because, as they put it, "it cannot turn off a control it does not run".

Check by rule, ask by risk

Sort every action your agent can take into one of three groups:

  1. Blocked by rule. Things it must never do. Remove the tool or the permission, so it cannot.
  2. Checked by rule. Things a system can verify: within budget, right customer, right format. Let it act, and let the check decide.
  3. Asked by risk. Costly, hard to undo, or genuinely ambiguous. These, and only these, go to a person.

In that picture, your reviewers now see ten requests a day instead of two hundred, and every one deserves their attention.

Watch what it did

Green dashboards are not proof. As AWS notes, "a healthy system is not the same as a successful task". Keep a trace of what the agent saw, what it called and what it decided, and read the ones that went wrong.

Do this today

  • Count how many approvals each reviewer handles per day, and how many they rejected last month.
  • If the rejection rate is near zero, that queue is a rubber stamp, not a control.
  • Move each approved-every-time action into "checked by rule".
  • Keep people for the actions that are costly or hard to undo.

Your next step

I write one practical note like this every day. Sign up just below to get them by email.

Drafted with AI help, edited and approved by me.

subscribeemail

// I will never sell your details, for any reason, and I never send spam.

How I use your details
  • I, Abhijit Chatterjee Chattopadhyay of The Logic Atelier, hold them, in my own newsletter system hosted on Fly.io. Emails go out through Resend.
  • Your name is for greeting you, your email is for sending new articles, and your phone number, if you give it, is so I can reach you directly about my work.
  • If you do not confirm your email within 7 days, I delete the sign-up. If you unsubscribe, I delete your details 30 days later.
  • If Google Analytics is running on your visit, it is told that a sign-up was made, and nothing you typed.
  • To see or delete what I hold about you at any time, email hello@thelogicatelier.com.
  • The full detail is on the privacy page.

vi.Contact

06 · contact

Talk to me.

Talk to me.

Production AI and software architecture. Based in Dubai, working with teams in the Emirates and Ireland.

I work Dubai hours, Monday to Friday, the same week you do. Dubai is three hours ahead of Dublin in summer and four in winter, so my afternoon is your morning.

I will never sell your details, for any reason, and I never send spam.

How I use your enquiry
  • Your enquiry comes to me, Abhijit Chatterjee Chattopadhyay of The Logic Atelier. The form is handled by Formspree, which passes it to my inbox and keeps a copy in my Formspree account.
  • I use it to read your enquiry and reply. If no work follows, I delete it within 12 months, and sooner whenever you ask.
  • If Google Analytics is running on your visit, it is told that an enquiry was sent, and nothing you typed.
  • To see or delete what I hold about you, email hello@thelogicatelier.com.
  • The full detail is on the privacy page.