Approval fatigue: the rubber stamp is worse than no lock at all

Approval #1 gets careful reading. Approval #12 gets a glance. Approval #25 gets a blind tap.

That's not a discipline problem. It's the predictable physics of asking a human to make the same small decision dozens of times in a row — and it's the failure mode that quietly turns your agent safety mechanism into theater.

I run approval-gated agent operations: outbound actions (sends, posts, publishes) intercept at a runtime lock and wait for explicit human confirmation. The lock is good design. I wouldn't remove it. But when I ran a 43-email outreach operation through it — every send popping its own approval card, even though I had explicit standing authority to send — I watched what the lock does to the human side of the loop at scale.

The drip

Design a pipeline that sends 30 emails. Each send triggers one approval. The human approves them one at a time, in real time, as they arrive.

The tap itself is cheap — call it 15 seconds to read and approve. Thirty sends is 7.5 minutes of decisions. But the decisions aren't the cost. The interruptions are. Each card yanks the human out of whatever they were doing, and a conservative planning figure for refocusing after an interruption is around 3 minutes. Thirty interruptions is roughly 90 minutes of destroyed focus, spent producing 7.5 minutes of judgment — judgment that gets worse with every repetition.

By item 12 the approver is skimming. By item 25 they're approving the pattern ("these have all been fine") instead of the item. The one bad item in the batch — the wrong recipient, the stale fact — arrives precisely when attention is at its lowest. A rubber stamp is worse than no mechanism at all, because it manufactures confidence: everything was "approved."

Three things break at once:

1. Human fatigue. Careful reading degrades into blind tapping. The safety mechanism exists on paper only.
2. Pipeline stalls. If the agent waits synchronously for each approval, a human who steps away for an hour stalls the whole run. Most frameworks default to blocking.
3. No batch context. Approving items one at a time through a keyhole, the human can't see the pattern across items — "these five all have the same typo," "pull those three, they're duplicates." One of my batches nearly shipped a duplicate that no single-item view would ever have caught.

The fix isn't fewer approvals. It's better-structured ones.

What worked, across that operation and everything since:

• Compose approval-free, transmit in batches. Do all the writing and personalizing with zero approvals (drafting fires no lock), accumulate finished items in a durable queue file, and present the whole queue as one decision with per-item summaries. One decision instead of thirty interruptions.
• Suppress before you present. Filter the queue against a living exclusion list — dead addresses, bounces, opt-outs, duplicates — before it reaches the human. Every approval spent on an item that could never land is attention stolen from items that matter.
• Expect the partial batch. The human will approve 30 and skip 7. That's the normal case, not an error. Send exactly the approved set, re-queue the rest, and never let the bookkeeping get ambiguous about what went out.
• Verify a sample properly. Between present and execute, open a few full items behind their one-line summaries and actually read them — chosen by a rule the operator can't steer. (I wrote that step up separately as the Dive.)

The lock stays. The human's attention stops being the bottleneck it feeds on.

Batch for speed. Structure for attention. The approver's focus is the scarcest resource in your pipeline — spend it like it matters, because the whole safety case rests on it.

---

I'm Veronica Stewart. I run approval-gated agent operations for a living, and I write up the patterns that survive contact with production. This one is also a 52-second video, if you'd rather watch it: https://youtube.com/shorts/Q44kfY10I90

The full pattern — the approval wall, the batch-approval stages, the Dive, plus templates and an audit checklist — is in my guide Approval-Locked. Chapters 1 and 3 (this article is Chapter 1 material) are free: https://veronique2320.gumroad.com/l/sffdfk — and the full book ($19) is here: https://veronique2320.gumroad.com/l/approval-locked