CypherMonkeyAI omni-channel lab

A lifelong-learner lab on AI omni-channel agents, including voice and the contact center. Free to read. Nothing sold.

Agents · L2 · 15 min

Human approval

Learn in five minutes

  • Some actions are hard to undo: sending, deleting, paying, changing a record.
  • The loop may draft. A person approves.
  • Approval is a step in the trace, not a feeling.
  • The approver needs the draft and the passages.
  • This example uses no real customer and no payment.

A refund sentence stays a draft until a named role approves it. The model does not approve itself.

This section HTML

Visual

T5-R4 diagram

Diagram SVGDiagram PNGDiagram HTML

A draft and a gate

  1. The tool prepares the sentence 'example refund of 0 units'.
  2. The state is draft.
  3. A person marks approved or rejected.
  4. Only approved may be called sent, and this lab does not send it.

This section HTML

Quiz

Who approves an undo-hard action in this example?

The model drafts.

What must the approver see?

Approval without the source is a rubber stamp.

Does this lab send the example?

The example stops at the gate.

This section HTML

Ecosystem

Approval
A recorded human decision before an undo-hard action.

Sisters: Workflows, State machines, Function calling.

The same loop can sit behind chat or a voice turn. The channel does not remove the cap.

MCP is one way to plug tools into that loop. It is the next track.

Config

  • Write the goal in one sentence.
  • Name the input and the allowed output.
  • Save the trace next to the result.

Troubleshooting

  • The answer looks fluent and is wrong. Check whether a source passage actually contains the claim.
  • The loop never stops. Set a step cap and stop when the tool result repeats.

This section HTML

Sources HTML

Sources and go deeper

sources.json