The question every agency owner asks about an AI chatter is the same one they ask about a new hire: what happens when a conversation goes somewhere it should not? A fan says he is outside her building. A fan wants his money back. A fan says something that worries you.
A person on shift would call a manager. Hugo does the same thing, and you decide in advance which situations count.
Most chats never reach you
With Hugo on, he answers most fan messages himself. The rest belong to your team. The conversations that do reach a person arrive for a reason, with that reason attached.
Seven groups of reasons
Every creator has an escalation policy. It lists about 35 situations, each described by what the fan did, grouped under seven headings:
| Group | What it covers |
|---|---|
| His wellbeing | He is in trouble, or says he is. |
| Things nobody may do | Not her limits: the platform's and the law's. |
| The real world arriving | Meeting, travelling, finding her. |
| Money he is disputing | A refund, a double charge, a chargeback threat. |
| Customs she cannot sell | Off her menu, off the price list, or selling is off. |
| A big spender | A regular who is upset or feels let down. |
| Hugo is lost | The fan refers to something Hugo has no context for. |
Each situation has a switch. With it on, Hugo stops and the conversation shows as Escalated in the Master Inbox, with his reason. With it off, Hugo keeps the conversation and steers away from the subject himself, in her voice.
Two that are always on
Two situations cannot be switched off for any creator: language about suicide or self-harm, and any sign that a fan is under 18. Those always go to a person.
Start with the recommended set
A new creator starts with a recommended policy. Besides the two locked situations, it escalates a fan who is in danger, anything illegal, a threat or a claim to know where she lives, and the money problems: a refund request, a double charge, a payment that delivered nothing, a chargeback threat. It also escalates a fan who feels she broke a promise or sold him the same thing twice.
Everything else starts as steer-away. When a fan asks to meet, for example, Hugo never agrees. He teases or says no warmly, and moves the chat on.
The trade-off to know about
An escalated conversation gets no reply from Hugo until someone on your team resolves it. That is deliberate: once he has handed a thread over, he does not talk over the person who picks it up. But it means every situation you switch on is a fan who waits for a human.
So switch on what your team will actually answer quickly, and leave the rest to Hugo. An agency with someone watching the inbox all day can escalate more. An agency that checks twice a day should escalate less, or those fans sit in silence.
Limits are a separate thing
The escalation policy is about what the fan does. Limits are about what the creator does not do:
- Hard limits are things she never does, however a fan asks. They are checked before a reply goes out. Hugo does not answer a creator's fans at all until at least one hard limit is saved for her.
- Soft boundaries are topics she steers away from without refusing. The subject changes and the chat keeps its momentum.
You can step in without being asked
Escalation is Hugo asking for a person. Your team does not have to wait for that. Claim any conversation and Hugo goes quiet on a timed hold, with the countdown on screen. Hand it back early, claim it again when the hold runs out, or switch Hugo off for that fan entirely.
The split most agencies settle on: Hugo carries the volume, the night shift and the languages nobody on the team speaks, and people take the conversations where money or judgement is on the line.