Why I built SupportBridge around ONE rule: max 1 auto-reply per ticket

When I was designing SupportBridge, I had to make one fundamental decision:

How much should the AI be allowed to do?

Most AI support tools default to: "as much as possible."

I went the opposite direction.

Here's the one rule we built everything around:

The AI gets exactly 1 auto-reply per ticket. After that — human.

Here's why this matters:

Most support conversations that go wrong don't go wrong on the first message. They go wrong on the second or third — when the AI doesn't understand the customer's follow-up, guesses wrong, and confidently doubles down.

One reply is almost always safe. Two or three replies is where hallucinations compound.

So we cut it off at one.

The second rule we built around:

The AI can only use content from your approved FAQ. Nothing else.

No open-ended generation. No "filling in the gaps." If the answer isn't in your FAQ — the AI says it doesn't know and routes to a human.

This sounds limiting. In practice, it means:

• Zero hallucinations (you can't hallucinate from a constrained source)
• Zero brand damage from wrong answers
• Support team only handles what actually needs them

Currently in free pilot — you just forward your support@ email and see it in action.

What rules would you want an AI support tool to follow before you'd trust it?