Chapter 1 · The resource
The seven handoff triggers
Follow this course in your space: your progress is saved, and the community answers your questions.
Follow for freeA good AI assistant isn't measured by what it can answer. It's measured by what it knows not to handle alone. The moment it says "I'm passing this on" is the moment the customer decides whether to trust you.
In this lesson, you set the seven situations where your assistant hands off, no arguing, and you write the rule that goes into its instructions.
The seven triggers
1. The customer asks for a human
This is the simplest one, and the one most often missed. As soon as someone writes "I want to talk to someone," "a real person," "an agent," or "are you a bot?", the assistant stops answering the substance and offers the handoff. No "I'm sure I can help you myself." It can offer both options, but it never forces the AI on anyone.
Variations to cover in the instructions: "I want the owner," "call me," "phone number?", "who am I talking to?".
2. Two misses on the same question
If the assistant has answered off the mark twice on the same topic, it breaks the loop. A third try only adds frustration. The next lesson gives you the three exact steps: rephrase once, offer choices, then hand off.
3. A complaint
A damaged package, a job gone wrong, a disputed invoice, a delay. Your assistant can take in the complaint, ask two or three factual questions, then pass it on. It doesn't judge, doesn't get defensive, and doesn't offer any goodwill gesture. That's your call.
4. A refund, a cancellation, a goodwill gesture
Anything involving money or a commitment goes through you: refund, store credit, late cancellation, exceptional rescheduling, discount. France's data protection authority, the CNIL, makes the point in its chatbot guidance (2021): a conversation with a chatbot, with no human involved, can't by itself be the basis for an important decision about a person. Your assistant collects the details, you decide.
5. An emergency
A water leak, no heat in winter, a sick animal, a car that won't start, pain after a session. The assistant doesn't diagnose and doesn't reassure lightly. It gives your fastest contact method, and if someone could be in danger, it points to emergency services (911 in the US, 112 in the EU).
6. Strong emotion
Anger, distress, insults, ALL CAPS, a customer writing "this is unacceptable" or "I'm desperate." AI isn't built to handle strong emotion, even with nice phrasing. It acknowledges what the person is feeling in one sentence, then hands off.
7. Outside its information
Your assistant answers with your information: your services, your hours, your policies. If the question goes beyond that (a special case, a custom quote, specific legal, medical or technical advice), it doesn't guess and doesn't look elsewhere. It says it doesn't know and hands off. A made-up answer costs far more than "I'm passing this on."
What the assistant does at the moment of handoff
Whatever the trigger, the handoff always follows the same order:
- Say what's happening: "I'm passing your request to [name]."
- Say when: a response time you actually keep (lesson 3).
- Ask for what's missing: a way to get back in touch, and the useful detail if it hasn't been given.
- Prepare the summary for you (lesson 4).
- Promise nothing else.
The rule to copy into your assistant's instructions
WHEN TO HAND OFF TO [YOUR NAME]
You hand off, without insisting on answering yourself, in these cases:
1. The person asks to talk to a human, the owner, someone,
or asks whether you're a bot (you confirm you're an AI).
2. You've answered twice without resolving the same question.
3. The person makes a complaint or is unhappy with a service.
4. The request involves a refund, a cancellation, store credit,
an exceptional reschedule or a discount.
5. The situation is urgent: [your emergencies: leak, outage, pain...].
If someone could be in danger, you tell them to call 911.
6. The person expresses strong emotion: anger, distress, insults.
7. The answer isn't in your information.
When you hand off:
- you say you're passing it to [your name] and when ([your response time]);
- you ask for a way to get back in touch if it's missing;
- you write the handoff summary;
- you promise no refund, no service date,
no goodwill gesture, no diagnosis.
Triggers specific to your line of work
The seven triggers apply to everyone. Then add the ones for your trade:
| Business | Extra trigger |
|---|---|
| Plumber, HVAC | Active leak, smell of gas (point to the gas utility's emergency line) |
| Physical therapist, chiropractor | Unusual pain, a question about a treatment |
| Vacation rental, B&B | Guest already on site with a problem (key, heat, water) |
| Auto shop | Car won't start, red warning light on |
| Lawyer | Any question about an open case or a court deadline |
| Hairdresser | Skin reaction after a color treatment |
| Photographer | Request to edit or remove a published photo |
| Bakery | Allergy concern about an order already placed |
Common mistakes
- Hiding the handoff behind three menus. If people have to hunt for it, they leave.
- The conditional handoff: "Before I connect you, could you try...". No.
- The fake human: letting people think the AI is a person. In the EU, Article 50 of the AI Act requires people to be clearly told they're interacting with an AI, at the latest at the first interaction; lying at the moment of handoff breaks everything.
Your turn
- Copy the rule above into a document.
- Replace every bracket: your name, your response time, your emergencies.
- Add one or two triggers specific to your trade, using the table.
- Test your assistant with seven messages, one per trigger. Note the ones where it doesn't hand off: those are the first to fix.
Recommended tool
Flopia, the AI assistant from Flopix
Flopia is in testing. Join the waitlist to hear as soon as it opens.
Premium
Unlock every course of the group Flopia
Every lesson, new courses as soon as they come out, and the community. No commitment.