Human handoff
Also called live handoff, escalation to a human, agent transfer.
Human handoff is the moment an AI agent stops answering and passes the conversation to a person on your team. It happens when the customer asks for a human, or when the question needs judgment the AI shouldn't make. A good handoff brings the whole conversation along, so the customer never has to repeat themselves.
Doug's Field Guide · Updated
When the AI should hand off
Write the rules down before you turn anything on. Most businesses land on a short list.
Anyone who asks for a person gets one, no arguing. Emergencies go to a person right away, like a gas smell, water pouring through a ceiling or a patient in pain. Anything that sounds like advice goes to a person: a diagnosis, a legal opinion, whether insurance covers a procedure. Angry or upset customers go to a person. So do custom quotes, where the price depends on seeing the job.
Everything else, the AI can usually handle. Hours, service area, what a standard service costs, what to bring, how to prepare, and booking a normal appointment. That split keeps your team on the conversations that need them.
What a good handoff looks like
The customer should know what's happening. A line like 'I'm getting someone from the team, they'll reply here' is enough. If nobody is available, say so and give a real next step, like a callback time, instead of leaving them staring at a typing indicator.
The person taking over should see the full conversation, the customer's name and contact details, and what the AI already told them. Nothing frustrates a customer faster than explaining the leak a second time.
Afterward, look at why handoffs happened. If the AI handed off because it didn't know your weekend rates, that's a gap in your information you can fill. If it handed off because the customer was upset, that was the right call.
Write your handoff rules
- List the questions only a person on your team should answer, like emergencies, diagnoses and custom quotes.
- Decide who takes handoffs during business hours and what the customer is told after hours.
- Write the exact sentence the customer sees while they wait for a person.
- Review last month's messages and mark which ones should have gone to a person sooner.
Common mistakes
- Making customers fight the AI to reach a person, which turns a fixable question into an angry review.
- Handing off without the conversation history, so the customer has to start over.
- Promising a person will reply right away when nobody is watching the inbox after 5pm.
How Doug does it
Doug hands a conversation to a person on your team when a customer asks or when a question needs one, with the full conversation attached. On the Growth plan ($149 a month) your team can jump into a live conversation. Every question Doug couldn't answer shows up with a drafted answer for you to approve.
Terms on this page
- Callback
- A promise to call a customer back at a set time. A callback with a specific time keeps a lead warm, while 'someone will call you' often loses it.
- Deflection
- When an AI agent answers a question fully so it never reaches your team. Deflection only counts if the customer actually got what they needed.
- Handoff
- Your team jumping into a live conversation, with the customer still typing in the same window. You can also tell Doug to always hand off certain topics.
- Resolution rate
- The share of conversations an AI agent finishes without a person stepping in, like a question answered or an appointment booked.
Keep exploring
- AI receptionistSoftware that answers customers, takes their details and books them, day or night.
- How Doug learns your businessYour website, files, and Q&A pairs become the only place his answers come from.
- FAQ pagesThe questions customers ask on the phone, each with one short, clear answer.
- Speed to leadHow fast you answer a new inquiry. The first reply usually wins the job.
Sources
Written by Doug's team from these sources and our own work with local businesses. Search engines change their rules, so check the source before a big decision.
