A chatbot that answers every question is rarely the goal. The best WhatsApp setups use automation for the routine and a person for everything else, with a hand-off between the two that customers barely notice. Most of the frustration people feel with chatbots comes from a bad hand-off: being stuck in a loop, or having to repeat everything to a human who has no context.
What a chatbot should handle
Automation is strongest where the answer is predictable:
- Opening hours, location and delivery areas.
- Prices and service lists.
- Order or booking status.
- Sending an invoice, brochure or directions.
- Collecting a few details so a person can help faster, such as name, request and preferred time.
- Booking or rescheduling an appointment.
What a person should handle
Hand over quickly when:
- The customer is upset or has a complaint.
- The request involves money, refunds or a negotiation.
- The question is unusual or the bot is not sure.
- The customer asks for a person, in any wording.
- A high-value lead shows buying intent.
Rules for a good hand-off
- Always offer a way out. Every flow should include "talk to a person", and it should work every time.
- Set expectations. Say when someone will reply: "A team member will reply within 10 minutes." If it is outside working hours, say so and say when they will hear back.
- Pass the full context. The person taking over should see the whole chat, the details already collected and the customer's history. Nothing is worse than asking the customer to start again.
- Assign an owner. A chat that is handed off should land on a named person or team, not in a pool nobody watches.
- Do not trap customers. If the bot fails to understand twice, hand over instead of asking a third time.
- Return gracefully. After a person resolves the issue, the customer can go back to automated messages without confusion.
Design your flow around real questions
Look at your last month of chats and list the ten most common questions. Build answers for those first. Add new answers only when you see the same question repeatedly. A small, accurate bot beats a large one that guesses.
Tone matters
Write the bot in the same voice as your team. Use short sentences. Avoid pretending to be a person. A simple opening such as "Hi, I'm the ScaleDesk assistant. I can help with common questions, or connect you with our team" is honest and works well.
Measure it
Track:
- Resolution without a person: how many chats the bot completed successfully.
- Hand-off rate: too low may mean customers are getting stuck; too high means the bot is not helping.
- Time to first human reply after a hand-off.
- Customer ratings or comments, when you collect them.
Where AI agents fit
Newer AI agents can handle more open-ended questions than fixed flows, but the same principle holds: clear limits and a reliable path to a person. Read our guide to guardrails for AI agents for how to set those limits.
A shared inbox platform such as TalkForGrow is designed to combine automated replies with team assignment, so a chat can move from automation to a person with its history intact. To see where this fits in the whole picture, start with what WhatsApp automation is.
Frequently asked questions
Should the chatbot say it is a bot?
Yes. Being upfront builds trust and prevents frustration when the bot reaches its limits.
How many questions should the bot cover?
Start with the ten most common ones and expand slowly. Accuracy matters more than coverage.
What if a customer messages outside working hours?
Let the bot collect the request and tell the customer when a person will reply, then make sure that promise is kept.
Run WhatsApp conversations with TalkForGrow
TalkForGrow is our platform for sending, automating and managing customer conversations on messaging apps such as WhatsApp, from one shared inbox.