Training your team to work with an AI agent: the four habits, the first-week routine, and what to stop doing

An AI agent does not replace a receptionist; it changes the job. The receptionist stops typing the same twelve answers and starts handling the twelve conversations that need a person. Teams that are told this, and given a routine, adopt it in a week. Teams that discover the agent by surprise resent it. Here is the routine.

By Ritchie, ReplyKit · Published 5 September 2026 · 6 min read

Quick answer

Give the team four habits: check the inbox for chats waiting for a person at set times, take over by simply replying (the agent pauses), hand back when done, and teach the agent from the unanswered-questions list weekly. Assign one owner of the inbox per shift. Run a one-week routine of reading conversations together. Tell the team plainly what the agent does and does not do, and that their judgement is the part that cannot be automated.

The four habits The first-week routine What to stop doing What to tell the team FAQ

The four habits

Check the waiting list at set times. The inbox flags chats waiting for a person; the owner also gets an email. Three fixed checks a day beats constant monitoring: opening, after lunch, before closing.
Take over by replying. Nothing to click. The first human reply pauses the agent on that conversation. Read the transcript first; the customer has already explained.
Hand back when done. Click hand back, or let the automatic return do it after the quiet period set in Settings. Handing back means routine follow-ups are handled without the team.
Teach weekly. Open the unanswered-questions list every Friday. Each entry is a question the agent could not answer; type the answer once and it is knowledge from then on.

The first-week routine

Day 1. Show the team the test chat. Let them ask it the hard questions and watch it hand over. This removes the fear that it will say something wrong to a customer.
Day 2. Read yesterday's real conversations together for ten minutes. Note anything the agent said that the team would have said differently, and change the instructions.
Day 3. Practise takeover and hand-back on a live chat.
Day 4. Assign inbox ownership by shift and put the three check times in the roster.
Day 5. First teaching session from the unanswered list. Explain that this is how the agent gets better, and that they own it.

What to stop doing

Stop answering questions the agent already answered. Read the transcript before replying.
Stop apologising for the agent. "Sorry, that was the bot" undermines it; say "I'll take it from here".
Stop keeping a separate notebook of leads. The leads list is the notebook, and the agent fills it in.
Stop changing the instructions without re-running the twenty-message test.

What to tell the team

Say it plainly: the agent answers repeat questions and books appointments; you handle complaints, judgement, relationships and anything unusual. It is a tool that makes the job less repetitive, not a replacement. Show the handover rules in the instructions so they can see exactly when it stops. Then let them own the teaching. A team that has taught the agent its answers treats it as theirs. Our guide to handover covers the mechanics in detail.

Frequently asked questions

How does a team member take over from the AI?

By replying in the conversation. The first human reply pauses the agent on that chat; nothing else is needed.

How does the chat go back to the AI?

Click hand back in the inbox, or let it return automatically after the quiet period set in Settings.

Who should own the inbox?

One person per shift, with three fixed check times a day. The owner also receives an email for each handover.

How does the team improve the agent?

From the unanswered-questions list, weekly. Each answer typed once becomes knowledge.

Will staff feel replaced?

Less so when told plainly what the agent does and does not do, and when they own the teaching.

Give the team a colleague, not a surprise

Test chat, inbox, takeover and teaching, all in one place. 7-day free trial.

Start free

Written by Ritchie, ReplyKitPart of the small team in Malaysia that builds and runs ReplyKit. Writes about WhatsApp automation, AI customer service and small business lead response.

How we know this. The routine below is what we walk new ReplyKit teams through, built around the inbox, takeover, hand-back and the unanswered-questions list as they exist in the product. Product figures in this article (limits, prices, per-reply costs) are taken from ReplyKit as it runs today and are re-checked when we update the page. Where we cite outside research, the source is linked below. We sell ReplyKit, so read our product claims with that in mind; we say where a different tool or no tool is the better choice. About ReplyKit.

Sources and further reading

Harvard Business Review, The Short Life of Online Sales Leads (Oldroyd, McElheran, Elkington, 2011) · firms that responded within an hour were about seven times more likely to qualify the lead
ReplyKit Terms of Service, section 4 · human oversight of AI-generated messages

ReplyKit is an independent product and is not affiliated with or endorsed by Meta or WhatsApp. This article is general guidance for running a small business, not legal, financial or regulatory advice.