E-commerce teams on WeChat answer customers in two different arenas: one-to-one direct messages, and group chats — buyer groups, VIP groups, store announcement groups. Many teams point one auto-reply bot at both and wonder why it either goes silent in groups or spams them. The failure is structural: a group chat is not a private inbox with more people in it. It needs its own routing rules.
Why group chats break naive auto-reply
A naive bot replies to every inbound message. In a DM that is correct. In a group of two hundred buyers it produces three failure modes at once:
- Noise amplification. Ten members discussing sizing among themselves get ten bot answers. Members mute the group, and the store loses its one direct channel to them.
- Answer misattribution. The bot answers member A's question in a thread where member B asked something else seconds earlier. Both customers now distrust the answer.
- Risk exposure. Aggressive auto-posting in groups is exactly the behavior WeChat's risk systems flag. An account that took months to build can get restricted in an afternoon.
A routing table that holds up
Treat DM and group as two rule sets that happen to share one knowledge base:
- DM rule: every inbound message enters the reply pipeline. Speed matters; a first reply under a minute is achievable and expected.
- Group rule: respond only to explicit triggers — an @-mention of the service account, or messages containing configured keywords like "order" or "refund." Everything else is read-only.
- Contact filter: in groups, whitelist which senders the bot may address. In practice that means the store's own staff and verified buyers, not every member.
- Rate cap: a per-hour ceiling on bot posts per group, separate from the DM cap. When the ceiling hits, the bot holds messages for human review instead of queueing them for later spray.
The trigger design matters more than the answer quality. A bot that answers one @-mention correctly per hour builds more trust than one that answers forty messages inclusively.
Session isolation is not optional
Context bleed is the quiet killer. If a buyer asks about a refund in a DM and the same bot instance carries that context into a group reply, it can leak personal transaction details into a public thread. Each DM is one session; each group is one session; contexts never cross. This also keeps metrics honest — per-session transcripts are the raw material for quality checks, and merged sessions make those checks meaningless.
What to measure after switching routing on
Three numbers tell you whether the routing table works: the share of group messages that required a human takeover (should trend down, not to zero), the bot's answer-to-mention ratio in groups (should stay near 1:1 — more answers than mentions means over-triggering), and DM first-reply time (should not regress after the change, since the same engine now carries two rule sets).
Weike AI Assistant exposes DM and group routing as separate rule sets over one shared reply library, which is the layout this article describes — the alternative, per-channel reply libraries, doubles the maintenance work and drifts out of sync within weeks.
Top comments (0)