Keyword auto-replies that don't feel robotic
A practical pattern for building keyword-triggered automations that still feel like a human answered.
Key takeaways
- Match on intent phrases, not single words, to cut down on false triggers.
- Always give the flow an escape hatch to a human agent.
- Keep automated replies under three sentences — long replies read as a bot.
On this page
Keyword automations are the fastest way to handle repetitive questions — pricing, hours, order status — without hiring a 24/7 team. Done badly, they frustrate customers who feel stuck talking to a wall. Here's how to avoid that.
Match phrases, not words
Triggering on the single word "price" catches "what's your price" but also "I priced this elsewhere and left", which needs a human. Match on short phrases instead, and keep the list specific to your business.
Always offer a way out
Every automated flow should end with an option to reach a person. A simple "Reply AGENT to talk to our team" line, honored immediately, keeps automation from feeling like a trap.
| Pattern | Good for | Avoid for |
|---|---|---|
| Single keyword | Store hours, simple FAQs | Anything with pricing nuance |
| Phrase match | Order status, returns | Open-ended complaints |
Keep replies short
Three sentences or fewer reads as a quick, helpful answer. Longer than that, and it reads as a script — even when a human wrote it.
Frequently asked questions
Can automations run alongside a live agent?
Related articles
Avoid a WhatsApp ban: messaging limits explained
What actually gets WhatsApp numbers banned, and the pacing rules SpeedWhat enforces automatically to keep your...
Connect your first WhatsApp number in under 10 minutes
A step-by-step walkthrough of linking a WhatsApp number to SpeedWhat, whether you use the QR-based connection...