The most common mistake small businesses make with DM automation isn't automating too much — it's treating it as a binary switch instead of a spectrum. The better question isn't "should I automate my DMs," it's "which of my DMs are repetitive enough to automate safely, and which ones need a person."
Good candidates for automation
- Pricing and product-availability questions you've answered the same way dozens of times
- Shipping and return-policy questions that are already written down somewhere
- Restock and launch-timing questions during a drop
- "How do I order" and "where do I buy this" — the highest-intent, lowest-nuance messages you get
Conversations that still need a human
- Complaints, damaged items, or anything involving a refund
- Anything ambiguous enough that a wrong answer would embarrass the brand
- Partnership, wholesale, or press inquiries
- Anything an AI flags as low-confidence — that flag exists for a reason
The right mental model
Think of automation as clearing the repetitive floor traffic so you have bandwidth left for the conversations that actually require judgment — not as a way to avoid talking to customers altogether. The businesses that get the most value out of DM automation are the ones that keep a close eye on what's being flagged for review, especially in the first month, and use that as a signal for what their AI still needs to learn about their catalog, their policies, and their voice.