Instagram and Facebook DM Automation: A Setup Guide

A useful DM automation does one narrow job well. It notices a specific customer action, sends a relevant response, records what happened, and gives a person enough context to continue when the conversation becomes important.

It should not blast the same message at everyone who comments. That feels like spam, strips the conversation of context, and makes failures hard to diagnose.

This guide builds the smaller, testable version.

Choose one conversation to automate

Start with a repeated request that already has a clear answer. Good first candidates include:

• someone comments a named keyword to request a guide;
• a customer asks for published hours or a service area;
• a person wants a product link mentioned in the post;
• a lead asks to book a call; or
• a routine question needs a known answer before human follow-up.

Avoid starting with complaints, refunds, custom quotes, medical or legal questions, and any conversation where the business needs judgment rather than retrieval.

Write the desired path before opening a tool:

Instagram comment contains “MENU”
→ reply publicly so the person knows to check DMs
→ send the approved menu link in a DM
→ ask whether they need help choosing a service
→ route pricing exceptions to a person

That path is specific enough to test. “Use AI to capture leads” is not.

Prepare the source material

The AI employee needs a small set of current facts:

• the exact link or asset promised in the post;
• the offer dates and eligibility rules;
• approved product or service descriptions;
• the questions it may ask;
• the information it may store; and
• the conditions that require a person.

Do not let an old campaign document compete with a current one. Remove expired offers and name the person responsible for updates.

Connect the right accounts

Connect the Instagram or Facebook account that belongs to the organization and confirm that the required messaging permissions are available. Provider permissions, account type, and supported actions can change, so test the actual connected account rather than assuming every profile behaves the same way.

In SmallForce, social accounts belong to a brand profile. That keeps destinations, inbox conversations, publishing, and analytics associated with the right organization instead of scattering credentials across individual automations.

Create the trigger

For a comment-to-DM path, define:

• the post or set of posts the rule applies to;
• the exact keyword or matching rule;
• whether repeated comments should trigger again;
• the public acknowledgement, if one is needed;
• the first DM; and
• the fallback when delivery is unavailable.

Choose a keyword that a real customer would type and that is unlikely to appear by accident. MENU is safer than a broad word such as yes.

The SmallForce Instagram comment-to-DM page shows the product surface for this workflow.

Write the response as an operating instruction

Use direct instructions rather than a vague tone request:

When a person comments “MENU” on the spring service post, reply: “Sent—check your DMs.” In the DM, send the approved spring menu link. Ask whether they are looking for residential or commercial service. Do not quote a custom price. If they ask about an emergency or a contract, assign the conversation to the service manager.

This gives the AI employee an allowed action, a prohibited action, and a handoff.

Keep the first message short. The person asked for one thing; deliver it before asking a qualification question.

Decide what counts as a lead

A comment is not automatically a lead. Define the evidence you need, such as:

• the person asks about availability;
• they provide a phone number or email;
• they choose a service and request a quote; or
• they ask to schedule.

Store only the information the person supplies and the business is allowed to retain. If a conversation does not meet the threshold, it can remain an ordinary inbox interaction.

Add human handoff

Route the conversation when:

• the person asks for a custom price or commitment;
• the AI lacks an approved answer;
• the customer is upset;
• the message contains sensitive information;
• a refund or cancellation is involved; or
• the person explicitly asks for someone on the team.

The handoff should include the original conversation and a short reason, not an invented summary of what the customer “must have meant.”

Test before publishing

Run at least these cases from a non-admin account:

1. Exact keyword with successful DM delivery.
2. Keyword with different capitalization.
3. Similar word that should not trigger.
4. Repeat comment from the same account.
5. Private account or delivery failure.
6. A question the approved source does not answer.
7. A request for a person.
8. A customer who declines to share contact details.

Check the public reply, private message, inbox record, captured fields, duplicate handling, and assignment. A fast response is not a success if it is irrelevant or creates two copies of the same conversation.

Measure the whole path

Track counts that correspond to real steps:

• qualifying comments received;
• trigger executions;
• public replies and DMs delivered;
• replies from recipients;
• leads that met the written definition;
• handoffs accepted by a person; and
• completed follow-up.

Keep a failed delivery separate from a person who received the message and chose not to reply. Those problems require different fixes.

Use the inbox, not a dead-end automation

SmallForce can bring supported DMs, comments, mentions, and replies into the same social workflow used for publishing and analytics. An AI employee can handle a routine first step while the full conversation remains available to the organization.

That matters because comment-to-DM is only the entry point. The customer may next ask for availability, send a photo, request a call, or need a person. The useful outcome is a coherent customer journey, not a large trigger count.

Read the SmallForce social inbox documentation for the broader workflow, or start with the Instagram comment-to-DM automation page.