I’ve been running a small Instagram account for a coaching business, and the DMs got out of hand around month three. People asking the same five questions over and over, comments asking “link please?” that I’d forget to reply to for two days, and me manually typing out the same responses at 11pm because I hadn’t gotten to them earlier. So I finally sat down and built something to take the repetitive stuff off my plate, and I figured I’d write up how it actually works since I spent way too long piecing this together from scattered tutorials.
The problem I was trying to solve
Not full automation — I didn’t want a bot pretending to be me for every conversation. I just wanted the boring 80% handled (comment triggers, FAQ-style DMs, sending someone a link when they comment a keyword) while I stayed in the loop for anything that actually needed a real answer.
The stack
Make.com is doing the orchestration — watching for triggers, routing data, handling the logic. For the actual Instagram side (comments, DMs, sending replies back through Instagram’s API without me writing raw API calls myself), I’m using InstantDM, which handles the messy Instagram-specific plumbing so Make.com doesn’t have to.
How the scenario is put together
- Trigger — A watch module on new Instagram comments and DMs, filtered so it only fires on my business account, not every notification under the sun.
- Router — Splits based on whether it’s a comment or a DM, since I wanted different logic for each. Comments with certain keywords (things like “price,” “link,” “info”) get treated differently than a comment that’s just someone dropping an emoji.
- AI step — For DMs specifically, the message text gets passed through an AI module that checks intent — is this a common question I’ve already answered fifty times, or something that genuinely needs my attention? If it matches a known pattern, it drafts a reply in my tone (I fed it a handful of my actual past responses so it doesn’t sound like a corporate chatbot). If it doesn’t match anything, it skips straight to a Slack notification so I know to jump in personally.
- Send step — This is where InstantDM comes in. Once Make.com decides on the response, InstantDM actually sends it back through Instagram — whether that’s a DM reply, a comment reply, or auto-DMing someone who commented a specific keyword.
- Logging — Every interaction gets logged to a Google Sheet so I can see what’s actually being asked and adjust the FAQ logic over time instead of guessing.
What surprised me
The keyword-comment-to-DM flow was the single biggest win. I used to lose people because I’d forget to reply to a comment with a link, and by the time I did, they’d moved on. Now if someone comments “info,” they get a DM within seconds. Conversion on that specific flow noticeably went up once the delay disappeared.
The AI intent-check step took the most tuning. My first version was way too eager to auto-reply and a couple of responses came off stiff and out of context. I ended up narrowing what it’s allowed to auto-answer to a shorter list of genuinely repetitive questions, and routing everything else to me. Less automation than I originally wanted, but it feels a lot more honest to how I actually want to run the account.
A couple of things I’d tell past-me before starting
- Don’t try to automate tone from day one. Start narrow, watch what people actually ask, and expand from there.
- Rate limits are real. Instagram will throttle you if you’re firing off messages too fast, so build in small delays rather than hammering the API the second a comment lands.
- Keep a human-review step somewhere in the flow, even if it’s just a Slack ping. It saved me at least twice from a reply going out that would’ve looked bad.
Happy to share the actual blueprint/export if anyone wants to poke at it or has questions about how the router logic is set up. Also curious if anyone’s tried something similar with Stories replies instead of just feed comments and DMs — that’s the next thing I want to tackle.