I use InstantDM for Instagram conversations and lead capture. The problem was that not every conversation needed to become a CRM contact.
Someone asking for a price probably belongs in the CRM.
Someone replying with a random question probably does not.
Someone requesting a free resource might be worth keeping in the conversation history, but I do not necessarily want them sitting inside my sales pipeline.
At first I was sending almost everything into the CRM. After a while I had a lot of contacts that were not really leads. It made the pipeline harder to read and made follow ups more confusing.
So I changed the workflow.
Now every new InstantDM conversation first comes into Make.com through a webhook.
Make.com pulls out the basic details I want to keep, such as the username, original message, keyword, source, and time of the conversation.
Instead of sending everything directly into the CRM, I log every conversation into Google Sheets first.
This gives me one simple place where I can see everything that came through InstantDM without filling the CRM with every single person who sent a message.
After that, I added a filter.
The filter checks whether the conversation actually looks like a lead.
For example, if someone asks about pricing, a service, booking a call, or something clearly related to buying, the workflow continues and creates or updates that person inside the CRM.
If it is just a general question or a casual reply, the conversation stays logged in the sheet and nothing gets added to the CRM.
That small change made the whole setup much cleaner.
I still have a complete history of the conversations if I need to go back and check something, but my CRM now contains mostly people I actually need to follow up with.
My current flow is basically:
InstantDM conversation → Make.com webhook → clean the data → log it → check if it is a real lead → send only qualified conversations to the CRM
I am still improving the qualification part because right now most of it is based on keywords and a few simple conditions.
The next thing I am considering is adding an AI classification step that reads the message and decides whether it is sales, support, general inquiry, or something else.
I am not completely convinced that I need it yet though. The simple filter is already doing most of the job.
Has anyone here built something similar?
I would be interested to know whether you keep every conversation outside your CRM first, or whether you send everything directly into the CRM and organize it afterward.