ManyChat, replaced with software the brand owns.
- What shipped
- I built the brand its own Instagram messaging system to replace ManyChat. A keyword in a comment, DM or story reply gets an answer in seconds, and new leads get the current promotion. It runs in the same system as the promo calendar and team portal, on infrastructure the brand owns.
- Why
- Everything ManyChat did, the brand's own system could do, and manage better. Building more inside ManyChat meant building on someone else's platform, out of reach of the rest of the business.
- What it was worth
- It's part of the brand's own system now, not a third-party tool connected from outside and worked around. The funnel, promo calendar and portal run as one, so whatever automation comes next is written straight into that system and works with everything else from the start.
A funnel, rented
A membership brand I work with turns Instagram comments into customers. Someone comments a keyword under a post. A DM arrives with the guide they asked for, they leave an email on the way, and the conversation carries an offer to start the membership.
ManyChat ran that funnel, and ran it fine. But the brand was renting one of the most direct lines it has to a new customer, and that line ran outside the rest of the business.
What renting it cost
Not much in dollars: a few hundred a year. The cost was distance. ManyChat couldn't see the promo calendar, so the offer in a DM was whatever had last been typed into it. Matching it to a live sale meant opening each automation by hand, and again when the sale ended.
Every new automation was built by hand too, by copying the last one and changing a few values. And the contacts, the conversations and the rules all lived in ManyChat's account, on ManyChat's terms.
The same funnel, built around the business
I rebuilt it as software the brand owns. It talks to Instagram through Meta's official API, on an app Meta reviewed and approved, and it lives in the same system that already runs the team's portal and promo calendar.
Everything the old flows did came across: the keywords, the messages, the email question, follow-ups held for daytime hours, no second pitch inside two weeks. A member and a new lead get different conversations. The team reads every comment and DM in one inbox.
It knows what's on sale today
This is the part a rented tool couldn't do. The engine reads the same promo calendar that drives the homepage. Most days a new lead gets the standing welcome offer. When a promotion opens, the DM carries that promotion instead, with a link tagged so the sale traces back to it. When the promotion closes, the standing offer returns. Members aren't pitched at all.
Nobody edits anything. The calendar changes and the conversations follow on their own, so what a follower hears in a DM matches what the site says that day.
Automations written, not dragged
Each automation is now a short file. I can write one, run it against a simulated conversation and publish it without opening a builder, and more than 800 automated checks run before anything goes live. A list of fifty-five automations had been waiting to be built by hand, one per blog post. That list is now a script.
We did give up ManyChat's visual builder. What replaced it fits this brand better: each automation is drawn as a flowchart the team can follow, and changes are written and tested rather than dragged into place.
What the brand owns now
The brand now owns the funnel: the contact list, the conversation history, the rules, the code. Nothing bills by the contact, so growth doesn't raise the price. And it decides how the funnel is used, day to day and promotion to promotion, from the same place it plans everything else.
From the decision to live took under seven weeks, two of them waiting on Meta's review. The switch itself took an afternoon.
Software around the business
This one happens to be about ManyChat. The pattern is wider. Most businesses have a tool or two like it: it does its job, but it sits apart, and the business bends its plans to fit what the tool allows.
I'm not against rented software. If a tool is doing its job, it stays. The ones worth replacing are the ones that touch your customers and can't see the rest of your business. Those are the first thing I look for.
- Stack
- Instagram's official API, Convex, n8n, Klaviyo, a Next.js team portal. Built with Claude Code and Codex.
- Replaced
- ManyChat