Where the beauty brand started
The brand sold skincare and cosmetics across 12 physical locations and a busy Instagram presence. Instagram DMs were their single biggest inbound channel, and their biggest bottleneck. A two-person social team was drowning in "is this in stock?", "what's good for oily skin?", "can I order through DM?", and the replies came hours late, if at all.
They measured it: only a fraction of DMs ever got a same-day answer, and the rest quietly became lost sales. The team wasn't lazy, there were simply more conversations than two humans could hold.
What we built together
We deployed three MyChatBot agents on their Instagram channel. A Sales Agent answered product questions and guided purchases. A Catalog Search Agent, trained on their full range via the Product Feed in the Knowledge Base, resolved "something for sensitive skin under 500₴" instantly. And a Follow-up Agent re-engaged anyone who went quiet mid-conversation.
All three shared one instruction set built in the Configuration Wizard, so the brand voice stayed identical whether the customer asked about ingredients or shipping.
From pilot to full rollout in 4 weeks
Week one was setup and connecting the Instagram account. Weeks two and three were tuning, teaching the agent the brand's product language, the quirks of their categories, and exactly when to hand off. Week four was full rollout across all DM volume.
The biggest lever was hand-off discipline. Using Hand-off Control with a stop-word, the agent handled routine questions end to end but routed anything sensitive, reactions, complaints, bespoke gift orders, straight to the human team, who got a Telegram ping via Flight Control.
The result: 3.2× more DMs closed
Within the month, the brand was closing 3.2× more Instagram DM conversations into orders. Reply time collapsed from hours to seconds. The two-person team didn't shrink, instead they stopped doing data entry and started handling the high-value, high-touch conversations the agent routed to them.
Every closed conversation landed in their CRM via create_crm_deal with the chat link attached, so for the first time they could see exactly how much revenue Instagram DMs were actually producing.
What made it work
Three things. First, the agent answered from real catalogue data, not vibes. Second, it sounded like the brand, not like software. Third, it knew when to step aside. None of that is magic, it's configuration, and any brand running Instagram at volume can set the same thing up in MyChatBot. The playbook is repeatable; the 3.2× is just what happens when DMs stop going unanswered.