From 4-hour reply times to 2 minutes, at 13k SKU scale
84%
Orders auto-handled
2 min
Avg reply time
+38%
Conversion uplift
The challenge
Cosmy's WhatsApp inbox was drowning. Around 800 messages a day landed on a team that averaged a four-hour reply time, an eternity for an impulse-driven beauty buyer. By the time a rep replied, 38% of carts had already been abandoned.
The catalogue made it worse. With 13,684 SKUs, finding "that serum from the ad, but for oily skin" by hand was a slow scavenger hunt, and reps gave inconsistent answers depending on who picked up.
The brand was spending heavily on paid traffic to fill that inbox, then losing the buyers to nothing more than slow, manual replies.
The solution
A multi-channel AI sales agent now answers in seconds and knows all 13,684 SKUs cold, turning the inbox from a bottleneck into a sales engine:
Semantic search over the full catalogue, the entire product range is loaded via Product Feed into the Knowledge Base, so the Catalog Search Agent resolves "that serum from the ad, but for oily skin" instantly and accurately.
One agent across every channel, WhatsApp, Instagram and the website widget run the same agent, so reply time dropped from four hours to about two minutes everywhere a buyer might land.
Live inventory, no false promises, a connection to the 1C ERP gives real stock levels, so the agent never sells what isn't there.
Active cart recovery, the Cart Recovery Agent re-engages abandoned carts via Follow-ups, referencing the exact items, which directly attacks that 38% drop-off.
Leads routed into HubSpot, every qualified conversation is written to HubSpot via create_crm_lead with the chat link attached and the source tagged to the AI, so sales sees a clean pipeline.
Reply time fell from four hours to two minutes at 13k-SKU scale, and the paid traffic the brand was already buying finally converted instead of leaking away.
