A home-goods store replaced the sales desk with an AI seller: one tone, no missed nights, upsells from a knowledge base. After 3 months the numbers no longer looked like a trial.
- conversion to purchase
- 15% → 23%
- instead of 15 minutes to reply
- 30 sec
- ROI in 3 months
- 312%
Starting point
An online store for home goods. Before the rollout:
- 5 sales reps.
- More than 1,000 chats a day.
- Conversion to purchase: 15%.
- Desk cost: ₽350,000 a month.
Both people and shifts hurt: missed messages, uneven tone, churn and onboarding. Overnight leads sat idle. Orders had errors.
How we rolled it out
Prep. We pulled scripts, built a product knowledge base, wired CRM, wrote scenarios.
Test. AI ran next to the reps. Replies were tuned on live chats.
Launch. Load went up in steps, quality was watched, extra skills were added.
After 3 months
Requests: average reply 15 minutes → 30 seconds. Up to 500 chats at once. 24/7, none missed.
Sales: conversion 15% → 23%. Average order +35%. Repeat purchases +45%. Satisfaction +28%.
Money: ₽280,000 a month off payroll. Revenue +40%. ROI 312% in a quarter. Payback 1.5 months.
How a chat looks
Customer: “I’m looking for a bedding set.”
The AI seller asks size, colour, fabric, offers matching items, handles “a bit pricey” and upsells a throw at 15% off.
Why it holds
- One reply standard and real catalogue knowledge.
- Chat stats and behaviour — not “the team’s gut feel”.
- Capacity grows without hiring. New SKUs and campaigns go into the knowledge base, not into onboarding a sixth rep.