A fifty person clothing brand used to spend until spring clearing its Black Friday queue. An AI agent now handles more than half the messages, and the team keeps the hard ones.
Pepper sells intimates direct to consumer in the United States. It started as a crowdfunding campaign and now runs a fifty person business.
Support is where the brand is made or lost. Fit and sizing questions arrive before the purchase, so a slow answer is an abandoned cart rather than a complaint.
At peak the inbox took more than four hundred emails a day. The queue built through Black Friday and was still being cleared in the spring.
The old platform could not sort that queue by customer or by country. Everything arrived as one long line, so the team worked it in the order it came.
What was actually going wrong before anything changed, and why it stayed that way so long.
Most of the volume was not complicated. It was the same questions about fit, orders and returns, written by different people in different words.
The brand could not hire its way out of it. Expanding the outsourced contract for one holiday season would have cost at least twenty five thousand dollars, and it would have bought slower answers.
Meanwhile the valuable conversations sat in the same queue as the routine ones. A customer who needed sizing help waited behind a tracking request, and often gave up first.
Built inside the tools the company already paid for. Nothing exotic, put together in a different order.
An AI agent answers the questions it can answer, using what the brand already knows about fit, orders and returns. It has handled more than 27,000 conversations.
KlaviyoPre purchase questions get a recommendation instead of a link. The assistant turns a sizing question into a basket, which is where the 18 percent uplift comes from.
KlaviyoAnything personal goes to a person with the history attached. Customers who need more are offered a video fit session, and those sessions end in exchanges rather than refunds.
KlaviyoThe before column is where the company started. The after column is where it landed once every system was running.
| Metric | Before | After |
|---|---|---|
| Tickets answered without an agent | None | 54 percent, 60 at peak |
| First response time | Over a day | Minutes |
| Resolution time | Days | Hours |
| Holiday support contract | $25k more | Not needed |
We would have had to expand our BPO contract by at least $25K for the holiday season.
We are not threatening those really personal interactions that make our brand Pepper.
Every story leaves one. Reading it is cheaper than learning it the slow way.
The automation rate is not the interesting number. The interesting number is that average order value went up while the team got smaller jobs to do.
That happened because the agent was pointed at the pre purchase questions, not only at the complaints.
The line the brand drew is worth copying. The agent handles what repeats, and a person still takes anything that sounds like a person.
The points the story turns on, in the order they happened.
Lower operating costs, year one: 12%
Read the caseBillable hours recaptured weekly: 50 h
Read the caseThe auditoría de IA y operaciones finds it, prices it, then hands you a plan in two working days. Your fee comes off the first thing you build with us.
