How a DTC sleep supplement brand implemented a 24/7 AI retention agent in weeks

MoonBrew sells sleep treats formulated without melatonin, built instead on magnesium, L-theanine, and a proprietary herb blend. Subscriptions drive the business. Arjun Shroff runs CX and fulfillment with a team of about 20 agents, and retention lands squarely on his desk.

“With CACs going up pretty high, driving retention is a huge growth lever. It gives us a lot of flexibility to continue scaling on the acquisition side if we can get our retention numbers up.”

A canceling customer had two paths. The portal had a cancel button, a preset list of reasons, and an offer wired to each one. “Most people were usually just clicking through it.”

The other path was email. Agents had an SOP to ask why the customer was leaving, but most customers skipped the question, and the team canceled the subscription rather than add friction.

Most cancellations went through unchallenged.

Arjun wanted to move retention to live chat. A customer who's in a chat window will actually talk about why they're leaving, where an email thread drags across days and usually dies. Staffing that queue around the clock was the problem. His math said at least 4 to 5 agents, more in peak season.

MoonBrew already ran Gorgias AI on product and order questions. He kept it away from cancellations.

Hoop started in sandbox mode, writing its replies as private notes on hundreds of real cancellation tickets while MoonBrew's agents worked them. The team read the two side by side and sent feedback on anything off.

Then they pointed it at their own employee subscriptions before any customer touched it, and watched it cancel what should be canceled and apply the discounts it was supposed to apply.

The pilot after that stayed deliberately small. A few tickets at a time, in a limited window of hours, widening as the team got comfortable. The first weekend it ran with no agents on the queue, Arjun checked in from home a few times to make sure nothing was on fire.

“Customers were saying thank you so much for the help, sending smiling emojis. Not only is it doing the job we're asking it to, it's doing it pretty well.”

The implementation work sat on Hoop's side, including a few requests that weren't on the platform yet.

The save rate reached 10%

The save rate is 10% now, with offer tests still running.

24/7 coverage, one agent's cost

The cost math was the easy part of the decision. Staffing live chat around the clock meant hiring 4 to 5 agents.

Response quality held the bar

Quality was the part he was watching.

“I do pride myself in having a great team of agents. I was really impressed to see that Hoop was able to not only maintain that level, but provide even better responses in certain situations than our agents would have.”

Subscribers regularly list two or three cancellation reasons in a single message, the kind of ticket that separates a tenured agent from a new hire.

“Being able to capture all three of those intents, figure out which ones will have the highest success rate, and frame the response addressing all three concerns: those things take time. Our most tenured agents are good at it, but for a newer agent that's a pretty big learning curve. We were able to solve that pretty quickly with the AI.”

Customers can't tell

The team watches for “please transfer me to a human” comments, which show up on their other ticket types all the time.

Offer changes in minutes, not weeks

Policy changes used to mean updating the SOP and waiting while 20 agents re-learned the rules a ticket at a time.

“With the AI agent, it's like turning on a switch. Instead of offering the customer 20%, we want to offer them 25%. It's almost instantaneous. Being able to quickly test different things is super beneficial.”

Arjun wants to double the save rate over the next 12 months, and bring Hoop from live chat into email, with SMS and phone after that.

“Meeting the customer where they are is a huge part of our strategy. It's better to start testing, test small, and continue to expand. It's a huge lever to pull to not only cut costs, but provide an overall better experience.”