AI BUILDERS STACK
Case studyAI ChatbotE-commerce

How Loop Studio went from a 3-day support backlog to same-hour replies

A two-person design marketplace was losing customers to slow support. Three weeks later, a bot trained on their own docs was resolving 4 of 5 tickets — in their voice.

82%

tickets resolved without a human

3 days → 1 hr

median response time

3 weeks

from first call to production

+31%

repeat purchases, 90 days after

01 — The problem

Support was eating the founders alive

Loop Studio sells design templates to 30,000+ customers — with a team of two. Every morning started with 60–80 unread tickets: expired download links, license questions, refund requests. The same ten questions, endlessly.

Replies were taking three days. Refund disputes were climbing, reviews mentioned "no response from support," and the founders were spending afternoons answering email instead of shipping new products.

They didn't need more staff. They needed the repeat 80% off their plate.

INBOX · BEFORE

Download link expired3 days old
License for client work?2 days old
Refund request — wrong file2 days old
Invoice copy needed4 days old

+ 63 more unread

02 — The build

Three weeks, built in the open

No new app for customers, no new dashboard for the team. The bot lives in the site chat widget they already had, connected to the systems they already used.

W1

Train on what already exists

Ingested 214 help docs, 4,000 past tickets, and the refund policy. The bot learned Loop Studio's actual answers — and their tone.

W2

Give it hands, not just answers

Wired to Stripe and the license system: it re-issues download links, resends invoices, and checks order status itself. Refunds over $50 escalate to a human.

W3

Shadow week, then launch

The bot drafted replies while the founders reviewed every one. After 96% approval-as-written, it went live — with a one-click "talk to a human" escape hatch.

BUILD LOG — LOOPSTUDIO

> abs build --client loopstudio

Ingesting knowledge base…

✓︎ 214 docs + 4,000 tickets indexed

✓︎ Stripe + license API connected

✓︎ Escalation rules set: refunds > $50

✓︎ Shadow mode: 96% approved as written

Deployed to production. Day 21.

StackOpenAIn8nStripe APIIntercomSlack
03 — The result

The backlog disappeared in the first week

Within 7 days the queue was at zero for the first time in a year. Ninety days in, the bot handles 82% of tickets end-to-end, escalates cleanly, and the founders spend support time only on the conversations that actually need them.

Response time

3 days → under 1 hour

Founder hours on support

20 hrs/wk → 3

Refund disputes

down 64%

"We went from a 3-day support backlog to same-hour replies. The bot they built actually sounds like us — customers thank it by name."

Dana R., Founder, Loop Studio

Got a backlog of your own?

A 30-minute call, no pitch deck. We'll tell you honestly whether a system like this fits your business.