Case Study · Customer Service Agent
Fitness Studio Customer Service Agent: One Inbox Voice Across Every Studio
A multi-location Lagree studio group was answering member email from several hands, in several tones, quoting fees from memory. This is the reference build behind the Customer Service Agent I sell: what it looked like before, and what changed.
The starting point
The inbox was the brand, and nobody owned it
Member email is where a studio makes its hardest calls: a cancellation during a contract, a fee someone wants waived, an injury in class. At this studio group those replies were written by whoever had a free minute. The words were kind, and they were inconsistent. Two members with the same problem could get two different answers, and in a multi-location group they talk to each other.
The second problem was memory. Fee amounts and policy windows lived in people's heads and in a few old documents. When a policy changed, the inbox found out last.
The goal was never to remove the human. It was to give the studio one voice, one source for fees and policies, and a clear path for the emails that should never be answered by a machine at all.
The results
Before and after
Every staffer answered in their own tone, so a refund email read differently depending on who opened the inbox.
One voice profile built from real sent mail, so replies sound like the studio in every category.
Late cancel and no-show amounts were quoted from memory, and they varied by location.
One settings file holds every fee and policy. Change it once and every reply updates.
Hard emails sat in the inbox until someone decided who should handle them.
Injuries, billing disputes, and legal threats route by name to the person who owns them.
Promo language landed in the wrong conversations, including complaints and cancellations.
Upsell rules block any offer to an angry, cancelling, or injured member.
Nothing stopped a well-meaning reply from making a medical or refund promise.
A never-say list keeps those phrases out of every draft.
No shared record of what the studio told a member last month.
Every draft is reviewed by a human before sending, and the decision is on file.
The build
Four steps, four to six weeks
Read the inbox
Months of real sent mail, sorted by category: cancellations, injuries, billing, class changes, membership questions. The voice comes out of what the studio already writes, not from a style guide.
Write the responses
A drafted reply for each category, reviewed line by line with the owner. This is where the studio decides what it actually wants to say when a member gets hurt or asks for money back.
Lock the rules
Fees, policies, escalation names, upsell limits, and the never-say list go into one settings file as placeholders. No amount is ever hardcoded into a template.
Run it with a human first
The agent drafts, a person approves. Automatic sending only comes after the drafts are consistently right.
What to take from this
Four lessons that transfer to any inbox
Voice is an extraction problem
The brand voice was already in the sent folder. The work was pulling it out and holding it steady, not inventing something new.
One settings file beats twenty templates
Fees change. Policies change. Anything written into the templates becomes a hunt later.
Rules matter more than fluency
The model writes well on day one. What makes it safe to use is the escalation map, the upsell limits, and the never-say list.
Human in the loop is not a phase to skip
Approval runs first so the owner sees the edge cases before anything sends on its own.
Want this running on your inbox?
Bring two weeks of sent mail and your current policies. We start with the categories that cost you the most time, and nothing sends without your approval.