< All work

Crunch Fitness · AI sales agent

Put an AI agent in front of every lead, around the clock

How I took a conversational AI agent from business case to a signed agreement and a proof of concept built to be measured.

An always-on conversational agent for every prospect

Role
Director of Innovation
Scope
Franchised and corporate clubs, company-wide
Partners
Marketing, digital, operations, analytics, legal, finance, training
Timeline
Business case to proof of concept
Conversational AI engine connecting chat, voice and text to business outcomes
24/7
instant replies for every prospect, at any hour
Instant
first reply to every prospect
24/7
coverage for every prospect conversation
3
journeys: inbound, FAQ and outbound
3
channels: chat, voice and text, in stages
Context

A recurring-revenue fitness brand with a large franchised and corporate club network. A prospect's first conversation often decides whether they join, and front desks cannot answer every message at every hour.

The bet

A conversational agent grounded in each club's real information could answer prospects instantly, around the clock, and turn more of them into members, with people stepping in whenever they should.

What I did
  • Built the business case and defined success: lead-to-member conversion, response time, resolution rate and customer sentiment
  • Chose where an agent should start: three journeys covering inbound membership conversion, FAQ resolution and proactive outbound follow-up
  • Designed the agent around real club data, brand voice, guardrails and clean handoffs to people
  • Sequenced channels by risk and volume: web chat first, then voice, then text
  • Designed the measurement: a controlled test across clubs at different stages of maturity, so results reflect the agent and not the season
  • Aligned marketing, digital, operations, analytics, legal, finance and training on one plan, and secured the agreement covering the POC and long-term use
What changed
  • From idea to an agreed scope, a signed agreement and a proof of concept designed to be measured
  • A repeatable playbook of journeys, channel sequence, testing method and stakeholder plan for the next agents
  • A clear path to extend the agent beyond sales into retention and member service

How I designed the agent

Six decisions that make an AI agent safe to put in front of customers.

01
Grounded in real data

Answers come from each club's live information and approved knowledge, not from guesses.

02
Guardrails by design

Brand voice, approved topics and clear limits on what the agent can promise or do.

03
People in the loop

Clean handoffs to staff when the agent cannot resolve something, and between channels without losing context.

04
Staged channels

Web chat first, then voice, then text, ordered by integration risk and conversation volume.

05
Tested before launch

Simulations, transcript reviews, staff spot checks and a go or no-go gate before any customer sees it.

06
Measured fairly

Matched control clubs by maturity, with consent and compliance built into outbound messaging.

How it came together

01
Case
Business case and success criteria
02
Scope
Three journeys, channels launched in stages
03
Build
Grounded in club data, with guardrails
04
Prove
Controlled test across different club types

The takeaway

An AI agent earns trust the way a new hire does: clear scope, good training, a manager nearby, and results measured fairly.

Back to the start

7-Eleven · Gold Pass

A subscription that generated tens of millions in incremental revenue in its first year

Read the case study

Contact

Got a product bet worth making?