AI for Gyms: Book More Classes, Win Back Lost Members

Your busiest hour for missed calls is probably 6:30 pm, right when the front desk person is out on the floor running a class. Someone phones to ask if there's room in tomorrow's 7 am. They get voicemail. They book somewhere else.

That gap is what most owners actually mean when they ask about AI for gyms. Not a robot coach. Something that picks up, books, reminds, and follows up while your team does the job you hired them for. Here's what the evidence supports, what it costs, and where it falls apart.

Canadian adoption is no longer an early adopter story

Statistics Canada's Q2 2026 analysis of AI use by businesses reports that 19.2% of Canadian businesses used AI to produce goods or deliver services in the preceding 12 months, three times the 6.1% recorded in the second quarter of 2024. Among adopters, virtual agents and chatbots were the third most common application at 28.2%, behind data analytics at 36.6% and text analytics at 34.5%.

Small operators aren't sitting this one out. Businesses with 1 to 4 employees reported 19.9% adoption in that same release. But 40.0% of all businesses said AI simply isn't relevant to what they do, and the leading barriers were cybersecurity and privacy concerns at 13.4% and cost at 10.6%. Both of those land hard on a business that holds health information and payment details.

Reminders cut no-shows, and here's the honest version of that claim

The cleanest evidence for automated reminders comes from healthcare, not fitness. In a randomized controlled trial published in the International Journal of Pediatrics, Lin and colleagues (2016) randomized 169 patients. The group that got a text message on top of the standard voice reminder had a 23.5% no-show rate against 38.1% for the voice reminder alone, a difference of 14.6 percentage points with p = 0.04.

Be careful how far you carry that. It's one small trial, in a US pediatric clinic, for medical appointments people had waited weeks to get. Skipping a Tuesday spin class is a much cheaper decision. Take the direction as well supported and the size of the effect as something you measure in your own studio, not something you copy from a paper.

The lapsed member problem, including the part nobody advertises

Class booking is the easy half. Win-back is where the money is, and where the research gets uncomfortable.

DellaVigna and Malmendier studied 7,752 members at three US health clubs over three years in Paying Not to Go to the Gym, published in the American Economic Review, volume 96, number 3, June 2006. Members on a flat monthly contract above $70 attended an average of 4.3 times a month, paying more than $17 per visit when a 10-visit pass would have cost $10 a visit. On average they gave up about $600 in savings over the life of the membership.

Then there's their fourth finding. On average, 2.31 full months passed between a member's last visit and the day they actually cancelled, worth $187 in membership fees paid for nothing, and that lag ran to at least four months for 20% of members.

Read that twice, because it cuts both ways. Part of your revenue is people who stopped coming and haven't got around to quitting. An AI that texts everyone who hasn't scanned in for three weeks will bring some of them back and will remind others that they meant to cancel. That isn't a reason to skip win-back. It's a reason to run it on one segment first, measure reactivations against cancellations, and decide with numbers instead of hope.

The version that tends to hold up is narrower than messaging the whole lapsed list. Go after people whose behaviour shows intent: a trial that ended without a conversion, someone who cancelled two bookings in a row, a member whose visit frequency halved but hasn't hit zero yet.

What to automate first

• After hours and overflow calls, so the 6:30 pm caller gets a real booking instead of a voicemail box.
• Waitlist backfill, where a cancellation fires an instant offer to the next few people and the first reply takes the spot.
• Class reminders with a one tap reschedule link, so a "can't make it" becomes a different booking instead of an empty bike.
• A quiet check in at 14 to 21 days of no attendance, before the habit is fully gone.
• Failed payment and renewal notices, which quietly churn members who never intended to leave.
• Post trial follow up inside 48 hours, while they still remember the class.

The rules you can't skip

Every one of those messages is a commercial electronic message under Canada's Anti-Spam Legislation. Per ISED's guidance on consent, express consent has "no time limit unless the recipient withdraws his or her consent," while implied consent from an existing business relationship "may be valid for up to 2 years, or just 6 months in the case of inquiries or applications." Your message needs your business name, a mailing address plus a phone number, email or website that stays valid for at least 60 days, and an unsubscribe mechanism actioned "within 10 business days or less and at no cost to the recipient."

Texting counts too. ISED's page on texting clients is blunt that offering a STOP option isn't enough, you have to honour it, and that you stay responsible for third parties sending on your behalf.

Then there's the data itself. The Office of the Privacy Commissioner's e-marketing guidance states that individuals must consent to having their electronic addresses collected and used for marketing, that PIPEDA prohibits address harvesting, and that it is your responsibility to confirm the company you work with abides by its provisions. For a studio in Burnaby or New Westminster, that translates simply: the number a member gave you for class confirmations is not automatic permission to send them a win-back offer. Ask separately, log the consent, keep the log.

What it actually costs

Model inference is the cheap part. As of August 2026, Anthropic's published pricing lists Claude Haiku 4.5 at $1 per million input tokens and $5 per million output tokens, and the same page works through an example of 10,000 support conversations averaging roughly 3,700 tokens each for about $37 total. That example is customer support, not gym bookings, so treat it as illustrative. It still tells you the direction: a single studio's message volume is not what breaks the budget.

The real costs sit elsewhere. Integration with your booking platform, telephony and SMS carrier fees, and a staff member reading transcripts every week for the first month. Anything that talks to members needs a human checking what it said.

Where this doesn't apply

The productivity case is weaker than the sales pitch. In Artificial intelligence adoption and productivity in Canadian firms, released by Statistics Canada in April 2026, Jiang Li and Huju Liu found AI adopters looked 16.8% more productive. That premium fell to 10.2% after controlling for pre-existing productivity, then to 5.1% and statistical insignificance once complementary capabilities were accounted for. Their conclusion is that there is no statistically significant direct association between AI adoption and productivity, and that firms already using data analytics, cloud computing and ICT training were far likelier to adopt in the first place.

That matches what happens on the ground. Automation pays when the system around it is already decent. If your attendance data is patchy, your win-back list will be wrong and you'll annoy paying members. If your booking software has no usable API, you'll spend more on glue code than on the AI. And if you run one location with 150 members and know every one of them by name, the relationship is the product, and a bot answering the phone takes away the thing they're paying for.

Sources

• Statistics Canada. "Analysis on artificial intelligence use by businesses in Canada, second quarter of 2026." Released June 11, 2026. https://www150.statcan.gc.ca/n1/pub/11-621-m/11-621-m2026010-eng.htm
• Li, Jiang, and Huju Liu. "Artificial intelligence adoption and productivity in Canadian firms." Statistics Canada, catalogue 36-28-0001, released April 22, 2026. https://www150.statcan.gc.ca/n1/pub/36-28-0001/2026004/article/00002-eng.htm
• DellaVigna, Stefano, and Ulrike Malmendier. "Paying Not to Go to the Gym." American Economic Review, volume 96, number 3, June 2006. https://eml.berkeley.edu/~ulrike/Papers/gym.pdf
• Lin, Chia-Lei, Nila Mistry, Jordana Boneh, Hong Li, and Rina Lazebnik. "Text Message Reminders Increase Appointment Adherence in a Pediatric Clinic: A Randomized Controlled Trial." International Journal of Pediatrics, 2016. https://pmc.ncbi.nlm.nih.gov/articles/PMC5227159/
• Innovation, Science and Economic Development Canada. "Getting consent to send email." Canada's Anti-Spam Legislation. https://ised-isde.canada.ca/site/canada-anti-spam-legislation/en/getting-consent-send-email
• Innovation, Science and Economic Development Canada. "Texting for good client relations." Canada's Anti-Spam Legislation. https://ised-isde.canada.ca/site/canada-anti-spam-legislation/en/texting-good-client-relations
• Office of the Privacy Commissioner of Canada. "Guidance for businesses doing e-marketing." https://www.priv.gc.ca/en/privacy-topics/privacy-laws-in-canada/the-personal-information-protection-and-electronic-documents-act-pipeda/r_o_p/canadas-anti-spam-legislation/casl-compliance-help-for-businesses/casl_guide/
• Anthropic. "Pricing." Claude platform documentation, accessed August 2026. https://platform.claude.com/docs/en/about-claude/pricing

If you run a studio or gym around Metro Vancouver and want to know which of these is worth wiring up first, Autana Solutions will work through it with you. Book a free call and we'll look at your real booking data and no-show pattern, then tell you straight whether automation is the right next move or whether your money is better spent somewhere else this quarter.