AI & Intelligence for Combat Sports & Fitness — The Practitioner’s Playbook.

SectionThe honest reframe most AI agencies won't tell you

Generic "AI marketing" agencies are selling commercial gyms, boxing clubs, BJJ academies and CrossFit boxes a ChatGPT-blog-post package. Twelve posts a month on "ten benefits of HIIT," lightly rewritten by a model the agency does not understand, published unedited under the head coach's name, and then they wonder why the gym's organic traffic flatlines, why the ASA flags the transformation imagery, and why members keep churning at month four.

That is not AI strategy. That is AI cosplay.

The high-leverage AI use cases in combat sports and fitness are not blog automation. They are: form-correction video analysis that gives a member technique feedback on their boxing combination, BJJ guard pass or CrossFit clean before the next session; member-churn prediction from attendance and payment data so the front desk knows who is two weeks from cancelling and rings them before they do; an intro-offer-conversion classifier that tells you which six-week-trial members will convert to full membership and which need a coach intervention; sales-call summarisation so the membership team's enquiries land in the CRM as structured records, not voicemail notes; and ASA-compliant review of every AI-generated transformation post before it publishes.

Each of those is measurable. Each saves coach hours, lifts conversion or cuts churn. None of them is "more blog posts." Read the playbook. Run it yourself, or have us ship it on retainer.

SectionThe eight-point audit we run on day one

Score your own AI stack red / amber / green this week. Three or more reds means the foundation is broken — fix that before any new tooling spend.

1. Form-correction video analysis pipeline — A member uploads a 15-second clip of their jab-cross-hook, their armbar entry, their kettlebell swing, or their downward dog. The pipeline returns frame-level technique feedback against a coach-defined rubric: hand position, hip rotation, knee tracking, breathing cadence. The output goes back to the member inside 24 hours with a coach signoff line. Removes 20–40 minutes of in-person correction per member per week, and the asynchronous loop turns one head coach into a multiplier.
2. Member-churn prediction model — Attendance frequency, payment status, last-class-attended date, intro-offer cohort, class-type mix and direct-debit health all go into a churn-risk score that updates nightly. The front desk sees a ranked daily list of members at amber and red risk. A coach rings them inside 48 hours. Saves 30–50% of avoidable cancellations on a typical 600-member commercial gym, which on £85 average monthly fees is real five-figure retained revenue per year.
3. Intro-offer-conversion classifier — Every six-week-trial or 30-day-pass member gets scored on conversion probability after their second visit, using attendance pattern, class diversity, coach interactions and booking lead time. Below a threshold, the member gets a coach call. Above a threshold, they get the standard upgrade path. Front-of-funnel intelligence that lifts trial-to-member conversion by 15–25%.
4. Sales-call and enquiry summarisation — Every inbound enquiry — phone, web form, walk-in, DM — transcribed where applicable, summarised into a structured CRM record (goal, prior experience, injury history, schedule constraint, objection, next step), and lead-scored on intent. Coach time goes to A-leads first. Marketing reads the actual objection patterns at the end of every week, not at the end of every quarter.
5. ASA-compliant AI-generated transformation content review — No unedited AI on the public site or social feeds for transformation imagery, weight-loss claims, or "before and after" content. Every AI-drafted caption, blog post or social asset goes through a review gate against the ASA's CAP Code, particularly the rules on weight-control claims, atypical results, and substantiation. Misleading transformation content is the single most enforced area in fitness advertising and the ASA publishes ruled-against installers monthly.
6. Coach-reviewed AI-drafted content workflow — Every AI-drafted post on technique, programming, nutrition or recovery goes to a qualified coach (REPs Level 3, CIMSPA, England Boxing or equivalent) for review and sign-off before it publishes. The coach's name on the byline. Author schema points at the coach. Member trust on technical content lives or dies on whether the named author is qualified — and one wrong cueing instruction in a published article on the deadlift is a member-injury exposure.
7. Data hygiene on member health data — Member intake forms, PAR-Q responses, injury notes, body-composition scans and progress photos are special-category health data under UK GDPR Article 9. Lawful basis logged, explicit consent captured, retention windows defined, opt-out path documented, sub-processor list maintained, and no member data shipped to a model provider that retains training rights. Boring, mandatory, and where most gym AI projects quietly fall over on the first ICO query.
8. Productionisation with explicit coach fallback paths — Every AI tool has a documented behaviour for when the model is wrong, slow or unreachable. The form-correction pipeline tells the member "low-confidence output, your coach will review this in person Tuesday." The churn classifier defaults to "human review needed" on edge cases. The conversion classifier surfaces its reasoning. AI without a coach fallback is a liability — and in an industry where one bad cueing instruction can hurt a member, the fallback is non-negotiable.

Three or more reds — fix the foundation before commissioning new tooling spend.

SectionSix productised deliverables we ship per cycle

On a Foundation, Compound or Architect retainer, the same six outputs land in your portal each cycle. Industry-tuned, fixed scope, dated, owned by you.

Form-correction video analysis. A member uploads a 15-second clip via your portal — jab combination, BJJ shrimp, CrossFit clean, yoga warrior pose. The pipeline runs an off-the-shelf vision and pose-estimation model, scores the movement against a coach-defined rubric, and returns frame-level feedback with timestamped notes. A qualified coach reviews and signs off the output before it goes back to the member. Built on public LLM and vision providers (OpenAI, Anthropic, Google) with explicit fallback to in-person correction on low-confidence frames. Time to first signal: 21–30 days. Owned by you, exported as written SOP plus the orchestration scripts.

Member-churn prediction. A nightly model run over attendance, payment, class-mix and tenure data, scoring every member 1–5 on churn risk and writing the score back into your gym-management system. The front desk sees a ranked daily call list — amber risk gets an automated email, red risk gets a coach phone call inside 48 hours. Drives a 30–50% reduction in avoidable cancellations on a typical 600-member commercial gym, and the model retrains monthly as your data grows. Time to first signal: 14–21 days.

Intro-offer-conversion classifier. Every six-week-trial, ten-class pass or 30-day intro member gets scored on conversion probability after visit two, with the score updating after each session. Below a threshold, the system triggers a coach call with a structured talking-points brief. Above a threshold, the member gets the standard upgrade path. Lifts trial-to-member conversion by 15–25% by the simple act of catching wobbling trial members before they ghost.

Sales-call summarisation. Every inbound enquiry — phone, web form, walk-in, DM — captured, transcribed where applicable, summarised into a structured CRM record (goal, prior experience, injury history, schedule constraint, top objection, next step, confidence), and lead-scored 1–5 on intent. The CRM updates the lead automatically; the coach sees the call summary before they ring back. Drives a 15–25% lift in close rate by routing coach time to high-intent enquiries first.

ASA-compliant transformation content review. A documented review gate for every AI-drafted transformation post, weight-loss claim, body-composition graphic or before-and-after asset before it publishes. The gate checks against the ASA CAP Code rules on weight-control claims, atypical results disclaimers, substantiation requirements, and the "in conjunction with a calorie-controlled diet" qualifier. Reviewer signs off in writing; sign-off is logged with the published asset for audit. Cuts ASA exposure to near-zero on a category that is the single most enforced area in fitness advertising.

Coach-reviewed AI-drafted content workflow. A documented draft-to-publish workflow: AI drafts a long-form post on technique, programming, nutrition or recovery; a qualified coach (REPs Level 3, CIMSPA, England Boxing or equivalent) reviews and edits; the coach's byline goes on the published post; author schema points at the coach. Throughput rises 3–5x versus all-human drafting; technical accuracy stays at the level a CIMSPA assessor or England Boxing coach educator would sign off. Never unedited AI on member-facing technique content.

SectionWhat to do this week

Three actions, ranked by leverage. Same first three steps we ship in week one of a Foundation retainer for a commercial gym, boxing club or BJJ academy.

1. Pull last quarter's cancellation list and look for the warning signs. Owner: founder or membership manager. Time: one afternoon. Pull every member who cancelled in the last 90 days. Look at their attendance pattern in the four weeks before cancellation. How many showed a clear signal — visits dropping from 3x a week to 1x, then to nothing — that a coach call could have caught? If the pattern is obvious in retrospect on more than half of cancellations, member-churn prediction is your single highest-leverage AI deployment.
2. Audit one of your last ten transformation posts against the ASA CAP Code. Owner: head coach or marketing manager. Time: 30 minutes. Pick a recent before-and-after, weight-loss case study or body-composition post. Read the CAP Code rule 13 on weight control. Does the post carry the required atypical-results disclaimer? Is the weight-loss claim substantiated? Is "in conjunction with a calorie-controlled diet" present where required? If the answer to any is no, your transformation content is shipping non-compliant and you have an ASA exposure waiting for a complaint.
3. Decide DIY, DWY or DFY for the next 90 days. Owner: founder. Time: 30-min discovery call. We will confirm the right way in writing within two business days. See the three ways.

SectionFive questions combat-sports and fitness operators ask us about AI

How accurate is form-correction video analysis — can it really replace a coach's eye? It cannot replace a coach's eye, and we do not deploy it as a replacement. It is a multiplier. Public pose-estimation models with a coach-defined rubric typically land at 80–90% agreement with an in-person coach assessment on the basics — hand position on a jab, hip rotation on a hook, knee tracking on a squat, breathing cadence on a yoga flow. The pipeline catches the obvious technique faults; the coach catches the subtle ones. The deployment pattern that works is asynchronous: member uploads a clip between sessions, AI returns frame-level feedback, coach signs off the feedback, member reads it before their next class. The head coach reviews a sample weekly. One head coach goes from correcting 30 members in person per week to multiplying that across 200.

What kind of lift can we expect from member-churn prediction? On a commercial gym with 600 active members, a 6-month tenure and an average £85 monthly fee, a 30–50% cut to avoidable cancellations is in the £25,000–£40,000 retained-revenue-per-year range. The mechanism is simple: members who churn almost always show a clear attendance signal in the four weeks before they cancel, and most gyms catch zero of those signals because the front desk is busy and the data lives across two or three systems. A nightly model run, a ranked call list, and a coach making three calls a day catches the majority of the warning signs. The lift is proportional to your member base size and your current cancellation rate; payback on the build typically lands inside two months.

What does the ASA actually require on AI-generated transformation content? The ASA's CAP Code rule 13 on weight control and rule 12 on medicines and treatments are explicit. Weight-loss claims must be substantiated with credible evidence; atypical results must be disclaimed clearly; the "in conjunction with a calorie-controlled diet and exercise programme" qualifier is required where the marketing implies weight loss from the gym alone. Before-and-after imagery must be representative of typical results, with disclaimer if not. The ASA does not care whether the content was drafted by AI or by a human — the rules apply equally. The reason AI raises the stakes is throughput: a gym shipping 30 AI-drafted social posts a month is shipping 30 chances to breach. The fix is a documented review gate against the CAP Code before any transformation asset goes live, with sign-off logged.

Is qualified-coach review on AI-drafted content really necessary, or is it a CYA exercise? It is necessary for two distinct reasons. The first is member safety: a wrong cueing instruction in a published article on the deadlift, the round kick or the kettlebell swing can hurt a member who reads it and trains accordingly. REPs, CIMSPA and England Boxing all hold qualified coaches accountable for technical content carrying their name. The second is brand: members joining a serious commercial gym, BJJ academy or boxing club are reading the technical pages with care — they are choosing a coach as much as a facility. One paragraph that contradicts the head coach's actual cueing destroys the trust the rest of the site is building. Qualified-coach review on AI drafts is not a CYA box-tick — it is the production gate that lets you ship 3–5x more content without ever shipping unsafe content.

Can we run this ourselves with the playbook plus £750 audit? Yes. Most of the audit-and-fix list above is achievable in-house if you have a marketing manager plus a developer half-week per cycle plus a qualified coach who can spend an hour a week reviewing AI drafts and signing off form-correction outputs. The £750 audit gets you a written red / amber / green of all eight points, a prioritised next-step list with named owners and dates, and a copy of the workflow templates and prompt patterns we use ourselves. If you sign for DWY or DFY within 30 days, the audit fee credits against the first cycle.

SectionWhere to go from here

If you want this shipped end-to-end on a productised retainer, book a 30-minute discovery call. Tailored proposal in writing within two business days.

If you would rather have a senior practitioner reviewing your team's AI deployments each week, the coaching plans start at £750/month with rolling cycles and walk-away rights. If you have a hard deadline — a January-window AI pilot, a new-class launch, an intake-cohort surge after a brand campaign, or a fight-camp prep build — the two-week embedded sprint lands a senior practitioner inside your tools for ten working days at £3,000 fixed.

Or run it yourself. Read this playbook end to end, run the eight-point audit, ship one deliverable a month for six months. Twice-quarterly office hours are open to anyone using the playbooks — bring your work, get reviewed, no charge.