Ironleaf Fitness · concept demo
An AI team
Six AI assistants. One shared memory.
Simulated demo · Ironleaf Fitness is a made-up gym
This is the gym's AI office, running live. Six assistants, each looking after one part of the business, and one shared memory. Everything is a simulation, but every job they do is a real gym job.
The problem
Members fade out before they leave.
W1 W2 W3 W4 W5 W6 W7 W8 W9 W10
Visits a week
Nobody noticed
Cancelled
This is one member's visits. They went from four a week to zero over two months, and nobody noticed until they asked to cancel. Most cancellations look like this. The signs are in the data; nobody has time to watch it.
How it works
Data in. Better decisions out.
You approve
The data already exists in their systems: check-ins, bookings, reviews, WhatsApp. The AI team reads it every day, does the routine work, and brings the big decisions to the owner. Whatever works is saved to the Gym Memory, so next month starts smarter.
The team
Six assistants. One per part of the gym.
The badge on an assistant is the step of its playbook. Under the building, the Timeline shows what each assistant did in the last 15 minutes.
Each assistant has one job, like a person on the team. The small badge over an assistant is the step it has reached in its playbook, for example 3/5, and amber means it waits for a person. Under the building, the Timeline shows what every assistant did in the last 15 minutes, and Replay plays it back. Member Care watches for members drifting away, Marketing runs posts and reviews, Coaching fills classes, Facilities gets machines fixed, Sales follows up every trial visitor, and Operations builds the owner's numbers. They take the repetitive work off the staff.
A normal Monday
While the gym runs, the details get done.
07:00 Numbers ready 09:00 23 check-ins sent 11:00 Repair booked 14:00 Trials followed up 16:00 Waitlist filled 18:00 Video drafted 22:00 14 questions answered
Walk through the day. 7am: the owner's six numbers are ready. 9am: 23 members who stopped coming get a friendly check-in. 11am: a broken machine is booked for repair in quiet hours. 2pm: yesterday's free-trial visitors get an offer. 4pm: cancelled spots go to the waitlist. 6pm: tomorrow's video is drafted for approval. 10pm: WhatsApp questions answered while the gym is closed.
Talk to it
Ask any assistant what it's doing.
Click any assistant to see what it's working on. The panel shows the step of its playbook it has reached (here 3 of 5, waiting for the Club Manager), its recent activity over the last 15 minutes, and its tasks. Then ask it something and it answers from the Gym Memory with the notes it comes from, or press Start a playbook to give it a job, like you would a member of staff. The answer is a demo reply: a run started here lives only in this browser.
The Gym Memory
Ask your gym anything. It answers from its memory.
Everything the AI team learns is saved in the Gym Memory: members, classes, decisions, results. The owner can just ask: "What's working to keep members?" It reads the notes and answers with the real numbers and links to where each fact came from. When a manager leaves, the know-how stays.
Playbooks
A written playbook for every job.
One social video, idea to results
Plan the slot, the brief Script 3 openings, the script Test poll results Check 9 rules, pass or fail You approve name and time Film shot list, consent Edit the final cut Publish when and where Measure numbers, lessons
25 playbooks · every step saved · nothing big without a yes
Open the Playbooks See a real run: the deadlift reel
Every job the AI team does follows a written playbook, like a staff handbook. There are 25, grouped into packs. The clip opens the Playbooks page, picks a pack, opens one playbook, and ends on Make your own, where the owner can write a playbook of their own (kept only in this browser). The Compliance checker, in the library, has a worked example read live from the Gym Memory. On the right is the video pipeline: eight small roles, one after the other. At each step the AI saves its proof to the Gym Memory: the script, the test result, the checklist, who approved it and when. Nothing big goes out without the Marketing Lead's or the owner's yes. Open the Compliance checker in the live demo: in its worked example the first script failed two checks (step 7, in red): a calorie claim nobody could prove and a member in the background with no consent form. The second script passed all nine (step 9) and the Marketing Lead approved it (step 10). You can open the whole run and see every step.
You stay in charge
AI suggests. You decide.
Check-ins Answers Repairs Drafts
Prices Spending Staff Memberships
This is usually the owner's first worry, so say it plainly: the AI never changes a price, spends money or touches a membership on its own. It does the routine work and asks on the owner's phone for anything big, with the reason and the cost. One tap approves it and the assistant gets on with it. In the live demo, let the investor approve one himself.
The calendar
Everything it plans lands on your calendar.
Solid: booked settled, in the diaryDashed: waits for a yes the AI proposed it, you decideNobody outside the AI team is told until someone says yesClick any item who, why, and the note behind it
Open the Calendar What waits for a yes
The calendar stops at Friday 16 Oct, 22:00, the newest note in the Gym Memory; the AI Office and the Cameras run on from there on a demo clock. Everything the AI team plans, and everything already booked, is on one calendar: classes, repairs, posts, reviews. A solid item is settled. A dashed item is only a proposal, and nobody outside the AI team hears about it until someone says yes. In the clip, the treadmill service is dashed. Click it and it says who planned it and why: Saturday has the cable fitting. Press Ask on the Owner's phone, approve it there, and the item turns solid with a line in What changed; the button Proposals waiting for a yes counts down from 5 to 4. Click any item in the live demo to see the reason and the note behind it.
The Council
Big ideas get debated. You make the call.
For bigger questions, like opening 24 hours or spending on January ads, the assistants argue for and against and vote. The owner sees both sides and makes the final call.
Cameras
The AI watches the floor, not the faces.
No cameras in changing rooms, showers or toiletsFaces blurred , live and recordedCounts, not names : it counts, it never recognisesVideo kept 30 days , then deletedEvery view logged for 2 years
Simulated: the pictures are drawn from the demo's 3D office. Demo policy; check your local data-protection law before using real cameras.
Open the Cameras See a real run: the evening peak
Owners worry about cameras, so start with the rules. There are none in changing rooms, showers or toilets. Faces are blurred by default. The AI counts people and describes what happens, such as "41 people on the gym floor, above the comfortable 35"; it does not recognise anyone. Video is deleted after 30 days, and every view is logged. Under the wall, Friday 16 Oct in numbers sums the day up: the busiest hour (18:00, 41 people), the one crowding alert, the one request that waited for a yes, the cable machine still out. Be clear that this demo is simulated: the pictures come from the 3D office, and the descriptions come from the simulation. With real cameras a vision model would write them.
What changes
In the demo gym, after 3 months.
Cancellations a month 57 44
Free-trial visitors who join 34% 49%
Days from fault to fix (average) 9 3
Monthly income SAR 312k SAR 337k
Be honest: these come from the demo scenario. The point is where the value comes from: fewer cancellations, more trial visitors joining, machines fixed faster, and more income. On the machines, say it exactly: 3 days is the average since the QR codes went up on 25 August. A repair that waits for a part takes longer: cable machine 2 has been out since 3 October, waiting for its new cable. The pilot proves the real numbers.
Your numbers
What could it be worth to you?
Members kept a year –
Extra income a year –
Hand the owner the sliders and put in their real numbers. Members times the share who cancel each month, times 12, gives cancellations a year. Preventing a share of them keeps that many members, each paying their fee for a few more months. It's an estimate to start the conversation, not a promise.
How it starts
One club. 90 days. One scorecard.
Weeks 1–2 Connect Weeks 3–6 Members first Weeks 7–12 Full team Day 90 Review together
1 pilot club
1 champion on site
1 agreed scorecard
Start small and measurable: one club for 90 days. Weeks 1–2 connect the booking system, check-ins, WhatsApp and reviews and agree what the AI may do alone. Member Care goes first because it pays back fastest. Day 90: review cancellations, trial sign-ups, staff time and income, before and after. If the numbers move, roll it out to the other clubs together.
See it live
Let's watch the team at work.
Calendar Playbooks Cameras Runs in the Gym Memory
Switch to the live demo. Suggested path: the 3D office, click an assistant and ask it something, then open the Gym Memory and search "cancel". After that, show the Calendar, the Playbooks page and the Cameras wall.