Introducing Spotter
Adebola Ali · 9 min readCoaching is not one job. It is programme design, nutrition, admin, sales, copywriting, scheduling and customer service, and most coaches do all of it alone, between sessions, on a laptop at nine in the evening. The coaching itself is the part you trained for. Almost everything around it is the part that eats the week.
Spotter is the assistant we are building into Backstage to take the mechanical, repetitive and frankly tedious parts of that list, and leave you the part that actually needs a coach. It lives in a panel beside whatever you are working on, it knows your library and your clients, and it drafts rather than decides.
The name is the promise. A spotter stands behind the bar while you press. They do not lift for you, they do not tell you what your programme should be, and they are not there to be impressive. They are there so you can push harder than you safely could alone, and they step in only when they are needed.
This was written before any of it was live, and it is mostly still written in the future tense. Some of it has since shipped: Spotter answers questions about your own data, and it drafts forms, splits, products, services and messages for you to approve. The rest is still ahead of us.
We published it early because we would rather be judged against a clear description than a vague promise, and because the coaches who tell us which parts matter most are the ones who decide what we build first.
Why a spotter, and not another dashboard
The instinct when a tool gets complicated is to add a dashboard. More charts, more filters, more tabs. The trouble is that a dashboard can only answer questions someone anticipated when they built it. The question you actually have at eleven on a Sunday night is never one of them.
An assistant inverts that. Instead of building a report for every possible question, you answer the question when it is asked. Instead of a form for every possible action, you describe what you want and confirm what it proposes. The interface stops being a map you have to learn and starts being a conversation about your own business.
That only works if it is grounded in your actual data rather than general knowledge about fitness. An assistant that can recite training principles is a search engine with better manners. One that knows your client has been complaining about their left shoulder for three weeks is something else entirely.
Building programmes
A training split is a structured thing: days, exercises, sets, reps, rest, progression. So is a nutrition plan: meals, foods, portions, and a set of macro targets they have to land on. Structured work is exactly what an assistant should be good at, and it is where a large share of a coach’s unpaid hours quietly goes.
The intent is that you describe the athlete and the constraint, and Spotter produces a real draft in your library, built from your own exercises and your own foods. Draft a four-day upper and lower split for an intermediate lifter with a bad left shoulder, avoiding overhead pressing. Build a 2,400 calorie day at 180g of protein from the foods already in my library, no dairy. Take this hypertrophy block and scale it to three days a week without losing chest volume. Swap every barbell movement for a dumbbell equivalent, for a client training at home.
The important constraint is that it drafts and you approve. A generated split arrives as a draft you can rewrite, not as something already sitting in a client’s app. You are still the coach, and a plan with your name on it should have passed through your hands.
Keeping up with clients
The hardest part of a growing roster is not writing programmes. It is holding twenty people in your head at once, noticing that the quiet one has gone quiet in a different way than usual, and remembering what someone mentioned about their knee six weeks ago.
Check-ins, measurements, progress photos, form submissions and message history already live in Backstage. Spotter will read across all of it and tell you what changed, so you walk into a call already knowing. Summarise the last eight weeks for this client: weight trend, adherence, what they have complained about, what they have been consistent with. Who has not checked in for two weeks. Which clients logged a drop in sleep or energy. Whose measurements have not moved in a month despite full adherence.
It will draft replies too, in your voice, acknowledging the travel week and adjusting the target. But drafting a reply is not sending one. Anything addressed to a client lands in the composer for you to read, change or delete, and nothing goes out without you pressing send.
Writing the words
A coaching business runs on a surprising amount of prose. Intake questions, package descriptions, welcome emails, form copy, the bio on your link in bio, the re-engagement message to clients who lapsed last quarter. It is all work that has to be done well and that almost nobody enjoys doing.
Write an intake form for online clients covering training history, injuries, diet, schedule and goals. Rewrite this package description so it speaks to postpartum clients rather than general population. Draft a weekly check-in form with questions that actually tell me something about adherence. Tighten this bio to under forty words without losing the credentials.
It should write in your language, not ours. The more of your own copy it can see, the closer a first draft lands to something you would have written yourself, which is the only standard that matters here.
Answering questions about your own data
Every coach eventually wants a number that no screen shows. Revenue from one package type last quarter. How many sessions ran in March and how many were no-shows. Which plan you assign most often. Average client lifetime in months. Everyone whose subscription renews in the next two weeks.
Building a report for each of those is impossible, because the list never ends. Answering the question when it is asked is not. This is the part we expect to be most useful and the part we intend to ship first.
Answers will come with their working shown: which records were counted, over what period, so you can check the number rather than trust it blindly. A confident wrong number is worse than no number at all, particularly when you are about to make a pricing decision on the back of it.
Doing the admin
Some requests are not questions, they are chores. Add chicken thigh with these macros to my food library. Duplicate this nutrition plan for the three clients I onboarded this week. Move Thursday’s seven o’clock session to Friday. Archive every package that has not sold in six months. Tag all of these meals as high protein.
Each of those takes a minute. Collectively they take your evening, and they are the reason the library never quite gets tidied and the archive never quite gets cleared.
Anything that changes your data will show you exactly what it is about to do and wait for a yes. Destructive actions, deleting or unassigning, will ask every single time, however many times you have already said yes. Convenience is not worth one silently deleted training block.
It knows where you are
Spotter opens in a panel beside your work rather than on top of it, and it can see the page you are on. If you are looking at a client, "summarise their last block" needs no name. If you are in the meal builder, "make this vegetarian" needs no explanation of what "this" refers to.
Add 20g of protein to this day without changing the calories. That sentence means nothing on its own and everything when you are stood in a nutrition plan with the day open in front of you. Context is most of what separates a spotter from a chatbot that happens to be in a sidebar.
What it will not do
A coaching tool that oversteps is worse than no tool at all. Some limits are deliberate, and they will not quietly loosen once the feature ships.
It will not give medical advice, diagnose an injury, or contradict something a doctor has told your client. It will tell you when a question needs a clinician rather than a coach. It will not message your clients on your behalf; it drafts, you send. It will not invent data, and when it does not know something it will say so rather than produce a plausible number. It will not change your library without showing you first.
And it will not replace your judgement on programming. It produces a starting point. The decision about what a specific human being should do with their body on Tuesday remains yours. A spotter does not take the bar off you.
Where its answers come from
General fitness knowledge is the easy part and the least useful. What makes Spotter worth opening is that it knows your library, your clients and your history, rather than the internet’s average opinion about training.
So it reads from your account first. Your exercises, your foods, your meals, your plans, your forms, your bookings, your check-ins. When it builds a nutrition plan it uses foods you have actually added. When it drafts a split it reaches for movements you already coach. When it cannot find something in your library it will say so and offer to create it, rather than quietly substituting something you do not use and would not have chosen.
Your data, and your clients’
Client data in a coaching platform is sensitive in a way that most software data is not. Injuries, weight, photographs, medical history, things people would not tell a close friend. That does not become less true because a feature is convenient.
Spotter will only ever see data from your own account, and one coach’s clients will never be visible to another. Your client data will not be used to train third-party models. Clients will not get an assistant that reads their coach’s private notes about them. You can see what it accessed in order to answer a question, and turning it off will mean off, with no background processing of your account.
One thing here has changed since this was written, and we said we would tell you rather than let you find it. Spotter now saves your conversations, so you can reopen one instead of losing it when you close the panel. That means your questions, and whatever client detail is in the answers, are stored in your account. Deleting a conversation deletes its messages with it, there is no archive, and nothing you ask Spotter is visible to another coach. The privacy policy says all of this in plainer terms.
If any of that ever has to change, it will be said plainly and in advance, not added to a policy page and left to be discovered.
What we are building first
Everything above is the shape of the whole thing. It will not arrive at once, and pretending otherwise would be the fastest way to lose your trust.
The order we expect to build in: questions about your own data first, because it is the most useful and the least risky. Then drafting, since a bad draft costs you nothing but a delete. Then the writing, then the chores, and last the actions that change your data, because those have to be right every single time rather than most of the time.
If you want a say in that order, tell us which part of this post you read twice. We would rather hear it now, while it is still cheap to change. You can tell us what you need any time.
Related posts
Ready to get started?
Marginal Improvements