What we refuse to automate

A parent messages a piano teacher at 10:47pm.

"Sorry to bother you so late. Is Ryan actually improving? He says he hates practice now and I don't know if we should keep going."

There is a version of the future where an AI answers that message. It would be fast. It would be grammatical. It would probably be warm, in the way that software is warm. And it would be the worst possible thing that could happen to that teacher's business.

We are building an AI assistant for coaches. So it's worth being clear, early and in public, about the parts of the job we have no intention of touching.

The distinction that matters

Almost every conversation about AI in small businesses collapses two very different kinds of work into one word: "admin."

But there are two kinds of labour in a coaching business, and they behave completely differently.

Administrative labour is the work that has a correct answer. Sending the invoice. Chasing the invoice. Rescheduling Tuesday's 4pm. Working out who has three sessions left in their package. Reminding fourteen parents that the studio is closed on Monday. Nobody chose to be a coach in order to do this. It is pure friction, it scales badly, and it happens at 11pm because that's the only hour left.

Relational labour is the work where the answer depends on judgement, history and care. Whether Ryan is actually improving. Whether to waive the cancellation fee for the parent who has never once been late before. How to tell a fifteen-year-old that she has plateaued without her quitting.

Automating the first kind gives a coach their evenings back.

Automating the second kind destroys the only thing they're actually selling.

Four things we won't do

We won't answer "how is my child doing?"

This is the single most valuable message in a coaching business, and it is not a status request. A parent asking that question is asking whether their money, their Saturday mornings and their hopes are being well spent. The answer requires a coach to hold six months of context and make a judgement call about a specific child.

Zuno can surface everything the coach needs to answer it well — attendance, what was covered, what they noted after the last session. It can put that in front of them in ten seconds instead of ten minutes of scrolling. It will not write the answer.

We won't apologise on a coach's behalf.

When a session is cancelled at short notice, or a coach forgets something, or a student has a bad week, the repair is the relationship. An automated apology is not an apology. It's a notification wearing an apology's clothes, and parents can tell — they can always tell.

We won't decide who gets let off.

Every coach has a cancellation policy, and every coach breaks it sometimes, deliberately, for the right person. That discretion is not a bug in the policy. It's the mechanism by which a small business builds loyalty that a large one cannot. Software that enforces the policy uniformly is software that quietly strips a coach of their most useful goodwill tool.

Zuno will apply the policy by default, tell the coach what it's about to do, and get out of the way when they say no.

We won't grade, assess or give feedback on a student's work.

Not the performance, not the essay, not the technique, not the swing. Even where a model could produce something plausible. A parent is paying for a specific human's opinion. The moment that opinion is generated, the coach is selling something they can be undercut on by anyone with the same subscription.

Four things we automate without hesitation

The flip side is that we're aggressive everywhere else, and we think most tools in this space are far too timid about it.

Money. Invoices go out on time, reminders chase themselves, and a coach should be able to see who owes what without opening a spreadsheet. Chasing payment is the most relationship-damaging admin task in the business precisely because it makes the coach do it. Take it off them.

Scheduling. Rescheduling a single session in a group class is a small combinatorial problem that eats twenty minutes of human attention. It should take one sentence.

Recall. What did we cover last time? Who is on their last session? Whose package expires this month? Which student hasn't been seen in three weeks? A coach carries all of this in their head or loses it. Neither is acceptable.

Drafting. This is the interesting edge. Zuno will draft the message to fourteen parents about Monday's closure — that's administrative. Our principle throughout is the same: Zuno drafts, the coach approves, and nothing leaves in a coach's name that they haven't read. The draft saves the time. The approval preserves the voice.

The test we use

When we're unsure which side something falls on, we ask one question:

Would the coach be embarrassed if the parent found out a machine wrote it?

Nobody is embarrassed by an automated invoice reminder. Everyone would be embarrassed by an automated answer to "is my child improving?"

It's a crude test. It has been right every time so far.

Why this is a business argument, not a moral one

It would be easy to read this as squeamishness. It isn't.

An independent coach competes with larger schools, franchised centres and cheaper alternatives on exactly one axis: the relationship is with a specific person who knows the student. That is the entire moat. It is why parents pay a premium, why they stay for years, and why they refer their friends.

Automating relational labour doesn't make a coach more efficient. It converts a differentiated business into an undifferentiated one, and undifferentiated businesses compete on price. A coach who saves four hours a week and loses their referral rate has made a catastrophic trade and won't notice for six months.

So the design constraint is commercial. Take the admin. Leave the relationship. Make the coach more present with their students, not more absent behind a convincing proxy.

What this commits us to

Writing this down publicly means we can be held to it. That's the point.

There will be a version of this product where adding an auto-reply to parent messages increases some engagement metric. There will be a customer who asks for it. We are saying now, before that pressure exists, that we won't build it — and if we ever change our minds, we'll say why here rather than shipping it quietly.

If you coach independently and disagree — if there's something on the "won't automate" list that you'd genuinely hand over tomorrow — we want to hear it. That argument is more useful to us than agreement.

Clazuno is a business management platform for independent coaches and instructors. Zuno is the AI assistant at the centre of it. We're onboarding a founding cohort now.