The offer in layers

An offer is more than the thing you deliver. Look at it in three layers.

• Core result: what the customer is really paying for. For an exam family it's a score high enough for the chosen university; for an adult, finally speaking up in meetings.
• Product: what delivers that result. At Klasna, a package of lessons with one tutor, say 16 maths lessons.
• Services around the product: matching a tutor, rescheduling, homework checks, a mock exam, the weekly progress note to parents.

Families buy the result and judge you by the layers around it. That's why two schools with equally good tutors can feel completely different: one sends a note every Friday, the other goes quiet until the exam.

One slip to avoid: what the family might choose instead (free videos with a chatbot, a school teacher) is not a layer of your offer. It belongs to the situation row of the plan, where the alternatives live.

Split every layer into table stakes and extras. Table stakes are what every serious alternative has: without them you're not even considered. Extras are a reason to choose you.

Two tests sort any part of the offer:

1. Would a family notice if it disappeared tomorrow?
2. Would anyone choose Klasna because of it?

Table stakes pass only the first test. A real extra passes both.

At Klasna in 2026, a qualified tutor who knows the national exam, video that works reliably from home and rescheduling a day ahead are table stakes: families would leave without them, but every school offers them. The weekly progress note is the extra. Tutors spend about fifteen minutes a week on each one, and parents mention it in almost every review.

The trap: leading your ads with table stakes. "Qualified tutors" is true, but so is it for everyone, so it gives nobody a reason to pick you, as you saw when positioning Klasna in module 4.

A part of the offer doesn't become an extra because it sits in the services layer or because it cost effort to build. Customers decide what counts, through the two tests: would they notice if it disappeared, and would they choose you because of it?

Klasna stores recordings of every lesson. It sounds generous, but hardly anyone watches them. Nobody would notice if they vanished, and nobody picks Klasna for them, so the recordings pass neither test. They are neither table stakes nor an extra: they're cost, in storage and in the time it takes to manage them.

What to do with such a part:

• keep it only if it's cheap, or a small group really needs it (a learner who missed a lesson through illness, say);
• otherwise drop it or stop promoting it, and put the money and effort into an extra families do value.

Before you add anything new, ask the same two tests in advance. "It would be nice to have" is not a reason anyone chooses a school.

"Let's add AI" is not a plan. Ask two questions of every AI idea before it goes near the offer:

1. Whose work does it take? The tutor's own part (the work with the learner: explaining, judging, correcting), or a chore around it (drafting, sorting, scheduling)?
2. Who gains? Does the family see more because of it, or do we only save cost?

Four Klasna ideas, answered:

• AI marks homework instead of the tutor: takes the tutor's part and mostly saves cost. The risky kind.
• AI drafts the weekly note from the lesson log, and the tutor edits and signs it: takes a chore, and parents get a fuller note on time.
• AI sorts rescheduling requests: takes a chore and only saves cost. Useful, but nobody chooses Klasna for it.
• AI marks extra practice between lessons, and the tutor goes over the mistakes next time: touches the tutor's work, but the learner gets more feedback and the tutor stays in the loop.

The questions don't ban any idea; they show what each one does to your difference.

The two questions about any AI idea (whose work it takes, who gains) lead to a rule of thumb: take chores so customers see more; be careful about taking the human core just to save cost.

Leo's idea is lessons with an AI tutor, much cheaper to run. That takes the tutor's part to save cost, and it puts Klasna on the same shelf as free videos and chatbots: the shelf the Club had to move away from in module 4, when families said they left it for a free chatbot. Cheaper to run, but with no reason left to pay.

The weekly note drafted by AI works the other way. It strengthens the extra families already value, frees tutor time and keeps the tutor's judgement in the note. One condition stays fixed: a person checks every note before a parent sees it, because Klasna is responsible for what parents read, not the tool.

Two other weak options: adding an AI chat button because competitors have one (no layer gets better), and saving time by sending personal data to any tool (a data-protection problem, not an offer).

An AI draft is only as good as what you feed it, and at a school what you feed it is often personal data: a learner's name, marks, notes on progress. Sending that to an outside tool is a decision with rules, not a shortcut.

Klasna's rule, in line with data-protection law: send personal data to an AI tool only with a lawful basis and an agreement with the provider, and only the data the task needs. If that isn't in place, the tutor removes names before pasting the log. For a weekly note, the draft needs what was covered and how the learner did, not who the learner is.

In practice:

• check whether Klasna has an agreement with the tool's provider; a free consumer chatbot usually comes without one;
• strip names and anything the note doesn't need;
• the tutor still checks the draft and signs it.

"It's only internal" doesn't make it safe: the data leaves Klasna the moment it's pasted. And asking a teen for consent in a chat doesn't fix it either: for a teen, it's the parent's say.

A chore taken by AI is worth something only if you know how much time it frees and what that time buys.

The calculation is simple: number of tasks a week × minutes saved per task ÷ 60.

At Klasna, about 600 exam learners get a weekly note in spring (from 430 in February up to 715 in May; 600 is a fair spring average). With an AI draft, the tutor's check takes 5 minutes instead of 15:

600 × (15 − 5) = 6,000 minutes a week; 6,000 ÷ 60 = 100 tutor hours a week.

Then decide what the hours buy. Good uses add something families notice: a monthly mock exam with a written score, or more lessons in the busy weeks before the exam. A bad use is cutting the human check to free even more, because the check is what keeps the note trustworthy.

Two slips: using the full 15 minutes instead of the 10 saved, and taking May's peak as if it lasted all spring.

The offer row of the plan now has its parts: the layers (result, product, services), what you keep as table stakes, the extra you add, and one AI change with its human check.

But an extra you plan to add is still a guess about what families value. Treat it as a hypothesis and test it cheaply before you build it, the same way you tested Klasna's position in module 4: ask a few spring families whether a monthly mock exam with a written score would matter to them, or offer it to one group first.

Two ways to get it wrong: building the mock exam for everyone at once just because 100 tutor hours are free, and banking the hours as a saving, which keeps the cost cut and adds nothing families see.

How it fits together: the layers show what families buy, the two tests sort table stakes from extras, and the two AI questions keep the difference intact. The next lesson asks how old each offer is: programming dipped all summer and the Club is three months old.