💥 The 10× question

An agent executes quickly. If the plan is wrong, it executes the mistake quickly, too. That's why the first decision isn't "which tool," but what problem.

One question can help you find the weak spot without a spreadsheet or consultant: "if my business grew 10 times overnight, which process would break first?"

No one actually needs to grow 10 times. The question just exaggerates the change so the weak point becomes clear. Whatever would break at 10× is already creaking today, just quietly.

How to read the diagram: the thick blue arrows show the volume coming in; the dashed gray arrows show what can get through. All the extra flow gets stuck at the lit-up box. Improving "service" or "billing" won't help while the lit-up box is clogged.

When you ask the 10× question, a list of complaints appears. Not everything on it is a bottleneck. Some of it is just nuisance: tedious and repetitive, but it doesn’t block anything.

The test is to ask what happens if no one does that for a week. If the business loses a customer, money, or a deadline, breaks. If it only gets uglier or slower, it’s annoying.

What to look for in the table: the first three questions are about customers, money, and the queue. The last is about feelings. Feelings matter, but they don't pay for the agent.

Found two or three candidates? Now move beyond guesses. For each one, count three things during a typical week: time expense, money lost and errors that happened.

You don’t need a system. Put a piece of paper next to the computer and make a mark each time it happens. Add them up at the end of the week.

How to read the diagram: each bar’s length represents time spent. The lit bar is longer than the other three combined—and it also generates a penalty (bottom line). Marcos thought the problem was “posting more”; the numbers show that it’s the smallest bar.

The most common mistake isn’t technical. It’s choosing the wrong process. We tend to automate what is fun of automating or what appears — post, video, “replying bot” — and leave out what it costs, because it’s boring and nobody sees it.

And each connected agent is one more door to control (tracks 3 and 4). It’s not worth opening a door for a small gain.

How to read the diagram: The higher it is, the more the problem costs; the farther to the right, the more often it occurs. The lit-up box (high and to the left) is where money quietly leaks away. The red box is where almost everyone starts—and therefore gets disappointed.

It’s easy to dodge the question on your own. An AI chat helps you stay focused: it asks, you answer. One question at a time keep the conversation short and go straight to the bottleneck.

Here, the chat is just a assistant: it doesn’t connect to anything or change anything. You know the business, and you’re the one who answers.

Me entreviste sobre o meu negócio: <descreva em uma linha, ex.: salão de beleza com 3 cadeiras e 1 recepcionista>.

Comece por esta pergunta: se o meu negócio crescesse 10 vezes amanhã, que processo quebraria primeiro?

Regras da entrevista:
- Faça UMA pergunta por vez e espere a minha resposta.
- No máximo 6 perguntas.
- Para cada processo que eu citar, pergunte: quantas vezes por semana acontece, quanto tempo leva e o que acontece quando dá errado.
- Separe o que QUEBRA (perde cliente, dinheiro ou prazo) do que só INCOMODA.
- Não invente números: use só o que eu disser. Se eu não souber, escreva "estimar".

No fim, escreva:
1. o processo que quebra primeiro, em uma frase;
2. os números que eu dei (vezes por semana, tempo, prejuízo);
3. o que eu deveria contar na próxima semana para confirmar;
4. um processo que parece urgente mas só incomoda.

The temptation is to solve everything at once. Don’t. One agent at a time, in one process at a time, is the only way to know whether it worked and stayed under control.

The other candidates don’t disappear: they go on a waitlist. Once the first one is running and measured, you can come back to them.