AI clone or Decision Model: two ways to put your expertise into AI

Search for a way to put your expertise into AI and most results offer a clone: a chatbot built from your content that answers in your voice. It does one job. People who follow you get to ask you things at any hour.

There is a second job a clone is not built for. Your clients already work inside Claude, ChatGPT, or Cursor. When they ask their own assistant a question from your field, it answers without you. A Decision Model is what that assistant reads first.

What a clone is

Delphi, the best-known product of this kind, describes what it makes as “an interactive digital mind, built from a real person’s knowledge, content, and experience”. Visitors get “their Delphi, answering in their voice”, by text or by voice.

A clone fits when:

• you have an audience that wants to talk to you;
• you have published a lot: podcasts, posts, a book, recorded talks;
• the questions are about your ideas, and the conversation itself is the product.

What a Decision Model is

A Decision Model is built from decisions you have made in your work, with the reason for each. It records what you weigh first, where you stop, and what changes your usual answer. Nobody chats with it. Another person’s assistant reads it over MCP in the middle of that person’s own task.

The assistant then answers as itself. It says which of your decisions it relied on, and it does not speak as you.

The differences, side by side

• What it is built from. A clone: what you have said in public. A Decision Model: what you decided, and why.
• Who asks. A clone: a person who came to talk to you. A Decision Model: someone’s assistant, at work on a contract, a roadmap, or a price.
• Who speaks. A clone speaks in your voice. With a Decision Model the buyer’s assistant speaks, and credits your model.
• What you check before it goes out. For a Decision Model you read every decision before it leaves your side, then the finished model before it is published.
• Where you have no position. A model can list the areas where your record gives no grounds, and the assistant is told to say so there.
• Money. Some clone platforms let you charge subscribers today. On avva, building and publishing are free during the beta, and charging for access is not open yet.

Four questions before you choose

Do people want to talk to you, or to work to your standard?

A reader who admires your thinking wants the conversation. A client drafting a statement of work wants their draft checked against how you would scope it. The first is a clone. The second is a model.

Do you have content, or decisions?

A clone needs published material. A model needs records of calls you made: emails where you explained a choice, proposals you turned down, review notes. Many working experts have little of the first and years of the second.

How would you feel about a wrong answer in your voice?

Before you publish a clone, ask it something you have never addressed and read what comes back. An answer in your voice reads as your position. In a regulated field, such as law, medicine, or finance, that test matters more.

Who is on the other side?

A clone has its own page, and people come to it. A model sits where the work happens. It needs the other person to use an assistant that supports MCP, which Claude, ChatGPT, and Cursor do.

You can do both

The two do not compete for the same material or the same reader. A clone can hold your essays for your audience while a model holds your pricing calls for your clients’ assistants.

If the second job is the one you need, this guide walks through building a Decision Model. You can also open a published one and see what an assistant reads.