About the working logic of thunk.Ai

I want to make sure I understand. Thunk.AI also has a text that determines the company’s policies. This text is created in our Google Driver as Drive\thunk.ai\policies\default_policies.txt.gdoc by default. Here we can state our policies in writing and in any way we want.
Then you train the AI assistant to understand these policies. AI is trained on our policies.

For example, when we add an expense receipt in expense templates, you have it checked whether the receipt complies with the expense policy.

Is this the way it works? Do I understand correctly?

If I understand correctly, my conclusion is this… AI can be trained individually, i.e. project-based, template-based and company-wide.

If so, then Ai becomes like a staff member working in our company, like an AUDIT staff member who controls everything.

So it’s like building a person from scratch.

If so, if I understand correctly, this opens up very new horizons.

I am waiting for your answer urgently because it is exciting.
If I don’t understand, please explain the working logic of Thunk.Ai.


Yes that is exactly right! You provide “training” via short and simple descriptions in natural language. Some of these are at the account level (eg: your company’s policies). Some of these are at the per-thunk (i.e. per project) level. And then your AI agent follows these instructions where appropriate, and uses its knowledge/intelligence to do what you asked it to do.
It is more than building an assistant person from scratch. Most of the assistant is pre-built. You are just providing direction where appropriate.

We need to write an article describing the concept and capabilities. Will try to do that this week.