Advanced - MAAX AI API Model Architecture

You don't need to grasp this section to get started. However, if you're a tech expert managing complex sales funnels, it's highly advisable to gain a thorough understanding of this content.

To craft a human-like response it is necessary that MAAX AI remembers past conversations with a specific prospect, as well information about that prospect. In addition, it has to understand what the conversation should achieve (goal): answer a question, persuade to buy a product, or convince someone to book a call with your team. To achieve that we had to develop a model architecture that can support all of those cases. Please take a look at the diagram below:

When you first register for a MAAX AI account you also register your company. Your company can now interact and manage multiple data sources and instances within the MAAX AI ecosystem:

  • Prospects: this model defines a human that MAAX AI will be interacting with. It has properties such as name, contact info, and free-from information that you can supply, and that is then used to personalize the messages to that prospect. Obviously, you can have as many prospects as you want.

  • Conversations: this model contains messages sent by the prospect and by MAAX AI. Internally we use this model to train MAAX AI to become better and better. There is no cap on the number of conversations you can have, and each prospect can be assigned to multiple conversations.

  • AIs: this model is pretty much defining how it will be acting and which AI instance should be used:

    • In case of sales, what product it will be referencing, what is the goal (a common goal our customers define is to get the prospect to book a call with a sales person), and which strategy should be used.

    • In case of customer support, the AI will be learning from your supplied content: video and pdfs.

Last updated