AI 2.0 element

The element is designed to recognize and respond to customer questions using Artificial Intelligence (or neural network).

To use the element, drag it with the mouse from the menu on the right.

To edit the element, click the pencil button.

Set the AI parameters:

• Leave the block without waiting for a response — allows playing a message to the customer while waiting for a response, for example, through the Speech synthesis element.
• Identifier — the identification number assigned to each scenario element.
• Message — enter the variable {{stt_answer}}. This variable determines which question the AI is answering. The AI takes information about which question to answer from the Question-answer element.
• Context — the rules that guide the AI in finding the answer to the given question. The context length is limited by the allowable number of tokens. This example includes not only options for finding the answer to the customer's question but also how the AI will respond depending on the customer's answer - nuances that affect the AI's performance. You can enter any context containing the logic by which the AI will process information received from the customer.
• Type— choose the AI type. Each type has its own performance, quality, and cost indicators. Some AI types contain multiple models/versions.
  • OpenAI >>Model
    • gpt-3.5-turbo: the fastest, cheapest, and most accessible AI model. Maximum number of tokens - 4096.
    • gpt-3.5-turbo-16k: this model is more expensive and can contain much more context. Maximum number of tokens - 16000.
    • gpt-4-1106-preview: much smarter, contains newer knowledge, and is more expensive. Maximum number of tokens - 128000.
• Tokens — sequences of characters found in the text set. This is the unit in which text is measured in the context and the message. Each token includes approximately 2-3 characters. 100 tokens ~= 75 words.
• Temperature— available value from 0 to 1, recommended - 0.7.
  • 0 — gives the most likely answers and constructions, less creativity.
  • 1 — more creativity, less predictability.

To limit the waiting time, click the stopwatch button, check the Limit waiting time checkbox, and set the timer.

In the scenario, the AI element works in conjunction with the Question-answer block. For more details, see AI IVR Scenario.

Also see:

Call Scenario Elements