AI Agent Setup and Configuration

AI Agent Setup and Configuration

Desk360: AI Agent Customer Assistant Setup and Configuration Guide

Introduction

The AI Agent developed by Desk360 aims to automate customer support processes, reduce the workload of representatives, and improve customer experience. This AI-powered system provides uninterrupted service even when customer representatives cannot provide live support, ensuring 24/7 accessibility.

AI Agent enables faster and more effective communication with expanding customer bases. This allows instant responses to customer demands while representatives can focus on more complex and critical issues. This both increases operational efficiency and enhances customer satisfaction.

AI Agent Activation Scenarios

There are two fundamental scenarios for AI Agent usage:

When representatives are offline: If a customer representative is not active in the system, AI Agent steps in to address customer questions and provide guidance.

Always active mode: AI Agent can be configured to respond in all situations, regardless of whether customer representatives are online or offline.

This flexibility provides organizations with an advantage in ensuring service continuity. AI Agent makes customer interaction seamless, making the support experience uninterrupted and fast.

Welcome Screen and User Interface

In the Desk360 platform, if a user has not yet created an AI Agent or if the existing AI Agent has been deleted, they encounter a special welcome screen featuring the ChatGPT logo. When they press the Start button, the steps for creating an AI Agent are presented clearly and understandably, directing them to the first step, the General tab.

After creating an AI Agent, users encounter a structure consisting of five main tabs in the user interface:

  • General: Section where basic settings are configured

  • Knowledge Base: Area where training data for AI is uploaded

  • Commands: Tab where commands are organized to customize responses

  • Agent Redirections: Settings for redirecting to customer representatives by defining keywords

  • Testing Area: Agent is tested and activated

Settings and content made in each tab are saved with the "Save and Continue" button, and users are automatically directed to the next tab.

Additionally, activity status control can be performed in the AI Agent interface; users can activate or deactivate the Agent at any time. The deletion process begins by clicking the trash icon and continues with a confirmation pop-up that appears on the screen. When the user confirms the deletion, the AI Agent is permanently removed from the system.

1) General Tab: AI Agent Basic Settings

The General tab is a critical configuration stage where the basic working principles and communication style of the AI Agent are determined. In this section, users are expected to make the following six basic settings:

WhatsApp Business Account Selection

The WhatsApp Business account through which the AI Agent will respond is selected here. If the user has multiple WhatsApp Business accounts defined, a selection field opens; if only one account exists, it is automatically assigned. When a new WhatsApp Business account is added, the selection screen becomes visible again.

AI Agent Responses

The condition for AI Agent responses is determined here. Two options are available:

  • Always: AI Agent always responds to customers

  • Only when operator is offline: Responds only when customer representatives are offline

The default setting is Always.

Conversation Tone (Tone of Voice)

The AI's communication tone is selected. Available options are Neutral, Friendly, and Formal. Selecting the appropriate tone according to user experience and brand identity is important. The default value is Neutral.

Emoji Usage

The preference for emoji usage in messages is determined here. By default, emoji usage is disabled. Users can activate this feature if they want to provide more intimate communication with customers.

GPT Model Selection

The AI Agent's basic language model is selected here. By default, the GPT-4o Mini model is assigned. Different models supported by Desk360 are presented in list form, and performance and features may vary according to the selected model.

Proper and conscious configuration of all these settings directly affects both the technical performance of the AI Agent and the quality of interaction with customers. Users ensure that selected preferences are saved by pressing the "Save and Continue" button after each completed setting and proceed to the next configuration step.

2) Knowledge Base Tab: Creating Information Database for AI Agent

The Knowledge Base tab creates the fundamental data source for the AI Agent to provide accurate, fast, and effective responses. In this section, users add various data types to train the AI and increase its knowledge capacity for customer demands.

All added information is processed by the AI Agent's algorithm and used in response generation. Providing accurate and diverse data sources enables the AI to better understand customer demands and produce quick solutions. Flexibility in uploading different data formats allows users to create knowledge bases specific to their sectors and needs.

Knowledge base creation steps and rules are as follows:

File Upload

Users can upload documents for the AI Agent to learn. Supported file formats include .pdf, .doc, .docx, .txt, .csv, .dot, .ppt. A maximum of 5 files can be uploaded per user with a 10 MB size limit per file.

Text Input

Users can also provide information to the AI by directly writing text. These texts are limited to a maximum of 1024 characters. Text input is ideal for providing additional information easily and quickly, such as explanations, brief instructions, or frequently used answers.

Website Addition

Websites can be added to the information base using two main methods: Crawl and individual website. When the Crawl method is selected, pages connected to the specified URL are listed under "Connected Links," and users can select or remove pages among these. A maximum of 20 different websites can be added. If there are more than 20 links, the "Add" button is automatically disabled.

Question-Answer Definition

Question and answer pairs are added for the AI Agent to provide correct responses to frequently asked questions. It is mandatory to enter both questions and answers in this field; if one of the fields is left empty, the user receives a "This field is required" warning. A maximum of 20 question-answer pairs can be added.

3) Commands Tab: Custom Response Creation and Optimization

The Commands tab is a critical step for the AI Agent to provide personalized and optimized responses in customer communication. In this section, users write their specific requirements briefly and clearly in the text input field on the left. Then, by pressing the "Generate" button, an optimized command generated by the AI model using the entered information is provided. The command displayed on the right can be edited by the user and customized according to needs.

The Commands tab directly affects the quality of the AI Agent's communication with customers. A well-configured command ensures the generation of accurate and purpose-appropriate responses. After making necessary adjustments, users save changes by clicking the "Save and Continue" button at the bottom of the page and are directed to the next step, the Agent Redirections tab.

4) Agent Redirections Tab: Trigger Settings and Redirection

The Agent Redirections tab is a configuration area that enables the AI Agent to make automatic redirections to customer representatives. In this section, users can define specific keywords so that when customers use these words, they are directly transferred to real representatives. For example, phrases like "discount, I want to buy" can be defined.

Additionally, when customer representatives are offline, the AI Customer Representative Continue Conversation feature provides informational messages to customers and continues communication seamlessly. This way, user experience is not negatively affected.

All settings are saved with the "Save and Continue" button, ensuring smooth transition to the next tab.

5) Testing Area Tab: Testing and Publishing

The Testing Area tab is the final stage where the AI Agent is tested and activated after completing its configuration. In this section, users experience the responses given by the AI Agent to customer questions, evaluating them in terms of accuracy, fluency, and consistency.

During testing, if users need to, they can return to previous tabs and make necessary adjustments; thus, the AI Agent is optimized.

This stage is created to ensure uninterrupted and quality customer experience, providing insight into the AI Agent's field performance.

When satisfied with the tests, users press the "Finish" button to create the AI Agent. When publishing is completed, the system directs users to the welcome screen and the AI Agent is opened for service in passive mode. Users need to switch the toggle to active position to activate it.

Only 1 AI Agent can be active at a time.

Last updated