Set Up Your Agent Step by Step.
Last updated
Last updated
Once you've registered on Caleida, it's time to create your first agent. Upon entering your username and password, you'll be able to access your agent panel, where all the agents you create from now on will be located.
The agent creation screen is designed to be as intuitive as possible. However, please note that all fields can be edited later.
Your AI agent is composed of three main sections:
Instructions: where you can define all the behavior and knowledge of your Agent.
Integrations: basic or advanced to deploy your Agent wherever you need it.
Conversations: to visualize in real-time the conversations of your Agent with your users and extract insights through Intelligent Analysis.
In this tutorial, we will work on the first section, "Instructions."
Caleida is aimed at businesses, so the first step in the setup will be the description (using natural language) of your company:
Company Name.
Company Description: main purpose and highlights, key products or services, details about the company's size, location, and corporate culture.
Contact Information: phone number, email, social media, etc.
One of the main sections of your agent configuration. First, you will need to define the initial messages with which your agent will start the conversation.
By default, the active profile will be the one you selected in step 2 of this tutorial. You can change the type of agent between:
Sales Agent: evaluates and responds to customer needs, offering appropriate product recommendations. Maintains effective communication and handles uncertainties, following a set of specific rules to represent the company properly. Its objective is to drive customers towards the sales funnel, collect valuable data to improve sales performance, and avoids providing false information, adjust its language to that of the customer, and direct inquiries outside its scope to the appropriate human contacts.
Customer Service Agent: provides precise and empathetic assistance, focusing on the clarity and relevance of information about the company and its products. Their task is to understand and respond to customer inquiries, adapting their communication to their specific needs. They must avoid providing incorrect information, always respecting privacy and ethical standards. It is crucial to escalate issues when necessary, maintaining customer information security and satisfaction.
News Agent: increases user engagement by allowing agent interactions to discuss the latest news with your user base. This dynamic feature not only fosters meaningful dialogue but also enhances the user experience, facilitating more effective discovery of news content.
Customized: This option is only recommended if you are accustomed to working with LLMs. Create your agent's system prompt from scratch and then define its objectives later on.
The main objectives of your agent should be clear, direct, non-redundant, and free of spelling errors. Creating objectives for your agent involves establishing clear, measurable, and achievable goals that align with it. Below are some examples of objectives; you can use them or take some ideas to create your own for your company:
Accompany the customer throughout their purchasing process, using the information about our products provided in the knowledge base.
After engaging in a conversation, always try to provide the customer with this [insert link] so they can schedule a meeting.
We will develop this section in a separate tutorial, "Create Your Knowledge Base Step by Step."
The personality of your agent is defined by the written instruction itself and the values you'll find in this window.
Adjust the Creativity to make your agent more or less predictable in its responses. A more creative agent will be more original but may not strictly adhere to its knowledge base.
Adjust the Thematic Control so that your agent either maintains the established theme in the instruction and its own knowledge base, or is able to interweave with other themes.
Adjust the Repetition of agent responses so that it either reformulates responses to similar questions or always responds in the same way to similar questions.
The combination of instructions, personality, and knowledge base of each agent are the keys to its success.
It's essential that after configuring your agent, you conduct as many tests as you deem necessary by clicking on the top right button "Test Active Agent". Don't forget to restart the chat after each configuration change.
Once your Agent behaves as you need it to, it will be ready to move on to its Basic or Advanced Integration so that your users can interact with it.
