AI agents use instructions, knowledge, tools, tasks, memory, models, and guardrails to handle user requests. Configure these sections in the agent editor, then preview and publish the agent when it is ready.
To use AI agents effectively, first understand the main building blocks.
Knowledge
Knowledge refers to the data, rules, and information an AI agent uses to understand and respond to various tasks.
Within UnifyApps, this knowledge can be pre-loaded or dynamically gathered in real time from sources such as databases, APIs, or user inputs. This may include an organization’s internal knowledge articles, external data, CRM data, websites, and more.
Insurance AI Agent can be fed with policy details, premiums, and coverage options, enabling it to provide quick, accurate answers and improve customer support.
The broader and more precise the knowledge, the more effectively the AI agent can perform.
In addition, UnifyApps supports Role-Based Access Control (RBAC) filters within its knowledge structure. By integrating RBAC filters, AI agents can dynamically filter out restricted information, ensuring that the knowledge provided is relevant and appropriate based on the user’s role.
In the agent editor, use Add Knowledge Set to attach an existing knowledge set, or Add Knowledge to add documents, websites, applications, or data catalogs directly. Added sources are indexed so the agent can retrieve relevant information when it responds.
Refer
To explore more about how to add your organization’s Knowledge base to your AI Agents, refer to this article.
Tasks
Tasks are separate journeys or intents an agent can carry out. Each task includes an objective, step-by-step instructions, allowed tools, optional conditions, and optionally its own execution model. At runtime, the agent detects which task fits the user's request and runs it.
Refer
Refer to this article to know more about Topics and how to configure it.
Prompts
Prompts in AI agents are sets of instructions or commands that guide the agent.
The accuracy and clarity of the prompts are essential, as they directly influence the agent’s ability to deliver precise and relevant results.
By using well-crafted prompts in AI agent, users can give clear instructions that ensure smooth task execution. High-quality prompts enable the AI agent to produce precise outcomes, enhancing the performance of business processes.
Refer
Discover more about how to add a prompt and the best practices to write instructions to your agent in this article.
Guardrails
Guardrails are a feature in UnifyApps' AI agent that apply to the agent's inputs and outputs. They include PII Masking, Hallucinations, Content Filters, Custom Word Filter, Denied Topics, and Blocked Messaging.
The Guardrails group is permission-gated and saves automatically as you configure it.
Why Are Guardrails Important?
Guardrails are essential for ensuring that AI agents:
Provide accurate and reliable responses.
Avoid harmful or inappropriate actions.
Respect privacy and data security regulations.
Stay aligned with company policies and ethical standards.
Refer
Learn more about how UnifyApps supports the Guardrail feature in this article.
Prerequisite Tasks
Prerequisite actions run before the agent starts handling a request. They are part of the Settings group, and this group may be permission-gated.
Refer
Refer to this article for configuring Prerequisite Tasks for your AI Agents.