Understanding the AI Agent in BoldDesk

The BoldDesk AI 2.0 Agent is an autonomous virtual support assistant that answers customer questions, uses configured knowledge, performs permitted actions through connected tools, and escalates unsupported requests to human agents.

Organizations can create specialized AI Agents for different support or business workflows. Each AI Agent can have its own instructions, Libraries, AI Tools, model, response settings, versions, and channel deployments.

The AI Agent and AI Assist are separate BoldDesk features:

AI Agent: Autonomously interacts with customers and performs configured support actions.
AI Assist: Helps human agents generate, summarize, translate, and improve content in the Agent Portal.

Role of the AI Agent in BoldDesk

The AI Agent can act as the first point of contact for customers. Depending on its configuration, it can:

Understand customer questions and conversation context.
Answer questions from linked knowledge Libraries.
Guide customers through documented procedures.
Collect information before creating or escalating a request.
Retrieve current information through connected tools.
Perform permitted BoldDesk or external-system actions.
Transfer unsupported or complex requests to human agents.
Provide consistent support across deployed channels.

The AI Agent’s responsibilities are controlled by its instructions, Libraries, tools, permissions, fallback behavior, and channel configuration.

Creating an AI Agent does not make it available to customers. The AI Agent must be configured, tested, published, and deployed to the required channel.

What the AI Agent Can Do

The AI Agent’s capabilities depend on the resources linked to its published version.

Capability	What the AI Agent Can Do	Required Configuration
Answer questions	Generate answers from Knowledge Base, Website, File, and Q&A Libraries.	Relevant and current Libraries
Guide customers	Provide documented procedures, policies, and troubleshooting steps.	Clear instructions and knowledge content
Create tickets	Collect required information and create a BoldDesk ticket.	BoldDesk MCP Server with the required write tool
Retrieve ticket details	Retrieve supported ticket information and conversation history.	Applicable MCP read tool
Update tickets	Perform supported updates such as status or assignment changes.	Applicable MCP write tool and permissions
Reply to tickets	Add a supported ticket response.	Applicable MCP write tool and permissions
Search support records	Search supported tickets, contacts, agents, and groups.	Applicable MCP read tools
Handle approvals	Perform approval operations supported by the connected tools.	Applicable MCP tool and permissions
Retrieve external data	Obtain current information from another business system.	Linked MCP tool or API Action
Perform external actions	Trigger a supported operation in an external application.	Linked and tested MCP tool or API Action
Escalate requests	Transfer unsupported or complex requests to a human agent.	Escalation instructions and channel workflow

Do not assume that the AI Agent can perform an action by default. Ticket operations and external-system actions require the applicable MCP tool or API Action to be configured and linked.

How the AI Agent Uses Knowledge

The AI Agent can answer questions using linked Libraries that contain:

Knowledge Base content.
Website content.
Uploaded files.
Custom Q&A content.

Responses depend on:

The customer’s question.
The available conversation context.
The linked Library content.
The selected AI model.
The AI Agent’s instructions and restrictions.

Example

Customer question
How do I reset my password?

AI Agent process

Identifies the password-reset request.
Searches the linked Libraries.
Retrieves the applicable procedure.
Generates a response that follows the configured instructions.

The AI Agent cannot reliably answer organization-specific questions when the required information is missing, outdated, inaccessible, or not linked to the AI Agent.

How the AI Agent Performs Actions

The AI Agent uses linked AI Tools when a request requires current information or an action in BoldDesk or another business system.

AI Tool	Purpose	Supported Use Cases	Requirements	Learn More
MCP Tools	Connect the AI Agent to operations exposed through a configured MCP Server.	Create tickets, retrieve ticket details and conversations, update ticket status or assignment, reply to ticket conversations, search supported contacts, agents, and groups, and perform supported approval operations.	The MCP Server must be configured and linked to the AI Agent. The required read or write tools must be enabled. Actions are limited by the authenticated connection’s permissions and accessible brands.	Overview of BoldDesk MCP Server
API Actions	Connect the AI Agent to a supported external API.	Retrieve current business information or perform an administrator-defined external action, such as checking an order status, retrieving a commitment timeline, scheduling an appointment, updating an external record, or triggering an external workflow.	The API Action must be configured, authenticated, tested, and linked to the AI Agent. The available operation depends on the connected API and its configuration.	Configure AI Tools for External Connections

Only AI Tools linked to the AI Agent are available during an interaction. Adding an MCP Server or API Action to BoldDesk does not automatically make it available to every AI Agent.
Tool actions are not built-in capabilities of every AI Agent. The required tool must be linked to the published AI Agent version, and the authenticated connection must have permission to perform the requested operation.

How the AI Agent Helps Human Agents

Reduces Repetitive Work

The AI Agent can handle routine and frequently asked questions that are covered by its knowledge sources, including:

Product setup and configuration.
Login and password-related procedures.
Business hours and availability.
Product features and information.
Company policies and guidelines.
Common troubleshooting and support requests.

By handling these repetitive inquiries, the AI Agent allows human agents to focus on complex issues, exceptions, escalations, and high-value customer interactions.

Collects Context Before Handoff

Before transferring a conversation to a human agent, the AI Agent can gather relevant information based on its configured instructions, such as:

The customer’s issue or request.
Account, order, or product details.
Troubleshooting steps already performed.
The reason additional human assistance is needed.

This helps ensure that agents receive the necessary context upfront, reducing the need for customers to repeat information and enabling faster issue resolution.

Supports Ticket Workflows

Depending on the configured workflows, permissions, and available tools, the AI Agent can assist with ticket management by:

Generating customer-facing responses.
Drafting replies for agent review and approval.
Adding supported public comments.
Adding private notes.
Retrieving and updating supported ticket information.

The AI Agent focuses on automating customer interactions and ticket-related tasks. For agent-facing productivity features such as content generation, response improvement, ticket summarization, and translation, use AI Assist.

How the AI Agent Works for Customers

Customers can interact with a published AI Agent through the channels where the AI Agent is deployed.

Customer Experience	How It Works
Ask AI in the Customer Portal	Customers and visitors can enter natural-language questions on the AI Answers page and receive responses from the AI Agent mapped to the selected Live Chat widget. For setup instructions, explore How to Configure the AI Answers Page in BoldDesk.
Instant and Guided Support	The AI Agent answers questions from linked Libraries, asks follow-up questions, and guides customers through supported procedures.
Current Information and Actions	When the required MCP Tool or API Action is linked, the AI Agent can retrieve current information or perform a supported action.
Human Handoff	The AI Agent follows its configured fallback or escalation instructions when a request is outside its scope, required information is unavailable, a tool fails, or human review is required.
Supported Channels	A published AI Agent can be deployed to configured channels such as Live Chat, Web Widget, WhatsApp, Facebook, Instagram, SMS, Telegram, and LINE.

The applicable AI Agent, Live Chat widget, Libraries, AI Tools, access settings, and customer-facing channels must be configured before customers can use the corresponding experience.

How the AI Agent Processes a Request

Step 1. Receive the Request

A customer sends a message through a channel to which the published AI Agent is deployed.

Step 2. Interpret the Request

The selected model processes the message together with the conversation context and AI Agent instructions.

Step 3. Select the Required Resource

The AI Agent determines whether the request requires:

An instruction-based response.
Knowledge from a Library.
An MCP tool.
An API Action.
A fallback or escalation.

Step 4. Retrieve Knowledge or Run a Tool

The AI Agent retrieves the applicable Library content or runs a permitted AI Tool.

Step 5. Generate the Result

The AI Agent produces a response using the retrieved information, tool result, response format, and configured instructions.

Step 6. Respond or Escalate

The AI Agent returns the result through the active channel or follows its fallback behaviour when the request cannot be completed.

What the AI Agent Needs

An AI Agent requires the following configuration and resources to answer customers, perform permitted actions, and operate through customer-facing channels.

Requirement	Purpose and Related Guidance
Defined Purpose	Defines the AI Agent’s intended users, responsibilities, supported requests, boundaries, and expected outcomes. Configure the purpose before adding knowledge or tools. Explore How to Create an AI 2.0 Agent in BoldDesk.
Instructions	Control the AI Agent’s role, response style, supported tasks, restrictions, tool-use conditions, escalation rules, and fallback behaviour. Learn How to Manage AI 2.0 Agent Instructions, Libraries, Tools, and Settings.
Libraries	Provide organisation-specific knowledge from Knowledge Base, Website, File, and Q&A sources. Linked content should be accurate, current, and accessible to the AI Agent. Explore How to Manage AI 2.0 Agent Instructions, Libraries, Tools, and Settings.
AI Tools	Enable the AI Agent to retrieve current information or perform supported actions through MCP Servers and API Actions. Only tools linked to the AI Agent are available during an interaction. Learn Overview of BoldDesk MCP Server.
Model and Response Settings	Determine how the AI Agent processes instructions, customer requests, conversation context, knowledge, and tool results. Explore How to Manage AI 2.0 Agent Instructions, Libraries, Tools, and Settings.
Permissions	Control who can create and manage AI Agents. Users responsible for AI Agent administration require AI Agent Builder Access: Full Access. Tool actions remain limited by the authenticated connection’s permissions and accessible brands. Learn How to Configure AI Agent Builder Access Permission.
Testing	Verifies expected answers, missing-knowledge behaviour, restrictions, fallback responses, tool selection, and tool results before publication. Use Playground to test representative customer requests. Learn How to Set Up, Configure, and Optimise AI 2.0 in BoldDesk.
Published Version	Makes a tested AI Agent version available for deployment. Publishing does not automatically expose the AI Agent to customers. Explore How to Create an AI 2.0 Agent in BoldDesk.
Channel Deployment	Connects the published AI Agent to configured customer-facing channels such as Live Chat, Web Widget, WhatsApp, Facebook, Instagram, SMS, Telegram, and LINE. Learn How to Assign and Deploy AI 2.0 Agents Across BoldDesk Channels.
AI Credits	Allow the AI Agent to process customer requests and generate AI responses. Credit usage varies according to request complexity, processing requirements, tool usage, and response detail. If no credits are available, the AI Agent cannot generate AI responses. Explore Understanding AI Credits in BoldDesk.
Monitoring	Allows administrators to review conversations and traces, identify missing knowledge or unexpected tool behaviour, and improve the AI Agent through a new draft and published version. Learn How to Review AI 2.0 Agent Conversations in BoldDesk.

Creating an AI Agent is only the first step. The AI Agent must be configured, tested, published, and deployed before customers can use it.
Publishing and channel deployment are separate actions. A published AI Agent is not available to customers until it is assigned to a configured customer-facing channel.

Recommended Implementation Flow

Define the AI Agent’s purpose, audience, responsibilities, restrictions, and escalation rules.
Create the AI Agent from AI Center > AI Agents.
Configure instructions, Libraries, AI Tools, model, and response settings.
Test supported, unsupported, and restricted requests in Playground.
Publish the validated version.
Deploy it to the required channels.
Test a complete customer interaction.
Review conversations and traces.
Correct instructions, knowledge, or tool behaviour in a new draft.
Test and publish the revised version.

Frequently Asked Questions

Is the BoldDesk AI Agent the same as AI Assist?
No. The AI Agent autonomously interacts with customers and performs configured actions. AI Assist helps human agents create, summarise, translate, and improve support content.
Can the AI Agent answer any question?
No. Responses depend on the instructions, linked Libraries, conversation context, model, and available tools.
Can the AI Agent create or update tickets?
Yes, when the required BoldDesk MCP Server and applicable read or write tools are linked and enabled. The connection must also have permission to perform the operation.
Can the AI Agent retrieve information from another business system?
Yes, when an administrator configures and links the applicable MCP tool or API Action.
Does publishing make the AI Agent available to customers?
No. After publication, the AI Agent must be deployed to the required customer-facing channels.

Was this article useful?

Dislike

Comments (0)