Getting started with AI Agents

Support teams spend a significant amount of time answering repetitive questions such as:

Hiver's AI Agents in Chat use generative AI to autonomously handle customer conversations. Unlike traditional rule-based chatbots that follow rigid scripts, AI Agents can understand intent, ask follow-up questions, and respond naturally, all while staying grounded in your trusted knowledge sources.

This can help your team:


How AI Agents in Chat work

Your AI Agent relies on the information you provide through knowledge sources such as help articles, websites, PDFs, and internal documentation. Using generative AI, AI Agents autonomously guide customers through conversations rather than just responding with static answers.

For example, if a customer says “I’m not receiving Slack notifications”, the AI Agent can:

This helps customers get answers faster while reducing repetitive support work for your team.


How AI Agents stay accurate and safe

Hiver's AI Agents are designed with built-in safeguards to ensure reliable customer interactions:


Prerequisites

Before you begin, make sure you have:

Note: AI Agents in Chat is currently a part of an early access program. To enable this in your account, reach out to us at support@hiverhq.com


Steps to set up AI Agents

To get started, go to Admin Panel → Shared Inboxes → Chat Inbox → AI Agents, and click Create your first AI Agent.

Alternatively, you can also go to Admin Panel → Hiver AI → Chat Agents, and click Create AI Agent.


Step 1: Train your AI Agent

Your AI Agent is only as helpful as the information it can access. You add information about:


Supported knowledge formats

Knowledge Base

Your published help articles can also be used as knowledge sources.

By default, your connected Hiver Knowledge Base appears under Added Sources but remains disabled. Toggle it on to include it as a knowledge source for your AI Agent.

Website

Hiver AI can learn from pages published on your website.

To add your website:

1. Click Website

2. Paste your website link. Enable Crawl links if you want Hiver AI to automatically scan linked pages such as:

3. Click Add.

Note: Hiver AI might take a while to go through the information depending on the volume of pages available. Once synced, your website will appear under the Added sources section.

Documents

You can upload text-based files (PDF, docx., txt files) or text-based files containing:

Note: You can select up to 5 files at a time. Ensure that each file does not exceed 30MB.

To upload a document:

1. Click Documents.

2. Select a file and click Upload.

Once you’re done adding knowledge sources, use the preview on the right to test how your AI Agent responds to customer questions based on what it has been trained on so far. Click Start Conversation to initiate a chat.

Try asking a few sample questions as customers would.

Review the answers returned and tweak your questions or your knowledge sources to improve accuracy. Once done, click Next.


Step 2: Customize your AI Agent

Customize how your AI Agent looks and communicates with customers. This section includes:

1. Identity

Add a name

Choose a name that matches your brand.

Examples:

Upload an avatar

Add an image that represents your brand or support team.

Once you're done, click Save and proceed.

2. Conversation style

Welcome message

Add a welcome message that appears when customers start a chat.

For example:

'Hey there!

I’m Plenix AI, and I’m here to help you today.'

To include another message, type it out, and click + Add

To remove a message, click the delete icon.

Tone

Choose how your AI Agent should sound. You can select between:

Instructions for your AI Agent

Use custom instructions to guide how your AI Agent should respond. These instructions act as guardrails to ensure your AI Agent operates within your support policies and maintains your brand standards. For example, you can instruct the AI Agent to:

You can test these changes in the preview.

Once done, click Save and proceed.


Step 3: Set up hand-off rules

AI Agents can autonomously manage repetitive and informational queries while handing more sensitive or complex conversations to your team. Hand-off rules help you define where that transition should happen.

For example, you may want to escalate conversations involving:

When a hand-off rule is triggered, the conversation is added to the Unassigned view along with the full conversation transcript.

There are a few hand-off rules added by default. Enable the ones you want to use.

To add a custom hand-off rule:

1. Click + New

Describe the situations where the AI Agent should escalate the conversation. For example,

2. Click Optimise. Hiver AI will rephrase the hand-off rule and optimise for accuracy.

3. Click Save and enable to start using the hand-off rule.

4. Once done, click Proceed.


Step 4: Deploy AI Agent

Once your AI Agent is ready, choose which Chat Inbox you want to deploy it in.

Under Deployment channels, select an inbox.

You can also refer to the checklist given to verify if you’ve configured required details including a welcome message and knowledge sources. When you’re ready, click Deploy.

Once deployed, the AI Agent will begin handling conversations from your chat widget on your website.

You can access your AI Agent from Hiver AI → Chat Agents or from your Chat Inbox’s settings, and make changes as and when needed.

Note: To disable your AI Agent, open it, go to Deploy and click Pause.


Viewing AI-managed conversations

You can access conversations handled by Chat Agents from the Assigned to Bot view in your Chat Inbox. Every conversation has a full transcript available.

You can also monitor chat as when they’re happening and assign it to a teammate to let them take over the conversation.

Reviewing AI-managed conversations regularly can help you:


Tips for training your AI Agent effectively

Here are some best practices for improving response quality and customer experience.

1. Cover your most common support workflows

Think about the questions your support team answers repeatedly every day.

For example:

Your goal is to ensure your AI Agent has enough information to guide customers through these workflows confidently.

2. Write knowledge sources the way customers ask questions

Customers often describe issues differently than how internal teams do. For example: