This guide helps you understand how AI fits into customer service in Crisp, and where to start if you want practical results quickly.

Crisp AI works best as a layered system. Some features reduce repetitive questions before a conversation starts, others let Hugo automate answers or actions, and others help your team work faster inside the inbox. In the current product, that setup is centered around Hugo and the AI Agent menus rather than a separate data hub.

How AI fits into Crisp

The goal is not to turn everything on at once. A better rollout is usually to train Hugo well, validate its behavior, and then add the AI layers that make sense for your workflow.

The main AI layers to know:

• Hugo → the customer-facing AI Agent that can answer, route, escalate, and use integrations when configured
• Search Chatbox AI → the self-service search layer in the chat widget that can reduce simple incoming questions before they become conversations
• AI Tools in the composer → agent-facing actions such as Predict, Rephrase, Fix grammar, Adjust tone, and Expand
• Copilot → an agent-facing assistant that helps your team ask internal questions and draft better replies using the same core resources as Hugo
• Conversation intelligence → summaries, topic detection, and voice transcription that help agents understand conversations faster

Start with Hugo

If you want the strongest AI foundation first, start from Crisp and open AI Agent from the left menu. That is where you configure Hugo, train it, test it, and activate it.

Configure identity and behavior

In AI Agent → Agent → Settings, define how Hugo should present itself and what rules should guide it. This is where you set the agent name, avatar, business description, source usage, escalation behavior, and office-hours awareness.

Clear configuration matters because Hugo does better when your business context and escalation expectations are explicit from the start.

Useful guides:

• Main setup guide → Getting started with Hugo AI Agent
• Instructions → How to prompt Hugo AI Agent for Instructions
• Routing → How does Hugo Routing work

Train Hugo on the right resources

In AI Agent → Train, give Hugo the content it should rely on. The best results usually come from combining website pages, knowledge base articles, files, and short Q&A snippets that answer recurring questions clearly.

The dedicated training article is especially useful here, because good training content makes a bigger difference than adding more prompts.

What usually works best first:

• Website content → product pages, docs, onboarding pages, pricing explanations, and policy pages
• Knowledge base articles → the support answers you already want customers to find on their own
• Q&A snippets → short internal answers for edge cases, policy clarifications, or recurring phrasing
• Files → TXT, CSV, or PDF resources when the useful content is not already published on the web

Useful guides:

• Training overview → How to train Hugo AI Agent on your data
• Hugo setup → Getting started with Hugo AI Agent

Test before activation

Use AI Agent → Evaluate → Playground to test Hugo like a customer would. This is the safest place to pressure-test your setup before you go live.

Once the answers feel reliable, move to AI Agent → Agent → Activation and decide whether Hugo should answer new conversations by default, work only on selected channels, or route to other logic first.

Add integrations if you want real actions

Hugo is not limited to answering questions from documentation. When you connect integrations or MCP servers, it can also retrieve live data or perform actions within the scope you define.

That is what unlocks use cases such as checking an order status, looking up a subscription, or retrieving account information without immediate human handoff.

Useful guides:

• Native Hugo integrations → How to use Hugo Integrations
• Custom MCP → How to build MCP integrations with Hugo

Add AI across the customer journey

Once Hugo is properly configured, you can extend AI to other parts of the experience for both customers and agents.

Search Chatbox AI

Search Chatbox AI helps visitors look for answers before they open a conversation. This is one of the fastest ways to reduce repetitive support volume when your content is already in good shape.

You can extend the scenario by adding buttons, fallback messages, escalations, or conditions if you want more control.

Before testing your AI Chatbot

Test the AI chatbot on a live website where the Crisp chatbox is deployed. The bot will not behave the same way in the chatbot testing environment.

If you do not want the bot to trigger for all visitors yet, use a very specific keyword as the entry trigger while you validate the setup.

How to make the AI answer your common questions

The AI learns from the content available in your Data Hub. A simple way to start is to add:

• Answer Snippets with short question-and-answer pairs for recurring questions
• Helpdesk articles so the AI can reuse your existing support documentation

Start small with your most common questions, then enrich the data over time as you identify gaps.

How the AI Chatbot really works?

11 AI chatbot best practices for always-on customer service

• Keep training resources clean → bad or outdated content creates weaker AI answers
• Add frequent searches → guide visitors toward the questions they already ask most
• Use categories when needed → especially helpful if you support multiple products or topics

Useful guides:

• Setup guide → How do I setup Crisp Search Chatbox AI on my chatbox widget?
• AI overview → Crisp AI

Hugo for automated conversations

Hugo is the layer that answers users directly, routes them to the right place, or hands the conversation back to humans when needed. That makes it the main customer-facing AI feature in Crisp.

If you use workflows, the current way to hand a conversation to AI is the Answer with Hugo block. That block replaces older legacy approaches based on MagicReply or legacy AI search blocks.

Useful guides:

• Workflow handoff → How to integrate Hugo in your Workflows
• Workflow basics → Getting started with Workflows

AI Tools in the editor

Inside conversations, agents can use AI Tools directly in the composer to improve or accelerate a reply before sending it. This is useful when the answer should still be reviewed by a human, but the drafting work can be faster.

The most useful actions to know:

• Predict → generate a draft reply that the agent can review, edit, or send
• Rephrase → rewrite the current message more clearly without changing its meaning
• Fix grammar → clean up spelling and grammar quickly
• Adjust tone → make a message friendlier, more formal, or more direct depending on the context
• Expand → turn a brief answer into a fuller explanation that is easier for customers to understand

Copilot for agent-facing help

Copilot is the internal assistant your team can open from the inbox sidebar. It is meant for agents, not customers, and it helps them ask questions, understand policies, or draft better replies while keeping the final send under human control.

Because it is trained on the same core resources as Hugo, Copilot is especially useful when an agent needs help finding the right answer quickly without leaving the conversation.

Summaries, topics, and voice transcription

As AI usage gets broader, operator productivity matters just as much as automated replies. Conversation summaries help agents catch up faster, topic detection helps teams identify patterns, and voice-to-text transcripts make audio messages or calls easier to process.

These features are particularly useful when conversations are long, multi-agent, or cross-channel.

A practical rollout order

A simple rollout usually works better than an ambitious one.

A strong order for most teams is:

• Train Hugo first → website pages, knowledge base articles, and Q&A snippets before anything else
• Test in the Playground → catch weak content, missing policies, and poor prompts early
• Activate Hugo carefully → start with the channels or use cases you trust most
• Add Search Chatbox AI → reduce repetitive pre-chat questions once your content is solid
• Scale with agent-facing AI → let teams use AI Tools, Copilot, summaries, and transcription to speed up handling

Next steps

Once your first layer is live, the best improvements usually come from better training content, clearer routing, and better use of integrations rather than more prompt complexity.

Helpful resources:

• Hugo setup → Getting started with Hugo AI Agent
• Training guide → How to train Hugo AI Agent on your data
• Routing → How does Hugo Routing work
• Integrations and MCP → How to use Hugo Integrations

Aktualisiert am: 30/04/2026