You integrate an AI chatbot into your service software at three points. It hands over conversations with the full history to your team. It reads and writes in your systems during the conversation. And it notifies your team about what they need to know. Your team continues to work in Zendesk, Userlike, or Microsoft Teams. The chatbot answers repetitive questions beforehand.
This guide shows what the connection looks like at each point. It also outlines the aspects you should define before getting started.
What role does the chatbot take on alongside your service software?
The chatbot operates upstream of your service software. It answers questions that it can resolve using your approved sources. It collects the details your team needs for a case. When a decision is required, it hands over the conversation.
Your service software stays right where it is. Your team continues to handle tickets and live chats in their familiar system, only receiving the cases that require a human touch. The Integrations page shows all the systems that Mercury.ai connects with via ready-made connectors.
At which touchpoints do you connect the chatbot and service software?
The table shows the six touchpoints where the chatbot and service software work together.
Touchpoint | What happens | How Mercury.ai solves it |
|---|---|---|
Handover to the team | An agent takes over the conversation | Handover to Zendesk, Userlike, or the Agent Desk with the full chat history |
Create ticket | The case is routed to the ticketing system | The chatbot creates the ticket during the conversation |
Read data | The customer checks a status | The chatbot queries a status in the CRM or ERP, for example |
Write data | The customer makes a change themselves | The chatbot updates an address or books an appointment, for example |
Notify | Your team is notified about a case | Follow-up actions via email or SMS, notifications from the Agent Desk in Slack |
Custom systems | Systems without a ready-made connector | Open REST API and custom webhooks |
The more of these touchpoints are connected, the more cases the chatbot resolves without a handover.
How does the chatbot hand over a conversation to Zendesk or Userlike?

The chatbot hands over to the live chat tool your team already uses. This can be Zendesk, Userlike, or Mercury.ai's Agent Desk. Your team sees the full conversation history and the details the chatbot has already captured. No one needs to ask the customer for their order number a second time.
Zendesk provides a dedicated interface for third-party bots. Your team handles tickets there as usual. The comparison of live chat tools with AI evaluates Zendesk, Userlike, and other providers.
In the Agent Desk, the conversation returns to a defined point in the chatbot flow after the live chat. The customer can then continue talking to the bot. The Agent Desk distributes conversations by department, language, or workload. It connects bidirectionally with Salesforce.
Our article on complex customer inquiries describes when a chatbot should hand over in the first place.
What does the chatbot handle in your systems before handing over?
The chatbot accesses your systems via interfaces during the conversation. It reads data and writes back to them. For example, it checks the status of an order, creates a ticket, or books an appointment. The dialogue then responds to the system's output. If the system reports an error, the chatbot asks for clarification or initiates a handover.
After a process is completed, the platform can automatically trigger a follow-up action. The customer then receives a confirmation via email or SMS, for example. The platform forwards structured conversation data to your middleware.
Our article on customer service with order data shows which service cases the chatbot resolves independently using order data. The guide to live chat integration into CRM describes which information belongs in the CRM after the conversation.
How do you connect custom systems without a ready-made connector?
Mercury.ai connects custom systems via an open REST API and custom webhooks. Your team defines the endpoint, headers, and payload in JSON format. Endpoints can change dynamically during the conversation, for example, using the customer's order number. The platform authenticates with your system using an API key, Bearer token, Basic token, or OAuth 2.0.
Custom apps and systems can also send messages to the chatbot as a channel. The platform secures these requests via Bearer token, API key, or HMAC signature. Zapier is also available for simple workflows.
A standard connector is ready in a few days. A custom integration takes two to four weeks. Your team provides access once, approves the data fields, and signs off on the result. Mercury.ai handles the connection, mapping, monitoring, and maintenance when APIs change. The guide to integration with SAP, Salesforce, and HubSpot explains CRM and ERP integration in more detail.
How does the chatbot work with Microsoft Teams and Slack?
Microsoft Teams and Slack serve as dedicated channels for Mercury.ai. Your employees can ask the chatbot their questions there. The chatbot answers using your approved documents and hands over to the appropriate department if a case requires a human. Microsoft Teams is deployed in production as a channel by Mercury.ai customers.
Your team also uses Slack for alerts from the Agent Desk. This allows them to see new conversations even when the Agent Desk is not currently open. The article on employee communication in Teams highlights which employee inquiries the chatbot resolves there. The article on omnichannel chatbots explains how a bot serves all channels from a single knowledge base.
What should you define before starting the integration?
An integration is completed faster when your team clarifies a few points beforehand. The following checklist helps with that.
Define the cases the chatbot should resolve independently.
Determine the points at which it hands over to your team.
Specify the details your team needs with each handover.
Clarify which systems and fields the chatbot is authorized to access.
Decide where the conversation routes after the live chat concludes.
Plan tests with real-world cases before the integration goes live.
Mercury.ai separates development, staging, and production environments. Your team verifies every change in the staging environment first. Automated tests check the core workflows of your configuration. This allows your team to detect early on if a change disrupts an existing connection.
How is customer data protected during integration?
Mercury.ai processes conversations on AWS Frankfurt, keeping data within Germany. Responses are generated by language models operated directly by Mercury.ai. The models are not trained on your customer data. Your team manages credentials such as API keys and tokens directly within the platform.
Data that the chatbot passes to Zendesk, Userlike, or your CRM then resides within those systems. They are governed by the respective provider's terms and privacy policies. Therefore, check with your data protection officer to determine which details are necessary during a handover. The article Where customer data from an AI chatbot flows describes the path of an inquiry across all systems.
Frequently Asked Questions
Do we need to switch our ticketing system?
No. The chatbot operates upstream of your ticketing system and routes cases to it. Your team continues to manage tickets in their familiar system.
How long does the integration take?
A standard connector is ready in a few days. A custom integration via REST API or webhooks takes two to four weeks.
Can the chatbot also transfer phone calls to agents?
Yes. Mercury Voice forwards calls to a phone number. The person answering the call hears a brief summary of the issue beforehand. Forwarding via SIP to a contact center system is available upon request.
Can our team see what the chatbot has already asked?
Yes. With every handover, your team sees the full conversation history and the details captured by the chatbot. This applies to Zendesk, Userlike, and the Agent Desk.
Want to find out how the chatbot fits into your service software stack? Let's clarify that in a 20-minute call. Talk to us.
About the Author
Dr. Maximilian Panzner
Chief Technology Officer @Mercury.ai
Dr. Maximilian Panzner is CTO and co-founder of Mercury.ai. He earned his doctorate in computer science at the CITEC institute of Bielefeld University, where he conducted research on multimodal machine learning and intelligent interaction systems. For over 20 years, he has been working on artificial intelligence, human-machine interaction, and conversational AI platforms for enterprise use.






