A Rasa chatbot can connect to Zendesk and other business systems to handle recurring customer requests. This integration example shows an order-status enquiry: the bot collects the necessary information, retrieves order data from a backend system and returns the status. The existing diagrams illustrate the workflow using the components in the original example. For a new project, the channel, Rasa version, APIs and handoff to your service team need to work together.
What you can achieve
The bot handles a clearly defined use case that your team currently processes repeatedly. This can shorten response times and leave more time for issues that need personal support. The result depends on data quality, available APIs and how exceptions are handled.
- answer standard questions or suggest relevant help articles
- retrieve order-specific information from connected systems
- hand unresolved requests to your team with the conversation context

Two different types of enquiry
Answering general questions
A question such as “Can I pay with PayPal?” can be answered using approved information from a knowledge base. The bot can provide an answer or suggest a suitable help article. The content needs to be up to date and relevant to the question.
Checking individual transactions
For “When will my order arrive?”, the bot needs order data. The integration retrieves this from a shop, ERP or CRM system, for example. Before personal information is returned, the system must check whether the person making the request is authorised to access that specific order.

How the order-status enquiry works
- Identify the request: The bot recognises the order-status question and collects the details needed to locate the order.
- Check access: The integration verifies identity and authorisation according to the sensitivity of the data. An order number, email address or postcode alone is not a general proof of permission to access a record.
- Retrieve the status: Once the check succeeds, the connected API retrieves the information that may be shared.
- Reply or hand over: The bot returns the status and, where available, a tracking link. Missing data, failed checks and unresolved requests follow the agreed handoff to your team.
Access checks belong on the server and must cover the specific record being requested. See the OWASP guidance on authorisation for the underlying principles.
Choosing the integration for a current project
Rasa describes custom channel connectors in its technical documentation. Zendesk’s Sunshine Conversations platform provides messaging APIs, webhooks and bot-handoff controls. These components can support a custom integration; they do not establish a ready-made connection for every Rasa version and channel.
Whether an additional integration platform such as Pipedream is useful depends on the available APIs and the required workflow. Define which component processes messages, how credentials are protected and where errors and handoffs can be traced.
Rasa integration or Zendesk AI Agents?
If you already run a Rasa solution or have specific architectural requirements, a custom connection may be suitable. For a new project, also evaluate today’s Zendesk AI Agents. Zendesk has removed the former Essential and Advanced plan distinction; its product announcement explains the change. Compare the use cases, channels, integrations and ongoing operational effort you need.
Start with a manageable use case
- Review recurring enquiries: Choose a frequent question with a clear benefit for customers and your team.
- Check the prerequisites: Review APIs, data quality, permissions and the handoff to people.
- Build a prototype: Implement one defined workflow, including its failure cases.
- Test real examples: Assess answer quality, follow-up contacts and the work that remains for your team.
- Improve step by step: Refine the workflow before automating additional requests.
Would a bot integration fit your customer service?
If you want to connect Rasa and Zendesk or assess a new automation use case, discuss your existing systems and requirements with us. Our Zendesk automation page explains further options.


