Continuing education provider, owner-led
We built an agent with the customer service team
Customer service at an owner-led continuing education provider wanted to build its own n8n workflow and be able to develop it further after the workshop.
| Industry | Education |
|---|---|
| Services | AI Workshops, n8n Automation |
Course inquiries required the same manual searches
Customer service received recurring questions about course schedules, access and participation documents. Staff found answers in approved course information and adapted them to each inquiry. The team had ideas for automation but had not yet built its own workflow.
The business owners wanted to understand how inputs, knowledge access and answer drafts connect. Receiving a finished system would only partly have met that learning goal. At the same time, the workshop could not leave customer service with uncontrolled sending or unclear technical access. Ongoing support needed to be settled in advance.
The team needed to explain its workflow independently
Our task was to build a manageable n8n workflow with an LLM step alongside the business team. Participants needed to run processing themselves, recognize errors and make simple changes. The initial workflow prepared answers from approved course information. External communication remained subject to manual review until the business team separately approved broader use.
The workflow made every processing step visible
Sanitized inquiries provided the input
Participants started the workflow with prepared service cases. Personal details were removed for the exercise; course references remained as necessary context.
Knowledge access was restricted
The workflow read approved course information from a controlled source. Missing content produced a visible flag for the person handling the case.
The LLM step created a draft
A jointly written instruction required answers based on the available knowledge. The draft included source references and was saved in a review interface.
Handover included the workflow and troubleshooting help
The team received the workflow in its own workspace, sanitized test cases and instructions. Credentials remained separate in the designated credential manager.
Participants built and tested the workflow themselves
Beforehand, we clarified the working environment, permitted data and support with service leadership and IT. In the workshop, participants broke a typical course inquiry into processing steps. We built the n8n nodes together and had the team check the visible result after each step. This let them identify where an error occurred.
Service staff wrote the answer rules themselves. We helped with technical connections and structured outputs. Participants then changed inquiries and course information. A deliberately missing section showed whether the workflow exposed the gap or generated an unsupported answer.
We limited the scope to a clearly defined inquiry type. Additional channels and write access would have required more operational decisions than the learning goal needed. We recorded those extensions for later assessment. For the existing workflow, service leadership and IT agreed on who would maintain content and who would handle technical errors.
At the end, the team started the workflow without our guidance and corrected a prepared misconfiguration. We added open questions directly to the handover documentation. Participants could then explain how to stop the workflow, discard a draft and report an error.
The business team left with a runnable workflow
The workflow was in the team’s own workspace and could run with the prepared cases. The team could review draft answers and make limited changes to the instructions independently. Contacts were named for technical changes and a later connection to live intake. Continuing the work no longer depended on external help setting it up again.
A learning project needs a runnable handover
Let participants fix errors themselves
A faulty connection reveals whether the team understands the workflow well enough.
Limit the first inquiry type
A clear case can be built and tested completely before additional channels extend the workflow.
Clarify access and support in advance
The workflow must remain available in the team’s own workspace after the session and have responsible contacts.
A first conversation takes 30 minutes.
Together, we select a service task your business team can build into a workflow during the workshop.