Skip to main content

Chatbot checklist

Assumptions​

  1. You are creating an industrial chatbot with a specific use case such as customer support, maintenance scheduling, information resource, engineering or spare-part replacement.
  2. You have an established brand voice and tone.
  3. Your team has the required domain-specific and user knowledge.
  4. You work in a team with developers, designers, and writers.
  5. You test continuously or iteratively.

Follow these steps and answer the questions with your teams to create your industrial chatbot​

1. Define your industrial use case

  • What is the purpose and scope of the chatbot?
  • Which resources does it need to access?
  • What constraints or limitations does it have?

2. Create your system persona (follow our steps)

3. Decide on the right technology stack

  • Which technology is most appropriate? Consider programming language, framework, NLP model/tools and deployment options.

4. Create user journeys

  • Where is our chatbot used inside the user workflow?
  • What is the intent of our users? Use sample dialogs to clarify explicit and implicit intent.

5. Develop chatbot with your technology stack

  • How does the chatbot integrate into existing systems?
  • How does the chatbot scale as use cases increase?
  • How do we implement authentication and authorization with various user roles?

6. Train your chatbot

  • How do we gather relevant industry-specific data?
  • How do we annotate our data?
  • How do we monitor and update our chatbot?

7. Test your chatbot

  • Does the chatbot respond with natural language? Assess NLP model and tool.
  • Does the chatbot use your brand's voice and tone?
  • Does it respond appropriately? Test error handling, task efficiency, and handling vague requests.
  • How do we collect user feedback? Consider rating systems, post-query surveys.
  • How do we assess the chatbot's performance? Consider response time, accuracy, successful/abandoned interactions.

Dos and Don'ts​

  • Do work with all stakeholders to ensure transparency
  • Do customize chatbot responses for project-specific terminology and processes
  • Do consider how your chatbot hands off to humans for complex issues
  • Do read and review interactions (with user consent)
  • Don't assume all chatbot interactions are successful
  • Don't forget to retrain, test and update your chatbot