My process

I’ve found that my process is ever-evolving, The process outlined below is a general overview. Some elements have a greater priority than others depending on each case. I typically use Google Sheets, Miro, Manychat, Dialogflow, and sometimes Intercom.

1. Define the objective of the chatbot

This is a workshop with the relevant stakeholders to understand key business goals. We discuss:

• • The brand
  • Technical requirements
  • Understand the users and their goals
  • Use cases

2. Create a bot Personality

• This is a workshop with the relevant stakeholders to understand and agree on the communication style of the bot. We discuss:
  • Key characteristics of the bot
  • Create style sheets to define communication style and tonality

At the end of steps one and two, I have a clear understanding of the brand messaging, the bot personality, as well as what use cases to design in the first design sprint. 

3. Map out the flow

I define the happy paths throughout the conversation using a flowchart. This helps communication with stakeholders and team members since everyone has an overview of the scope of the chatbot.

4. Define intents

• Review existing data from the customer support center, emails, chat logs, etc.
• Define the most important intents users have when interacting with the bot
• Map out the exact phrases they use to express themselves

5. Dialogue creation and build

• Create dialogue sheets
• Write dialogues for the welcome message,  the happy flow for the use cases previously defined, handover to live agent, local and global error handling, etc.
• I often test immediately after writing a dialogue, or only a part of a dialogue, to get a better understanding of the user experience. For example, if I have written a welcome message, I would set that up in my own chatbot testing environment to see how that message looks, reads and feels when expressed by a chatbot.
• If I’m the sole chatbot writer and builder in the project I build the chatbot flows and dialogues in the software we’ve agreed on, for example Manychat and Dialogflow.

6. Review

I present the solution to the stakeholders, get their feedback, and edit the copy if needed.

7. Testing

Testing the happy flows and error handling on my own, as well as letting the client test the solution. I often ask someone unfamiliar with the project to understand how well they navigate the chatbot conversation and interview them about their experience.

8. Launch & Optimize

After launch, I typically monitor the chatbot, run split tests, share reports with stakeholders and implement new use cases as necessary.