Firstly, we conducted a strategy workshop to fully immerse ourselves in the University’s thought processes and expectations of this project.
This workshop included the below talking points:
- THE PROJECT INSIGHTS: We reviewed how the project first came to light and discussed/reviewed any quantitative and/or qualitative data already collected.
- OVERALL GOALS AND OBJECTIVES: Collaboratively, we identified the main user and business objectives for the chatbot to achieve.
- UNDERSTANDING YOUR USERS - We created user stories, defined the user groups and, in a who/what/why format, understood how a user will interact with the chatbot.
- CONSIDER THE BRAND AND BUILD YOUR BOT’S PERSONALITY: together we completed interactive tasks to clarify the tone of voice and personality of the chatbot.
- TOPIC/CONVERSATION STRINGS THE CHATBOT WILL COVER: We Identified the phase one key topics the chatbot will cover, looked at the depth of each conversation and the end goal the user will achieve.
- INTEGRATION AND TECHNICAL REQUIREMENTS: We gained a strong understanding of what information was going to be displayed, stored and collected from where.
- PICKING A PLATFORM: We reviewed the different platforms available and made recommendations based on the requirements and capabilities of the internal team.
- DESIGN CONVERSATIONAL WORKFLOWS: During the final task we started gathering data for the conversations using slack.
- Conversation strings were mapped out: All the information collected from the strategy workshop was digested by the team and fed into conversational strings. These then went through an iterative process to receive sign off by the client
- The Chatbot was built using Chatfuel with AI integrations from Google Dialogflow to enable the bot to fully understand the context and intent of each user conversation
- Further integrations were made with Gecko forms to facilitate the user booking an open day or campus tour within the chatbot
From our learnings, testing is one of the most important phases when building a chatbot and Jet went through an extensive testing programme which included:
- Common sense testing: when someone who has had no visibility of the project interacts with the chatbot
- Peer-to-peer testing to ensure that all the code and conversation strings are internally signed off
- The client then completed user acceptance testing
- Once the client was happy, Jet then went to an extended user group for a final round of testing