November 2023 Summaries

The process of building a conversational AI system involves seven steps: gathering utterances and transcripts, building an intent model, defining the scope, planning happy paths, building out long tail design, QA and launch, and business as usual monitoring. This iterative process requires ongoing effort to continuously improve and refine the agent's performance. The timeline for building an agent can vary depending on factors such as call volume, data availability, and stakeholder feedback. A well-planned approach with clear goals, scope, and timelines is essential to ensure a successful launch and ongoing maintenance of the virtual agent.

The text discusses the challenges faced by a design team in conducting design reviews, including a time-consuming manual process and a lack of collaboration with developers. The team struggled with documenting designs in multiple static tools, leading to long meetings and frustration among designers. However, after adopting Voiceflow, a conversational AI tool, the team has implemented new processes that have improved efficiency and collaboration. Asynchronous design reviews, high-fidelity prototypes, and integration with developers have become key components of their process. Reviewers can now interact with live prototypes and access the project's conversation flow in one place, reducing questions and improving feedback quality. The team has also established standard operating procedures to drive consistency across projects, promoting stakeholder buy-in and better collaboration.