CINNOX: AI Assistant


To enhance the efficiency of the enquiry process for agents, we explored the integration of AI into CINNOX. Our goal was to embed support at key points in the customer workflow and deliver valuable insight to administrators.

AI Label and AI Summary serve as the main features in the project. To ensure usability for the majority of customer scenarios, we gathered feedback from customers and utilized Google event log data to define a scalable scope.

My Role

Leading the web and app UX/UI design with product owner in the project.

Design Highlight

Obvious, Accurate and Instant Support

The labeling process entails several invisible steps for agents, including identifying keywords, searching for labels, and ultimately deciding to tag an inquiry or contact. With AI support, agents can directly proceed to the final decision step. Moreover, the correct rate of AI-suggested labels is up to 99%.

Concise Automatic Summary

Auto-summarization reduces the time required for agents to handle inquiries. By referencing the AI-generated summary, agents can directly adjust the content, saving time on typing, content checking, and organizing briefs.

Due to commercial package and feature dependencies, call summaries are more complex compared to chat summaries. Successfully generating a call summary requires the system to go through 3 steps. Designing error handling was particularly challenging in this regard.


Competitor Product Benchmark

With the trend of AI, numerous competitor products have been released with many AI-related features. These provide valuable reference into industry application standards and process.

Google Event Log

The event log reflects how frequently customers use labels in the system. This can serve as a reference to ensure the effectiveness of development.

User Journey: Agent Handling Enquiries

Through interviews with customers, we gathered feedback and summarized a general process for handling enquiries. This journey helps us understand their workflow better, ensuring that our design is effective.

This is a simplified version of the user journey, focusing only on labeling and summary.

Outcome & Retrospective

After launching the feature for Proof of Concept (POC), we received positive feedback on usability from customers. However, a known issue that concerns us is the fee associated with OpenAI.

As expected, after some time, we heard from customers that they may pause enabling this feature due to the fee. This also reminds us that in the future, we need to consider not only agents (the end users) but also administrators and purchasers (the decision-makers).

Due to confidential restriction, please contact for more details.

Website design and content
© 2024 Chad Liao