Overview.

My role

With the explosion of ChatGPT and LLM in 2023, many tech companies have begun exploring the use of AI in their products. As a biological data analytics platform, Ragene has also started integrating AI technology to empower life science research.

Team

2 Product Manager

4 Developers

2 Bioinformatics Scientists

This case demonstrates how I, as a product designer, define AI features for integration with the platform based on user requirements and design these AI features for different stages of the product roadmap, taking into account business needs and development resources.

Timeline

2023-2024

Context.

What’s Ragene?

A no-code biodata analysis platform that accelerates scientific discovery.

In bioinformatics, mastering programming algorithms for data processing often presents a steep learning curve. Ragene's no-code data analysis platform significantly streamlines this, boosting researchers' efficiency.

If you'd like to learn more about my other contributions to this product, feel free to check them out at the following link :)

Why Ragene AI?

1. It's pretty obvious that AI is the biggest gimmick to attract user & investment right now.

From business side, with the rapid advancement of AI technologies, it's clear that AI has become a powerful tool for attracting both users and investors. Leveraging AI capabilities could make product stand out in the competitive biodata analysis market.

2. It's a good chance to embed AI for Ragene to boost users’ productivity in life sciences.

From user side, by leveraging AI capabilities, we can streamline complex processes, automate routine tasks, and provide intelligent insights that significantly enhance the productivity of life science researchers and accelerate scientific discovery.

Kick off.

Understand User

There are two types of user groups with different biodata analysis skill on our platform.

Based on previous user research, our platform is designed for various roles involved in biological data within life sciences, including doctors, researchers, students, biologists, and more. However, we've categorized these roles into two main types based on their data analysis capabilities.

Identify Opportunity

Facilitate workshop with internal users to ideate AI application scenarios.

To explore AI application scenarios within the Ragene platform that meet user needs during the data analysis process, I organized a workshop with representatives from both user types mentioned above. The workshop led to the following findings:

Develop Roadmap

Prioritize features based on business needs and development challenges.

After conducting the workshop, I identified several specific AI features that could be integrated into the platform. However, during discussions with the team, my PM highlighted the need to quickly demonstrate our platform's AI capabilities with minimal time investment, from a business standpoint. We then built a roadmap of AI features aligned with this business goal:

Design Phase 1.

Use Cases

There are 3 main use scenarios of user interaction with chatbot.

In the first phase of design, which focused on the AI chatbot, I aimed to enhance the chatbot's ability to provide help and guidance to users during the analysis process. To achieve this, I reviewed previous user research to understand the common questions users ask and summarized them into three main use cases:

Design Highlight

1. Streamline user-chatbot communication to make it effortless

As we know prompt is the most crucial part in AI product to help us get our desired answer, appropriate question is the prerequisite for getting accurate answers. However, it is difficult for users to accurately describe questions to the AI through short language within Ragene’s GUI.

Therefore, I set the anchor point "Ask AI" on the user's question target, so that the user can ask a question without having to describe the question target in text.

2. Optimize the Chatbot's content readability

User research revealed that bioinformatics queries to ChatGPT often yield lengthy, detailed answers. While standalone webpage of the ChatGPT can accommodate extensive text, the challenge on the Ragene platform is compressing ChatGPT's responses into a chatbot format without compromising readability.

So, I ensured user-friendly readability and aesthetics by designing the chatbot to be size-scalable and the parameter setting to be position-adaptable.

Design Phase 2.

Design Iteration

1. AI can automatically generate analytics pipeline based on prompts, rather than just offering textual guidance by Chatbot.

As the technical team developed the AI functionality, the application capabilities of the AI were optimized.

The first significant improvement is that, unlike in the first phase where users could only receive guidance on building an analysis pipeline through text interactions by Chatbot, the AI in the second phase can now directly generate an analysis pipeline for a data node based on the user's analytical needs. To accommodate this advancement, I designed the following user process:

2. AI can automatically fill out the parameters of the node, rather than just offering textual guidance by Chatbot.

The second improvement is that, whereas in the first phase users could only interact with the AI via text to receive guidance on setting algorithm parameters, the second phase of the AI can now automatically populate the appropriate parameters directly based on the user's analytical pipeline. I have designed the following user process to accommodate this enhancement:

Design Phase 3.

New Finding

Users are concerned with the analysis process because they care about the insights and output.

During user testing for the previous phase of AI functional design, we gained new insights from user feedback that the primary JTBD for users performing data analysis is to make meaningful biological discoveries from the analysis results and to document their findings. This insight also inspired the development of new AI features:

New Design

1. AI interpretation of data from completed run nodes.

In the traditional data analysis process, users interpret data visualizations, gain insights, and then record them. With the introduction of AI, this process can be automated: AI will automatically interpret the analysis results and generate text. Users will only need to:

• Click on the run completed node that needs interpretation.
• Click on the AI button to generate the interpretation.

2. AI generate report by combining interpretations of data from completed run nodes.

The current technical logic for generating the analysis report involves combining data interpretations from completed nodes to create the report. This requires users to first select the nodes they want to include in the report. For this user flow, I designed two approaches and worked with the team to select the best option.

Design Phase 4.

Design Iteration

Driven by user feedback & product strategy.

During user testing of the added AI functionality, I observed that users need to access AI features in different areas of the interface, leading to two issues:

User Perspective: Users cannot develop a consistent mental model for using AI features. Product Perspective: It is challenging to highlight the product’s AI capabilities when functions are scattered.

Therefore, in the design iteration, I systematically centralized the previously scattered AI functions into the data nodes, making the functions and user operations more intuitive.

Reflection.

Design Highlight

1. Differentiate AI content visually

At the same time, I am learning about the AI field and applying it to my design work. For instance, when designing AI products, it's crucial to differentiate AI-generated content from native content, enabling users to easily distinguish between them.

2. Explain how machines think

At the same time, I worked with the development team to ensure that the reasons for AI-generated content are clearly provided within the product, so users can better understand it.

🎉 Phew~ You made it!