Integrating Design Thinking principles into the development process may not be comprehensive enough; it is also essential to consider how AI will acquire the necessary data and continue learning over time. We combined the Design Thinking and Human-Centered AI models, which allowed us to create a structured approach that clearly delineates the primary responsibilities of each team. This fundamentally adjusts the research and analysis framework, enabling us to dive deeper into hardware technology and data requirements.

Analyze common user groups and their potential purposes for using AI tools. Considering that different organizations have varied understandings of AI applications, we used a questionnaire to gather information on the current AI tool usage of different users. This data was then analyzed both quantitatively and qualitatively to assess the AI capabilities across the company.

(Valid responses: 57, including employees both domestic and abroad)

Furthermore, we need to conduct user behavior surveys across various departments to gain a practical understanding of the specific impacts AI tools have on work. The Research phase includes conducting focus group interviews, brainstorming, outlining ideas, gathering feedback, and analysis.

The purpose of the focus group interviews is to fully understand users’ perspectives on AI tools and gather their expectations, concerns, and potential obstacles regarding AI implementation. The interview design followed a structured modular approach: "Expectations for AI in supporting the organization; sharing work scenarios; pain points in specific situations; support and concerns post-implementation; and key success factors." An AI engineer was also present to join the discussions, providing explanations and insights as needed. We encouraged participants to freely discuss any AI technologies and ideas, aiming to explore the optimal AI implementation solutions within a controlled design framework for maximum benefit.

Below is the interview plan and preliminary content outline :

Participants: Organization members & managers, AI Engineer, UX Researcher

Research Aspects : Workflow pain points, requirements, input/output scenarios

Technical Aspects: Opportunity evaluation, Data review, technical challenges

• • Each interview lasted for about 60 to 80 mins
• • All the interviews were audio-recorded and later transcribed to episode transcript
• • An Inductive approach was followed for analyzing the data
• 
  

• Once we have a clear definition of the problem, the next step is to conduct a feasibility assessment of the objectives, considering demand intensity and technical limitations. We aim to apply multiple criteria to filter out ideas that may be challenging for an AI tool to implement. We are in a stage of converging ideas, where considering solutions that effectively address a wide range of user scenarios is crucial.
• Therefore, here’s the several criteria for evaluation:

• Which issues can be resolved by implementing AI tools ?

• Which can be executed under current departmental conditions?

• • Which opportunities have the highest potential impact? What is the priority of each key point?

In the preliminary data evaluation phase, we aim to clarify a more defined path for execution.

After reviewing all the research data, the evaluation was ultimately approached from three main aspects:

1) Value Aspect : Which opportunities can maximize company benefits or enhance team collaboration?

2) Technical Aspect : The engineering team provides benchmarks for technology areas—Digitalization, Automation, BI & Predictive AI, and Generative AI. Each opportunity is evaluated to determine which is most suitable for AI tool integration.

1. 3) Data Aspect : Based on the completeness of data each organization possesses, we narrow down the priority for solution implementation.

First, Value Assessment is conducted to intuitively understand the relative importance levels of key points for each organization. We use the Task Exterprise-AI Performance Matrix, Created by researchers at Carnegie Mellon University’s Human-Computer Interaction (HCI) Institute.

Each organization evaluates key points based on their importance (such as contribution to core business functions, direct benefits, cost savings, work efficiency, competitive advantage, and operational smoothness) and impact (such as the number of people affected, scope of impact, usage frequency, and whether it influences other departments or larger processes). This comprehensive evaluation is followed by discussions among organization managers and members to reach a consensus on an appropriate matrix.

• Customizable colors for category differentiation.

• Importance assessment from the perspective of the respective department.

Secondly, Technical Aspect Considering that some solutions fall within the Digitalization phase or involve intelligent optimization of existing systems through Automation, we account for these prerequisites. Our focus highlights opportunities in BI & Predictive AI (highlighted in purple) and Generative AI (in orange), marking key needs in these areas. Through quadrant distribution analysis, we gradually identify the critical optimization points each department should prioritize.

Finally, Data Validation As priority opportunities become more focused, we conduct a more thorough evaluation of data inputs/outputs. In this phase, we carried out a second round of interviews, focusing on high-priority, actionable opportunities in the first, second, and third quadrants. The interview covered topics such as: the specific context of each key points; intended purpose of use; data sources and retrieval methods; data search actions; and data update frequency, among others. This approach helps us categorize the type of AI system users need and define the scope in system design.

The table below outlines the main items we ask users to briefly describe, including common tasks they might use AI for in their work and questions they might have in completing those tasks. Additionally, it includes sources, formats, and basic maintenance information for the data.

Based on additional data provided by the team, we have identified five key departments as the primary target users. AI engineers simultaneously begin assessing the technical framework, envisioning how the model should perform, learn, and maintain. At the same time, they work closely with us to determine specific outlines for foundational model design and data attributes, driving strategic solutions across multiple use-case scenarios.

(*Considering confidentiality, only a partial view is presented)