“User-centric AI (artificial intelligence)” is the most important challenge facing modern companies. This is because no matter how advanced the AI technology is, it is meaningless if it is difficult for actual users to understand and utilize.
McKinsey has analyzed that the introduction of AI can improve corporate productivity by up to 40% in a 2023 report, but many companies are experiencing unexpected difficulties in the process of introducing AI. In particular, the reluctance of practitioners and the difficulty of using the technology are pointed out as major obstacles.
No matter how advanced the technology is, it is ultimately people who use it. No matter how sophisticated the AI produces results, it cannot be called a true innovation if the process and results cannot be understood and trusted by the user.
Narrowing the gap between technology and users is the most urgent task that the AI industry must solve.
Deepflow, a demand forecasting AI solution, offers one answer to these challenges. Deepflow goes beyond simply providing accurate predictions, helping users understand the process by which the results were derived.
This is an important key to making AI perceived as a partner rather than a threat.
In this interview, we met with Eunjin Yoon, a product planner at Deepflow. She majored in UX design and worked as a UI/UX designer, and then switched to product planning to design a more in-depth user experience.
Currently, Deepflow is planning various features to help users use AI more easily and effectively.
Through this interview, we learned in detail how user-centric AI is being implemented and what efforts are being made to achieve this. This will provide important insights for companies considering the introduction of AI.
I majored in UX design in college. I studied user research methodologies and the process of discovering insights in depth.
After graduation, I worked as a UX/UI designer in the ESG field, but I wanted to work on the direction of the product and more in-depth aspects, not just the screen or UI, so I switched to being a planner.
Do you have any special reasons for choosing the AI solution field?
At Deepflow, we are working to solve this problem. We are making AI a partner that can provide more insights and help, rather than an adversary or competitor.
For example, when doing demand forecasting, AI helps us understand how the forecast value was derived and why this value was obtained. In this way, I believe that AI can be positioned as a helper rather than a replacement for work.
It was when I visited a certain client. The company was suffering greatly from excess inventory. The cost of inventory management was so serious that it was threatening the entire company.
That's when I realized the importance and necessity of our product. It was very rewarding to be able to create a valuable solution that could solve this problem.
We go through a lot of internal verification processes. When we provide the results to the user, we have a lot of discussions about whether the results are actually helpful and whether the values are derived through a clear logic.
At first, I didn't think that so much verification and discussion would be necessary. If we determine that a method is not sufficiently convincing to the user, we create alternatives and repeat the process of creating another alternative.
However, this process is not always negative. We are creating better results by sharing opinions from each of our perspectives as planners, engineers, and researchers. We are experiencing true teamwork.
I think my ability to implement is my strength. As a designer, I have the ability to visualize, so I can quickly show the actual appearance in the process of internal communication and discussion.
Rapid prototyping is important for startups, and my strength is being able to quickly visualize ideas and continuously develop them.
Providing insights is the most important thing. Our company's core goal this year is also to provide new insights to users.
AI is often seen as a black box, but we are aiming for explainable AI.
We believe it is important to not only present perfect results, but also to communicate them in a way that allows users to understand what factors influenced the results, what trends they show, and what variables they are composed of.”
Of course, this does not mean that our service is not accurate. Rather, it is important to provide accurate prediction results and to make it possible for users to understand the process by which the results were derived.
We intend to transparently disclose all processes, including which variables have affected the results, what trends are currently being seen, and even which variables the AI is operating based on.
I'm spending a lot of time choosing terminology. When developing AI solutions, I naturally use technical terms, which can be difficult for users to understand. I'm trying to use words that are as easy to understand as possible.
In addition, when users want more detailed information, the manual provides a conceptual explanation and instructions on how to use the service. While most guidebooks only list terms, we try to show the overall flow of the service.
For example, I explain the overall flow of how demand forecasting data leads to inventory management, and how that in turn creates insights, and how user input is reflected in the process to develop better forecasting and inventory management.
I take the lead in the bi-weekly sprints. I use this opportunity to continuously gather user feedback.
The sales team gathers the requirements of customers collected on site, the technical proposals of the development team, and the research results of the research team. Based on this, the entire team gathers to discuss and share their opinions from their own perspectives.
After all opinions have been expressed, the team prioritizes them and systematically reflects them. This process allows us to set a direction for development that harmoniously combines internal and external perspectives.
To be honest, it wasn't easy to lead. It took a lot of courage internally.
But I thought this system was necessary to develop our products. I thought it was important to start with small features and gradually develop them.
It can be difficult to communicate when we are focused on our own work. It is difficult to suddenly bring up a topic like, “I had this idea, let's hear it.” So we have regular meetings to share our thoughts.
The unique feature of our organization is that all members can freely express their opinions. We have people with different perspectives. Some want to keep moving forward, while others want to carefully consider things.
I think these different perspectives create a healthy organizational culture. If everything is calm in a startup, it can be a dangerous sign.
The sales team communicates customers' requirements and insights, and the development team not only develops, but also proposes new features with a sense of ownership. All team members record ideas and improvements that come to mind while working, and then discuss them together during the sprint.
The Wednesday Gourmet Club is a small group meeting held within IMPACTIVE AI. It is not actually held on Wednesdays and is not related to food, but it is an important meeting held every two weeks with a different agenda each time.
We usually share ideas for product advancement and make annual plans together. Sometimes we invite outside experts to strengthen our internal capabilities.
Through these meetings, we share our thoughts and discuss the direction of the company together. In particular, inviting external experts to gain new insights is very helpful.
In both the sprint and the demand-based food club, all members look at the product from their own position and share their perspectives. The sales team presents the voice of the customer, the development team presents the technical possibilities, and the research team presents the research perspective.
I think the process of aligning these various perspectives in one direction is really important. Sometimes there may be conflicting opinions, but that is what drives us to create better solutions.
In the end, the CEO adjusts the priorities, but all members can freely participate in the process of giving their opinions.
This culture is possible because our organization aims for horizontal and open communication. All opinions are respected regardless of rank or seniority, and discussions are held in a constructive manner. I believe this is the greatest strength of our organization.
A year from now, Deepflow will be focusing on time series forecasting as it is now, while developing and commercializing various innovative services based on this.
I hope that it will contribute to increasing the productivity of more companies in Korea and grow into a global service by expanding its presence in overseas markets. In particular, we plan to further enhance the AI-based insight-providing function to help users better understand and utilize hidden patterns and opportunities in data.
Yes, we are planning to develop an interactive feature that will allow users to input data and adjust various variables to see the changes in the results. The goal is to help users experience the interaction between data and models through experiments and derive their own insights, rather than simply receiving the results.
Of course, this is still a stage that requires verification. However, we expect that if such a feature is added, users will be able to use the service more actively.
That's right. We are currently focusing on predictive AI, but we are preparing to provide better predictive capabilities and services through a combination with generative AI. In particular, we are focusing on using generative AI to explain the results of predictions in a more rich and insightful way.
We want to go beyond simply interpreting data and deliver results in a form that users can easily understand and apply. Currently, we are thoroughly verifying the accuracy of generative AI response results and continuously reviewing the possibility of commercialization.
My goal is to grow this product together. I want to be involved in every step of the process, just like how a woman goes through the process of conceiving and giving birth to a child.
And even after the product is released, I will continue to develop it so that it can grow properly and be loved by many people. To do this, I want to constantly find what is needed and develop it to grow it into a product that delivers true value.
I would like to tell them not to be afraid. As you grow as a junior product planner, you will always face new challenges. It can be scary to face situations where you don't know if you can do it or not.
However, if you are gripped by fear, you will not be able to start anything. You should try what you can do now one by one, and even if you fail, think about the next step and move forward step by step.
I think there is a big difference between not starting and starting and failing. If you take one step at a time, the path will surely open up.