Enki.ai
AI Search & AI Training Design
Rethinking & Redefining Research Iterations and Increasing User Engagement through Tailoring Accuracy.
My Role
Lead UX Designer
Timeline
March~May, 2023
Team
CEO, Engineers, Other Stakeholders
Deliverables
End-to-End Delivery
The State:
Shipped
Main Focus
Service Design;
User Research;
Design Strategy
Understand the Big Picture
The Context
Enki.ai is a forward-looking market intelligence platform. It utilizes artificial intelligence to help businesses predict potential customers and opportunities.
​
The company is in the midst of a transitional phase. After discussing with the CEO, I conducted generative research with current and prospective customers and discovered essential user needs.
After synthesis analysis, I identified and prioritized design decisions that adapt to the transition period and enhance competitive advantage with future scalability.
Scope Overview
In THIS design deck, I will mainly talk about 01, AI Search; 05, AI Training and 06, Navigation Improvement.
The Problem
After conducting generative research, I identified user behavior patterns in researching and analyzing user-defined business scopes. 01, 02, 03, &04 are the pain points associated with this design scope.
​
Also, Enki.ai strives to provide users with a reliable and comprehensive data pool, delivering highly accurate results tailored to specific business requirements. As a UX designer, I collaborate closely with the CEO, engineer, and other stakeholders to deliver end-to-end UX design. My responsibility and passion lie in assisting Enki.ai in achieving data pool customization from a user-centered standpoint.
How might we provide users with an effortless way to access comprehensive, reliable, and accurate information tailored to their specific business needs at any time?
The Solution
Evolving | Customizable | Convertible
Taking into account user experience pain points, current artificial intelligence trends, and the long-term strategy and short-term objectives of companies, our approach creatively addresses these challenges.
​
We achieve this by designing a scientific, customizable, and intuitive user journey with the data pool customization naturally embedded, enabling users to access accurate information with empowerment.
AI Search
AI Training
Navigation Improvement
Impact
With higher information accuracy through data pool customization, there's more user engagement and happiness, leading to higher retention rates, and lower marketing costs. It will also lead to increased subscriptions and sales.
How did I get there?
Following generative research, I adapted the hierarchy of effects model, modified it to best fit our design scope, and utilized it as a framework to evaluate design decisions and consistently prioritized features while considering the company's blueprint and existing constraints to ensure future scalability.
​
We acknowledge and respect that the nature of learning & analysis is an iterative process. Therefore, this enhancement of the Enki product focuses on fostering seamless iterations of understanding and analysis, with both subjective and objective factors playing crucial roles.
After rethinking the learning iteration process with the hierarchy of effects model, I redefined the research process by turning the action into part of the iteration process with AI Training.
The further defined features.
Decision Evaluation
Design Nested Loop, Enhance Accuracy
Current State Map
Future State Map
Solution 01: AI Search
Let Enki scan work with ChatGPT
Merge the strength of both entities to enhance competitive benefits.
AI Search Design
Design Purpose:
-
Assist users in gaining a thorough and objective understanding of an unfamiliar field.
-
Streamline keyword extraction, editing, and input processes.
-
Facilitate strategic analysis, question-answering, and guidance for wide recognition, exploration, and evaluation.
Solution 02: AI Training
Tailoring the Data Pools
Improve the research iterations in a broader scope.
The Challenge
Many Users do not have the patience or time to read the entire article online, unless it is exceptionally captivating or brief, which made the sample selection for AI training is challenging.
The Research
Primary Research:
Method:
01, I provided a reading sample to observe how users efficiently scan and extract key information within a 10-15 minutes.
​
02, Then I asked sample-reading related questions, such as What did you gain from this article? What information captured your attention most, and could you point it out to me? Etc.
​
03, I delve into their reading habits by asking questions like Could you describe a time when you struggled to comprehend professional details and explain why? What is your reading habit and why? Etc.
Seconary Research:
I conducted secondary research on the nature of eye movement during reading and how human beings allocate their attention when presented with a large amount of information.
Reference:
https://www.dummies.com/article/academics-the-arts/language-language-arts/reading/speed-reading-fundamental-eye-fixations-193383/
Raymond JE, Shapiro KL, Arnell KM. Temporary suppression of visual processing in an RSVP task: an attentional blink? . J Exp Psychol Hum Percept Perform. 1992 Aug;18(3):849-60. doi: 10.1037//0096-1523.18.3.849. PMID: 1500880.
The Insights
Features & Loop
I designed the feedback loop. The keywords play a vital role because they can facilitate the reading speed, as well as provide the measurement of data pool customization.
The system enables users to modify keywords seamlessly, integrating the sample selection process into their daily usage and resulting in a user experience that is effortless, enjoyable, and efficient.
AI Training Design
Solution 03: Navigation Improvement
Respect the Nature of Attention
Create a delightful user experience.
The Problem
In the generative research, many users complained about the lengthy webpage and the lack of a compact layout.
Website Critiques
I run website critiques to identify approaches for effectively organizing information with a strong visual impact.
I chose the website of Apple ( https://www.apple.com/) and the Metropolitan Museum of Art (https://www.metmuseum.org/) to let people point out what information captured most of their attention and why.
Findings
Visual hierarchy interaction is more important than visual hierarchy itself.
Primary Visual Impact > Closest information > Text in the other places with saturated color
The Improvements
Before
After
Before
After
The Prototype
AI Search
Merged our strengths with ChatGPT.
Let users quickly build essential knowledge and objective awareness through question-answering.
AI Training
Customize data pools for specific business scope
We empower users with data pool training for accuracy enhancement.
Currently we are building and improving design system with latest Figma features. The finalization of the UI will be updated later.
Learnings & Reflections
-
​​​Respect and understand the nature of human behaviors & limitations and utilize this to generate innovative solutions. Continuous research is essential in this endeavor.
-
A deep understanding of controls, flows, and feedback loops empower us to create a positive cycle for adjusting future designs.
-
Being able to see the feedback loop in more extensive and smaller scopes and build the correlation between both can restructure the solutions with more controls. It also helps me to understand the company's business from a scalable perspective.