New Adventures to Solve
I had the amazing opportunity to pioneer the new AI Client Assistant from scratch as the UX/UI Designer. Our main objective was to reduce the client's time spent posting jobs, pulling reports and searching for data by 30%.
We wanted to create an AI assistant that would allow clients to ask questions about information within the application and provide accurate answers within the perimeters provided. Our first iteration was to create UI concepts for FAQ functionality, while working on the AI tone of voice.
Research & Discovery
If clients find the AI assistant easy to use and helpful with accurate information, it will be their "go to" in completing their tasks and reduce their overall time.
Understanding the User
I reviewed current user personas for the talent management product, I identified the different types of users who would be interacting with our AI Client Assistant product and updated information about the needs, goals, and pain points of these users; establishing detailed AI profiles of each user persona. These personas were then used to guide the design and development of the talent management product, ensuring that it meets the needs of our target audience.
Understanding the User & Problem
Workflows
I created workflows for each task the personas took, and it played a crucial role in ensuring that everything step ran smoothly. We were able to easily track progress, identify bottlenecks, and make adjustments to optimize efficiency.
-
Posting a New Job
-
Clinician Extension Requests
-
Evaluating & Downloading Candidate Documents
-
Reviewing Clinician "Start/Stop" Dates
-
Adding New Users to User Maintenance
-
Investigating Financial Reports for Facilities
-
Analyzing & Approving Candidates
Understanding the User & Problem
Journey Maps
Working through the different users' actions and steps for specific goals, I documented various feelings and pain points that established opportunities for our success.
-
Posting a New Job
-
Clinician Extension Requests
-
Evaluating & Downloading Candidate Documents
-
Reviewing Clinician "Start/Stop" Dates
-
Adding New Users to User Maintenance
-
Investigating Financial Reports for Facilities
-
Analyzing & Approving Candidates
Understanding the Problem
Timeframes
Using FullStory, I documented the average time it took the client to complete a specific task in their workflow.
In order to reach our goal of client's reduced time by 30%, each timeframe will be compared to the AI assistant's timeframe to complete the same specific task.
Ideation & Design
Starting with crazy 8s, sketches and wireframes, I researched inline and modal pros and cons, dark mode and light mode ideas, gathered my concepts and presented to my UX / UI team on other products for their feedback. After incorporating some great ideas, I worked with the Product Manager to narrow down and create high fidelity screens. I presented a story to our internal stakeholders for approval to move forward. Once we decided on a concept, I kicked off the components.
I was also tasked with creating the Tone of Voice for the AI Assistant. I documented primary and secondary words to describe our AI's personality, along with tips, guidelines, permissions and information about the AI Assistant that would be used internally by our MLabs development team on how our AI would respond and interact with users. I created a list of common questions for MVP and future questions, and then worked with the Product Manager to document responses, instructions and links that would be used.
Brainstorming & Inspiration
-
Brainstorm / Crazy 8s
-
Sketches
-
Wireframes
-
Inline / Modal Pros & Cons
-
Light Mode / Dark Mode Concepts
-
Components
-
High Fidelity Concepts
Tone/Voice
Through a focus group, it was determined that we wanted to be the business side of the "mullet" (business in the front, party in the back). We wanted our AI to reply / sound conversational but still straight to the point.
-
Primary & Secondary Communication Principles
-
MVP FAQ Questions and Answers
-
Future Questions
Concepts
After research and working through different brainstorming techniques, it became apparent that visualizing the UI would be an important task to get started on and spend time iterating the different concepts.
From inline, split screen and modal options to light mode vs. dark mode. Ultimately, a modal in a light mode UI was the path we were going to take for desktop and modal.
Usability Testing
Over 95% of our clients use desktop devices, I started to develop a prototype and gather questions, assumptions, steps and success criteria for remote usability tests. Our objective was to get feedback on the following:
-
Successful Task Completion: Could they locate how to get the AI modal started? Were they able to successfully ask a question? How did they feel about the response?
-
Critical & Non-Critical Errors
-
Error-Free Rate
-
Time on Task
-
Task Level Satisfaction
Lessons Learned
The most obvious constraint was AI being fairly new, so a lot of research went into best practices, but our application limitations were discovered during the journey mapping and what development could ultimately do within the back end.
We also discovered during usability testing that a link to their ultimate destination was the top priority (in addition to displaying steps on how to get there). Participants found the UI to be easy to navigate, but also had strong emotions on how the assistant was presented (your personal assistant vs. the onboarding specialist's assistant.