Discover The Perfect Therapist With The AI-powered Matchmaker

An AI-driven mental health platform design





Sept. - Nov. 2023
Desktop Application
Ziyu Song, Yuchuan Yu
Contextual Research, User Research, Wireframe, Design Iteration, Hi-fi Mockup, Prototyping, Usability Test

Project Overview
JanesForest is an AI-powered platform crafted for psychotherapy and mental health management, which offers individuals seeking therapy an efficient, cost-effective, and personalized therapeutic journey. This is a school work that I conceptualized and developed in collaboration with Yuchuan Yu for a Human AI Collaboration class. Our aim was to delve into innovative possibilities for applying AIGC technologies within the mental health industry.


Product Overview

JanesForest is an AI-powered platform crafted for psychotherapy and mental health management, which offers individuals seeking therapy an efficient, cost-effective, and personalized therapeutic journey.

Design Background

Primary Research

People are spending too much money and time on matching with a therapist

More than half of adults in the US who are receiving mental health treatment are not taking any therapy treatments. While at the same time, it normally takes 3-5 therapy sessions to meet a matched therapist. Meanwhile, therapists are facing an overwhelming workload which makes 45% of them feel burned out.

Why AI providing opportunities?

Help users lower the threshold for understanding psychological counseling

With the development of AI technology, we believe AI can help users quickly grasp the basics of psychological counseling, making the counseling process smoother, while also reducing the workload of counselors, enabling them to assist more people.

Thus, we asked

HMW create an satisfying psychotherapy experience for first-time users by leveraging the power of AI to provide efficient and accurate therapist matching and personalized services

Check The Outcome

Assessments with AI before matching up with a therapist. Therapists were able to view results before the first session

Simulated conversation styles and psychotherapy methodology with AI to help users match up with therapists

AI automatically generate tasks on daily task session as reminder for users

AI automatically taking notes during therapy sessions for users

Dive Into The Problem - Primary Research


Initially, I conducted a competitive analysis to understand how existing products contribute to mental health development.

I categorized existing products into 6 different types, and then conducted further study on three types that may relevant to our product. And I found some insights that promoted later design process.

1. Unfriendly to new users: Exiting online therapy application lack of assistance of onboarding. Besides, with tons of unfamiliar knowledge, it is hard for users to get on the track.
2. Have difficulties to keep daily practice: Many of methods require daily practice. It's hard to keep practice without reminder and daily progress tracker.

3. No prompt or visualized progress: It's not easy for users to find their progress and how they make the progress individually from the symptom tracking apps.


Subsequently, we delved into user research by creating personas and conducting interviews with both users and experts to gain deeper insights.

Following the completion of 10 user interviews and 2 expert consultations, we categorized our target users into two groups according to the duration of their mental health challenges. Individuals grappling with mental health issues for more than six months were classified as experiencing long-term mental pressure, while those facing such issues for a shorter period were considered to be under short-term mental pressure.

After the interviews, we tried to come up with porduct opportunities by conducting theming process and an affinity map.

Selected Themings (Pain pointes):
1. Struggling selecting a therapist: Users struggle to select a therapist with limited knowledge of psychotherapy methodologies.
2. Having problem conduct and keeping auxiliary treatment: Psychotherapy treatment path is very personalized with various methods that users are not familiar with. It takes time for users to cultivate and stick on the routine.

3. Having problem understanding and capturing insights from therapist sessions: It is hard to remember and understaning all insights from sessions since users are having emotions and tend to fall into the same cognitive cycle.

4.It is hard to end the treatment: It’s even harder to end a companion with a therapist since their relation is very intimate.


We ended the research by understanding the whole jurney of users' experience by a users journey map

We detected three steps that will happen on the app and identifying emotional low points during the process

The Down Points:
1. Matchup: Users struggle to match with a therapist with limited therapy knowledge
2.First Session: Users have difficulty recalling and comprehending therapists' guidance while dealing with both emotional and cognitive challenges
3.Reflecting: Users find it challenging to initiate and track daily mental care without post-therapy reminders

Research Wrap-up

Time-consuming browsing and repetitive assessments

Struggles in matching with an ideal therapist

Challenges in initiating and maintaining daily mental care routines

Tough to grasp or remember therapy insights.



After that, To better form the application and came up with all the features users need from the app, we did a brainstorm and kano model study.

Patent mapping of final features on the app.

Design Process

Users workflow - Matching up with a therapist

Users workflow - Attending a session


After drawing the users flow and information architecture, we created four main screens of the web app and draw the whole flow with amid-fidelity wireframe

I also did some iteration during the wireframe phase after testing with my colleagues

Hight Fidelity Prototype

Test and Iteration

Usability Tests

After finishing the high-fidelity prototype, we conducted a round of usability tests and iterated our design according to the insight from the test session.


And here is what I changed