Optimizing medical documentation with a secure, HIPAA-compliant AI scribe tailored for healthcare professionals.
**To comply with my non-disclosure agreement, confidential information has been omitted or modified in this case study. The information presented does not necessarily represent the actual product.
Skriber MVP
Vision & Objectives
Implement an automated documentation process to increase the quality of patient interactions;
Reduce the average documentation time per patient;
Achieve 100% HIPAA compliance certification for all documentation and data storage processes.
Team
I worked as the lead product designer, collaborating closely with two product managers (Brandon W. and Anushree A.), three pairs of software engineers, and a resident doctor as our advisor. Since my company invested in the product and received equity, we collaborated daily with the founders to drive its success.
Gen-AI x Healthcare
This is a 0-1 product called Skriber, a startup founded by a group of experienced entrepreneurs in collaboration with my company. Skriber aims to revolutionize the current healthcare workflow using cutting-edge Gen-AI and LLM.
Our team is dedicated to developing a HIPAA-compliant AI that securely encrypts, transcribes, and generates medical notes. The goal is to streamline doctors’ workflows, allowing them to focus entirely on patient care without distraction.
🔥 Startup Mode 🔥
The team sees a strong product-market fit in a highly demanding domain, with a substantial total addressable and serviceable market. However, competition is fierce.
Speed is crucial—we must launch quickly, iterate rapidly, and maximize agility to stay competitive.
From 0-1 in Just 7 Weeks
Rapid Ideation
Step 1: IA Mapping
First, clarify the needs, then prioritize the features. This ensures we know the direction and can take a step-by-step approach to achieve our goals.
Step 2: Rapidly Solutions Visualization
I had to kick off design discovery and map out key user flows before receiving the branding assets from the in-house marketing team. To move quickly, I used simple blocks and shapes to represent UI components, allowing me to visualize the high-level user flow.
I’ve always believed this approach is the most efficient way to align with the internal team and initiate strategic discussions.
For the stakeholder reviews, I added a touch of color and mocked up metadata to make the design more visually comprehensible. With minimal effort, this created a clear and easily digestible mockup for both the internal team and the investor group.
The design remained flexible and adaptable, allowing for endless changes, which led to highly collaborative and engaging review sessions.
Key Feature Set (MVP)
Start a New Session
The objective during the session is to reduce cognitive load and provide full clarity on the upcoming steps.
This allows doctors to concentrate solely on the patient, without any distractions from the tool.
Optional Notes & Comments
Recognizing that each doctor has unique workflows and that documentation standards can vary across organizations, we’ve introduced an optional step for adding notes or comments. This flexible feature accommodates individual preferences without disrupting the overall process.
Summary/Editing/Sharing
Once the session ends, transcription and summarization are processed within seconds and displayed in editable sections, making it simple to manage and share specific information.
Logic behind the layout
01
It’s an industry-standard layout, requiring minimal effort for doctors and healthcare professionals to learn and use the tool effectively.
02
Our rapid testing results revealed the most frequently clicked areas, and the data aligned with our initial assumptions.
03
Prioritize ease of use with a clear, upfront, and straightforward design that eliminates guesswork.
Prioritize ease of use towards doctors needs
This is a tool for healthcare professionals, it needs to be easy to use, minimum learning effort required, minimal cognitive load, consistent on all platforms.
Focus on Viablilty & Growth
Design for business
As a freemium model, we need to finalize the membership tier structure and identify and prioritize key upsell opportunities to convert free users into paying customers. The goal is to design a scalable, highly viable product that drives business growth.
Membership Tiers
Should we offer Free and Paid options, or adopt a tiered approach like Small, Medium, and Large, similar to the popcorn pricing strategy at the theater? 🍿🍿🍿
Sense of Urgency
For free users, we not only send reminders but also create a sense of urgency as they approach the limit of their free tokens, encouraging them to upgrade.
Premium Feature Trial
Another way to encourage free users to upgrade is by offering a free trial. This allows us to unlock certain features or lift limitations, giving users a chance to experience the full capabilities of SkriberONE.
Upsell Banner
We also utilize a portion of the bottom of the interface for free users as a billboard, primarily promoting our own upsell opportunities, while also considering its potential for generating additional revenue.
MVP to V2.0 Concept
Holistic Enhancement
In V2.0, we have the opportunity to fully distinguish ourselves from competitors, reinforce our brand identity, and refine numerous details that were previously lower on the priority list.
Smart sidebar
Enhance organization by introducing features such as folder creation, customizable labels, and smart sorting capabilities. These additions will provide users with a more streamlined, intuitive experience.
Personalization with AI
We’re also introducing more AI-driven features, including style customization, summary preferences, the ability to change voice and tone in summaries, and more.
All Devices
We’re also preparing to launch the Skriber app on both iOS and Android, ensuring full adaptability to various user behaviors and providing a seamless experience across all devices.
Note: To protect sensitive information, the screens have been blurred.
Summary & Insights
Things still need solutions…
Hallucination management is a significant challenge. It’s not just a development issue but also a design consideration—how do we clearly indicate potential hallucinated results without confusing or discouraging the user? Finding the best ways to alert users while maintaining their confidence in the system is crucial.
Additionally, constantly chasing the latest AI models consumes a considerable amount of time and resources for the team. This is a common challenge in today’s fast-evolving tech landscape. We need to adapt to this rapid pace by improving our ability to predict trends and implement changes more quickly as technology advances.
Where I hand it off…
I led the effort to ship the MVP for beta testing with doctors across various hospitals and domains. While I was midway through the V2.0 discovery phase, the investors became more involved in areas like sales, marketing, and operations. Eventually, they transitioned control of the assets and started building their in-house product team.
Although I’m proud to have delivered the first version, I wish I had the chance to implement analytics and use data-driven insights to continuously improve the product.
My view on designing Gen-AI tools
Trust and Reliability: Designing AI tools that inspire trust is essential. Ensuring the AI’s output is reliable and instilling confidence in users is key. Empowering users to feel in control of the LLM is crucial to achieving this trust.
Feedback Loops: Incorporating direct feedback mechanisms is vital for both users and the product. This not only enhances user experience but also helps improve the tool over time.
Minimizing Cognitive Load: Users don’t need to understand every technical detail behind AI’s functionality. Prioritizing an intuitive, low-friction experience is more important than ever in making the tool user-friendly.
Data-Driven Iterations: Establishing robust analytics to thoroughly track user behavior is critical. This enables meaningful, data-driven iterations that align with user needs and improve the product continuously.
Collaborative Experience: Generative AI should act as a partner to users, not just a tool. Designing interactions that encourage collaboration between humans and AI is a valuable area to explore.
What to know more?
I’m always up for a chat.
Or check out some of my other projects.