Optimizing doctor’s visit workflow and documentation tasks with generative AI for higher user satisfaction
TLDRDocGen is a conceptual tech-enabled healthcare SaaS app designed to optimize visit and documentation tasks for doctors. The goal was to reduce the time spent on documentation to enhance doctors efficiency and patient satisfaction, leading to higher retention and adoption rates, and improve health and business outcomes through generative AI.
WORK
UX / Product Design
INDUSTRY
Healthcare
DELIVERABLES
Desk Research
Information Architecture
Wireframe + Prototyping
User Journeys
Web Application Design
for Desktop & Phone
OVERVIEWI designed a healthtech SaaS product I designed a healthtech SaaS product from 0-1, targeting two user groups: doctors and patients. Working at an early-stage startup as the only designer wasn’t the easiest thing to do, as I faced several constraints:
- we haven't had the data from healthcare specialists, so I based decisions on academic articles and general public reports about the problems and potential use of generative AI in healthcare
- haven't had an opportunity to speak with users either, but I gathered feedback from stakeholders and developers and did my best
haven’t had an opportunity to speak with users either, but I gathered feedback from stakeholders and developers and did my best
I suggested using large language models (LLMs) and retrieval-augmented generation (RAG) to analyze patient data and automate documentation.
However, this feature was shelved.
After finishing my work with the startup, I decided to go ahead with this concept on my own to understand its potential impact on both end-users and the business. Because this design was never used, no real usage data exists. Instead, this case study shows the design vision and the hypothesized outcomes if it were deployed.
USER & BUSINESS OUTCOME
Save doctors time
By reducing time spent on writing documentation, the AI feature could descrease task time by ~20%—though this figure is an estimate based on secondary research and typical EHR usage patterns, but I would tracked feature usage data.
Engage with end patient, and benefit the doctor
With doctors spending less time on paperwork and more with patient, means higher satisfaction, leading to higher retention rates, more visits and word-of-mouth referrals.
Better documentation
There would be more details in their records that may have gone unnoticed or were not noted down on time, resulting in documentation of higher quality and fewer errors.
Grow the user base and increase revenue
If doctors would see the value in saving time, improving workflow efficiency, and increasing patient satisfaction and revenue, they’ll be more likely to opt for the paid version of the software.
Overwhelming amount of medical documentation to fill out, a patient’s history to get through takes time that could otherwise be spent with patients
PROBLEM
Doctors spend up to 50% of their visit time on documentation tasks (per various healthcare studies). Users in the short visit time have to read –– sometimes very long –– medical history, understand patient’s problem, take notes while engaging with patients and manually fill out the documentation.
There’s no way for the doctor to do it thoroughly if the patient has a huge amount of papers, test results, scans and notes from other doctors.
It became hard to fully focus on the diagnosis and not to leave some data unseen. Doctors spent excessive time on desk work, especially documentation for patient visits, often at the expense of patient care. This leads to patient dissatisfaction—either due to reduced time with the doctor or longer wait times.
Better diagnosis leads to higher patient satisfaction and thus retention
I focused on a feedback loop where improving the doctor’s workflow enhances patient experience, which ultimately benefits the business:
- faster documentation → more time for patients → better diagnoses → increased patient satisfaction
- higher patient satisfaction → more follow-up visits and referrals → increased revenue for doctors → higher adoption of the app
- reduced doctors' workload and increased efficiency → higher retention & upgrades to paid features → higher Life Time Value (LTV)
To have more time for patients, doctors must either: extend the visit duration, reducing daily patient volume, or reduce time spent on documentation, enabling a better patient experience without sacrificing efficiency.
How might we help doctors conduct more thorough visits without reducing patient throughput?
Hypothesis: if we reduce time spent on documentation, doctors can focus more on patient care, leading to higher retention and better health outcomes.
Helps with repetitive tasks doctors need to do regularly and enables doctors to fully focus on diagnose without worrying about not having enough time.
SOLUTION
Handle patient data
Doctors waste time reading unstructured patient data. DocGen extract & summarize medical data: using LLMs and RAG to distill data from years of patient history—even from scans and handwritten notes from other doctors—into organized summaries.
Auto-fill documentation & forms
Manually entering notes is tedious and time-consuming. The app helps in organizing medical records and standardizes data entry methods by automate note-taking during patient visits, transcribing dialogue, and then –– based on the trascryption –– completing documents. The doctor would no longer have to deal with taking notes.
RISKS
Doctors are used to doing these tasks manually, making adoption to new software potentially difficult. Non-technical users may feel hesitant to change their workflows.
Show the new technology in understandable way
DESIGN DECISIONS & FEATURES
Understand the technology
Users unfamiliar with AI may feel anxious about where to begin, get frustrated and stop using the app. To ensure high feature adoption rates I designed open-ended input with general suggestions that help users understand how it works. First-time users can use pre-defined prompts, while experienced users can write their own.
Offer suggestions to save time and effort
Writing prompts can be hard to put together, especially for less experienced users. With a sample set of prompts, we show users how to use AI and suggest the best strategies that’ll work for them.
Less friction with progressive disclosure
Revealing every feature up front can be overwhelming. Instead, show advanced functionality only when they provide the most value for a specific context.
Unlock features at the right moments
Show more detailed suggestions only when it’s relevant to the user and based on the context, so there shouldn’t be a ‘Summarize patient history’ option if there’s no existing data.
FINAL THOUGHTS
Health-tech was design space I was unfamiliar with. “You know nothing, John Snow”, that’s how I felt. But it turned out that learning about a specific field gives me that dopamine hit and enables me to make better design decisions.