Erez Druk is the Co-Founder & CEO of Freed AI.
Freed’s AI transcribes patient visit discussions, identifying key terms to create organized notes, including SOAP (Subjective, Objective, Assessment, Plan) documentation. This saves time and allows the clinician to fully focus on the patient.
Can you share the story of how moving to California to work at Facebook ignited your passion for startups and entrepreneurship?
When I moved to the US to work for Facebook, I lived with a Polish guy at the company’s corporate housing. He gave me a book called “The Lean Startup.” I read it in a few days and have been obsessed with startups since.
My passion for startups is that they very simple, but yet, running a startup is the hardest thing I’ve ever done, the most intense school I’ve been to, and if it works out, it’s the most impactful thing I could be spending my time on.
Your first startup UrbanLeap was a government procurement platform, can you discuss what this was and what were some of your key highlights from this period in your life?
The US government spends 4 trillion dollars a year through a process called public procurement. This process exists to ensure that taxpayer dollars are spent properly, but it also results in a massive waste of time and multi-million projects with no return on investment.
UrbanLeap helped 40+ local governments run procurement more efficiently and intelligently. We never managed to get it to scale and decided to shut it down.
One key highlight was to never build a product for a user I don’t know very well and care deeply about.
Can you share the genesis story behind Freed AI?
Gabi and I met 7 years ago, just before she started med school. We had a few months to fall in love before life became kind of terrible 🙂
I’ve watched Gabi and many of our friends chart at night, over the weekend, on vacation. I heard the sentence “I have notes to do” every day in the last 7 years.
So after 7 years, it seemed like a good idea to try and Free clinicians (wife included) from charting.
What are the main challenges Freed AI faces in accurately transcribing and summarizing complex medical dialogues?
Many. Clinician preferences, evolving medical terms, consistency, accurately identifying who said what, AI hallucinations, never dropping important information, not including redundant information, and more.
All these challenges are solvable, and I wish that we could solve them overnight.
What steps does Freed take to ensure that the medically relevant information extracted and summarized is accurate and secure?
Starting with accuracy, we generate more than 1 million notes every month and collect qualitative and quantitative feedback from all of them. This allows us to improve and learn quickly.
Every time we see an error we aim to develop a system to both prevent and identify a similar error.
We also make sure that the clinician has to review the note and place it into the chart themselves. Our goal is to create a great first draft for the clinician to use, but always keep the clinician in the loop.
This is a big topic we could go much deeper into.
Continuing with security, we follow industry best practices, have experts on the team, and have obtained 3rd party security certifications (SOC2 and HIPAA) and audits.
Our application architecture is very simple, which makes it relatively easy to keep it tightly secured, as long as we make it a top priority, which we do.
Given the sensitive nature of medical documentation, how does Freed ensure compliance with healthcare regulations such as HIPAA?
HIPAA is in essence a data handling standard. Our CTO is responsible to make sure that we fully follow the standard and we performed a 3rd party audit to ensure that our software is HIPAA compliant.
We also take extra measures on top of HIPAA such as not storing patient recordings, automatically and permanently deleting notes after 30 days, and more,
Could you explain how Freed AI’s transcription technology differentiates from other voice-to-text services available in the healthcare market?
Feed aims to be the best AI scribe for the clinician, not the clinic. We obsess about listening to our clinicians and building a product that truly frees them.
Specifically, we build Freed to be the simplest to use, most clinically accurate, and affordable solution in the market.
The market is definitely getting crowded and with good alternatives, but we have good things coming 🙂
Can you discuss any feedback from clinicians who have used Freed AI, particularly regarding its ease of integration with existing EHR systems?
Placing the note in the EHR is half of the charting problem and one that we need to solve as well.
Clinicians do tell us that our copy-pasting features are good enough, but they are asking for integrations as well, and it pains me to think that clinicians copy notes from Freed into their EHR 70,000 times every day.
We have our first integration in beta and more coming soon. Please join the waiting list if you’d like to try.
How do you see Freed AI impacting the work-life balance of clinicians in the long term?
I’d like my wife to be Freed from looking at the EHR, never do admin work, and go home when her last patient does.
How does Freed AI plan to expand its services or functionality to meet the evolving needs of healthcare providers?
Imagine the greatest medical assistant in the world. One that understands the clinician, knows every patient and handles every administrative task for the clinicians.
We want Freed to be that assistant, and for every clinician to have one. Or two.
Thank you for the great interview, readers who wish to learn more should visit Freed.
