After securing $3 million in seed funding led by Canvas Ventures, startup Parakeet Well being has launched a generative AI voice platform targeted on affected person engagement. Two of the San Francisco-based firm’s executives lately spoke with Healthcare Innovation about what’s concerned in bringing generative AI to doctor apply name facilities.
CEO Jung Park beforehand held government positions at One Medical and Epocrates, and co-founder Eric Mao led product and engineering groups at Microsoft and Primer AI.
Healthcare Innovation: Why did you resolve to focus your generative AI efforts on the affected person contact facet of healthcare?
Park: Folks might imagine specializing in utilizing AI to unravel administrative challenges shouldn’t be that horny, proper? However when you consider it, deploying AI on the scientific facet has challenges of its personal. One of many issues that I realized after I labored at Epocrates was do not change the workflow. So if you happen to really attempt to do one thing that takes over what a doctor does, chances are you’ll not essentially get the adoption that you just anticipate.
HCI: What are a few of the challenges doctor teams face with managing affected person contacts and calls?
Park: It is one thing that I feel all of us can relate to. It isn’t unusual to attend 5 minutes earlier than a name is answered. Nobody desires to be placed on maintain. Everybody’s time is effective, and if you happen to do not deal with these points, sufferers really feel like they don’t seem to be valued. Additionally, for a lot of medical practices, staffing is the largest problem. There’s turnover. Labor prices hold rising, and for the people who find themselves really addressing plenty of the assist points, the idea of empathy fatigue is actual. It is solely so typically that you would be able to even have a smile in your face, and do that consistently, day in, day trip. In some unspecified time in the future folks burn out. That is a problem that all of us face as healthcare directors.
One other problem, I might say, is income leakage. Operating a medical group, you could have a lot of alternatives to generate income, but additionally there are plenty of alternatives to lose income. Let me offer you an instance: When you could have a late cancelation, which means the affected person who is meant to return in and take that supplier’s time slot does not present up. So it creates these gaps within the schedule, which leads to decrease income.
When a specialty apply receives a referral from a major care apply, it is advisable to convert this referral, as a result of they’re like a golden ticket, proper? If a health care provider says to you, I might like so that you can go see this dermatologist about this rash, then you definitely go. Nevertheless it seems folks do not at all times go. So whenever you’re really capable of convert all these referrals, that is actually the best way that you will generate income.
HCI: So how does Parakeet work to deal with these challenges?
Park: To start with, it improves the affected person expertise by answering each name 24/7, with zero wait time. Parakeet additionally automates repetitive duties like scheduling and answering often requested questions, in order that the workers, as a substitute of truly doing these repetitive duties, can deal with r extra precious issues that sufferers want, like checking in, testing and actually attending to sufferers’ want after they’re really within the bodily workplace.
From a income leakage perspective, Parakeet really makes outbound calls. So if somebody cancels, Parakeet can really name the subsequent one who’s within the wait record, or possibly there is a affected person two weeks out that is likely to be out there to return in and fill in that hole in that schedule. That is a method to really bolster the supplier utilization and ensure the income is captured for the healthcare group.
HCI: Over time, most individuals in all probability have expertise with a chatbot on a retail or airline web site, and so they typically get pissed off with it as a result of it is not capable of reply their query. However has the know-how improved exponentially now?
Mao: I have been in AI for a really very long time. I do not suppose it is prepared to exchange medical doctors, however we see firsthand that it now can tackle these difficult scheduling questions, and I feel it is prepared for prime time for this set of use circumstances. The transcription high quality now’s so significantly better now with this wave of huge language fashions. And the speech appears like a human, which is a really latest factor. It is within the final two years that we have had this.
HCI: Does your system need to combine with the apply’s EHR, billing system, buyer relationship administration system, to ensure that this to work?
Mao: We spent the previous 12 months constructing out all this infrastructure to do this. We combine together with your EHR, together with your telephony system, together with your CRM ticketing system — all these issues to allow that very pure expertise. So whenever you ask for an appointment, it seems it up in actual time, it provides you real-time availabilities, after which we guide it immediately into the system so there is no human interplay wanted all through that entire course of.
HCI: Is that course of difficult or do all these techniques now have APIs that make it go fairly easily?
Mao: You’ll hope so. However what we have realized is plenty of these EHRs shouldn’t have excellent APIs, in all probability on function typically, as a result of they need you to do the whole lot on their platform. However we spent plenty of engineering effort to make this reusable. We’ve made it simple to onboard the subsequent buyer, as a result of we have performed all of the heavy legwork already.
HCI: It seems like just a few of the shoppers that you’ve got introduced are digital themselves. Have you ever focused these sorts of companies initially?
Mao: We’re focusing on medical practices that usually have a name heart. So we’re not focusing on mother and pops. When you could have a name heart with 10 or extra folks, you are going to see extra price financial savings and extra use out of what our platform does. We now have dermatology; we’re in talks with OB/GYN and first care.
HCI: So wouldn’t it make sense for a big well being system like Sutter Well being or Kaiser Permanente?
Mao: We will goal the Sutters afterward. We’re focusing on outpatient first, however that is completely in our roadmap deal with the Sutters, the UCSFs, the college well being techniques.
Park: We’re already seeing that our resolution has hit a nerve. Whether or not it is these of us who’re really working extra on the normal facet or the digital firms, on the finish of the day, the chance to work together with sufferers is an actual factor. Ninety % of all of us nonetheless use the telephone to make an appointment, in order that signifies that there’s a chance throughout the board to actually have interaction with sufferers.
HCI: Are conventional distributors within the affected person engagement house making an attempt to carry AI options to market as effectively?
Mao: You have got incumbents within the affected person engagement house, however they’re utilizing outdated know-how. They’re making an attempt to steer this big cruise ship to make use of LLMs, whereas we’re LLM-native. We began on day one with this know-how, so our know-how works higher. Ultimately, the incumbents are going to attempt to undertake this new know-how, however we expect we are able to transfer quicker, and our know-how is already right here utilizing these issues.
Additionally the competitor actually is offshore name facilities. After we speak to clients, after they have challenges of too many missed calls, they’re taking a look at increasing extra folks through offshoring to unravel that problem. They are not even taking a look at know-how. So after we present them what we have now, they’re like, ‘Wow, I did not even know this was attainable right this moment.’