Guillaume de Zwirek, CEO: Hello! My name is Guillaume de Zwirek and I’m the co-founder and CEO of Artera. I’m joined here by Ashu Agte, our chief technology officer. Ashu, we’re here at Heartbeat. We just introduced two new AI co-pilots. Can you tell us a little bit more about the Staff AI Co-Pilot?
Ashu Agte, CTO: Happy to. I’m so excited about the AI journey at Artera. So let me talk about the Staff Co-Pilot that we just launched. The idea behind Staff AI co-pilot is that we want to provide skills to staff users so they can be more efficient, they can make decisions better, and they can operate in a faster capacity. So we introduced four skills. The first skill was if an inbound message is coming in a non-English language, you can translate it in place, and same thing goes for the outbound message – a message needs to go out to the patient, you can type it in English and translate it into 98 different languages in an instant.
Guillaume de Zwirek, CEO: Well, this is huge because we have tens of thousands of agents on our software every single day interacting in 98 languages.
Ashu Agte, CTO: Yeah, and some of the stories we’ve heard from health systems is that they have to have relationships with translation companies and there is a delay in figuring out what the request is, what the response might be, and that causes delays in the communication, and this is instant, in real-time – there are no services required. Incoming messages and outgoing messages can be super helpful with translation.
Guillaume de Zwirek, CEO: And when you think about some of the use cases that we heard at the conference today, we just had one around a patient who messaged in about suicidal ideation, and those aren’t situations where you can afford to wait for translation services. You need to know what the patient is saying right away and give them passionate care and direct them to the right resources.
Ashu Agte, CTO: And if you look at the big metros, they’re already multilingual in all these cases, so many of the big enterprise health systems that are out there, they require almost language supports by design, like English and Spanish given. But there are 8 other languages depending on the population they have to support by default. So this translation tool is going to be a game changer for them.
Guillaume de Zwirek, CEO: Now I know it has 3 other skills. But staying on translation for a second…in healthcare, there are always concerns around security and privacy and also accuracy. So, can you tell us a little bit more about how we built this capability?
Ashu Agte, CTO: Yeah, so we built this capability based on an LLM model, which we trained it using a lot of healthcare data and when we trained that using healthcare data, it became more accurate than just the Internet information that LLMS use as their default data set. So we take that data that LLs bring from the Internet, and we infuse it with the healthcare data and the conversation that most of the staff users are having.
Guillaume de Zwirek, CEO: Presumably we have billions of billions and billions of data sets that we can
Ashu Agte, CTO: Over the last 9 years, we have so much data set that we have tagged with different types of conversations and turned it into training data that off the shelf and then mixing all the data we have, the accuracy of these translations is way higher than an average off-the-shelf solution and it is trained in sort of the dictionary and vocabulary of healthcare information, so that the translation, the context, is very nuanced and fine-tuned for healthcare conversations.
Guillaume de Zwirek, CEO: Now let’s talk about the other aspect of data, which is data residency. So presumably we’re not sending this out to Google Translate, right? But this is a self-hosted LLM – data staying contained within our firewall system.
Ashu Agte, CTO: Yeah, the LLM technology is now decently commoditized. So there are off-the-shelf open-source models. What we have done is we are using a model, open-source model, and we are hosting it – we are not sending this information to Openai. We are not sending it to Google. We are not sending it to Aws. This is a hosted model we have, so that we can meet all the compliance needs around HIPAA and HiTrust for this particular feature.
Guillaume de Zwirek, CEO: Wonderful. Tell me about the other 3 skills.
Ashu Agte, CTO: Yeah, happy to. So the second skill is a lot of time when staff users are trying to respond to the patients, they’re typing a longer message. Sometimes those text messages have to be split into two different ones because there’s a word limit. So it’s always better to send a very concise and short message out to the patients. So we have created a second skill: when the staff user types the message, they can ask this AI to make that message concise, so the concise message will not get split – also, a lot of times a shorter message gets a better interaction from customers because they don’t have to read a whole thesis on the reply.
Guillaume de Zwirek, CEO: What’s that quote? If I had more time, I would have written a shorter book.
Ashu Agte, CTO: Yeah, yeah, that’s right. That’s right. Simplicity is better. The 3rd skill that we’ve introduced is an autocomplete skill. So when the staff user is doing or is typing a response, we make predictive text to make sure to complete the sentences, so they can just accept those changes on very frequently used sentences.
Guillaume de Zwirek, CEO: So like Gmail or Outlook.
Ashu Agte, CTO: pretty much.
Guillaume de Zwirek, CEO: And presumably trained on healthcare data, so it’s not…
Ashu Agte, CTO: That’s the beauty of this. Most of these responses are trained on healthcare data, so they are actually completing with the right context as compared to just using Google’s or any other off-the-shelf autocomplete features.
Guillaume de Zwirek, CEO: Well, and we have such a unique advantage here that we can offer to all of our customers, which is just our scale, right? We have a hundred 1 million patients who are engaging with our software every year, tens of thousands of staff members doing conversational for years, right? Since the infancy of this company. So we have those records of what happens when staff members and patients, all sorts of staff members, clinicians, nurses, you know, front desk personnel, revcycle…
Ashu Agte, CTO: Absolutely.
Guillaume de Zwirek, CEO: Alright, last skill.
Ashu Agte, CTO: Yeah, the last skill is a very interesting skill. So what happens is, there are times when an agent goes into a conversation with a patient and they want to look at sort of the conversation history, or they want to look back at the previous conversation. Another use case that we hear from our customers is that if there is a conversation happening. which is a little bit clinical in nature, they would prefer to also copy-paste that conversation in their EHR or the CRM system. So we have created the fourth skill called Summarize. So what staff users can do is they can go to the patient conversation, they can select the summarize skill, they can look at the date range, and they can select all or subset of messages and it will provide a summary of the whole conversation, a concise summary of the whole conversation that they can quickly read and catch up on the context, or if they want to copy, paste that into the EHR and the CRM systems, there is an easy button to send that information to the Notes Section of their CRM or the EHR System. So that’s going to help a lot on not having two sources of truth for the health systems.
Guillaume de Zwirek, CEO: If I’m not mistaken, we are working on getting that button to auto-post to the EHR, some of the work that we’re doing with Care Navigator.
Ashu Agte, CTO: That’s on our roadmap and the bidirectional communication between the EHR system is our highest priority at this point, so that we can have that automatically sent to EHRs.
Guillaume de Zwirek, CEO: So that was a staff AI copilot. But I understand there’s also an Insights AI Co-Pilot. So tell us more about that.
Ashu Agte, CTO: So the Staff AI copilot is supposed to superpower and provide a lot of skills that will help the staff user. The insights AI Co-Pilot is going to be behind the scenes, providing insights into operational improvements of Artera’s system as a whole. So we are launching two skills. The first skill is called a Smart No-Show Predictor. So what we have done behind the scenes is, we have looked at patients and their behavior: how many times they reschedule, how many times they have unconfirmed appointments and they don’t show, how many times they have confirmed appointments, and they don’t show. So with that behavior, we have come up with an AI model that flags a certain amount of patients as high-risk no-shows. And we are going to create sort of three different things with this particular skill. The 1st one is inside the inbox, we are going to start flagging patients who are at a higher risk of not showing up. The second thing we are doing is we are going to generate reports for predictions of high-risk, no-shows as well as validation on what happened to those high-risk no-shows, so that health systems can contrast and compare how their actions impacted these. And then the 3rd one is, we are going to take automatically these high-risk no-shows and provide ability to create triggers that will set up a different type of conversation to make sure that these high-risk no-show patients get additional communication to show up to their appointment.
Guillaume de Zwirek, CEO: So this is magical. If our machine learning model flags somebody at a high risk of not attending their appointment, we can actually automatically put them on a completely different pathway based on that flag. Maybe that patient would get a communication that says, Hey, just as a gentle reminder, we have a $20 no-show fee, and that will drive them to adhere to their appointment.
Ashu Agte, CTO: Yeah, or even say something like you missed your last wellness visit, it’s important, you know, it’s better for your health – so something encouraging as well, but it can create a unique workflow for them, and then that will get them to either contact, reschedule, or show up to the appointment – and both of these outcomes are really necessary for health systems and patients.
Guillaume de Zwirek, CEO: Second skill.
Ashu Agte, CTO: Yeah, the second skill is called next best actions. And then what this skill is doing is this Insights AI Co-Pilot is behind the scenes doing anomaly detection. So in the normal course of events for any health system, messages are going out, things are happening, but when something out of the ordinary happens, this insights AI Co-Pilot is flagging them, recording them, and then providing next best actions in a tile, and then it’s also recommending what you can do about it, and not just flagging them. So, for example, if a patient is getting tons of messages, 90 to 100 messages in a day, that tile in the new inbox will show that the patient is getting over-messaged, and they can click on the view patient, and they can look at the communication history. In some cases, communication might have been necessary – maybe something was going on and that communication was required, but in a lot of cases staff users can look at those conversations and figure out something is not going well. So those kind of situations, in order to improve the patient experience, to improve the operability of Artera’s platform, the next best action is going to be a game-changer.
Guillaume de Zwirek, CEO: And you know what’s so special about that is, it’s providing insight not only into Artera traffic, but if our customers and if our partners are pulling in data from other systems, other communications data – take the call center system, their reputation management system, their billing system. Any other system that’s communicating the Ehr, right? The portal communications, we’ve actually started flagging errors in other systems, right? We’ve seen rogue events where a surgical communication tool is texting a patient incessantly 80 times in a row. And that’s horrible for patients. Today, nobody knows. So I think part of the power, and why I’m so excited about what your team has built, is that we’re not just looking within the four walls of Artera. Everything we’re building looks at the universe of interactions that folks are putting into Artera.
Ashu Agte, CTO: Absolutely. The 360-degree view of patient communication is going to be a game changer with this because then this Insights AI Co-Pilot is looking at every single message going out across the ecosystem, and that’s out there.
Guillaume de Zwirek, CEO: Well, thank you, Ashu. There you have it: two new AI co-pilots from Artera. The first is our staff AI Co-Pilot. 100X, your employees – a little bit of hyperbole, 20%, 30% better. What do you think?
Ashu Agte, CTO: I want to say 50%.
Guillaume de Zwirek: Oh, wow! Okay, we heard it from the CTO. So it must be true. And our second AI co-pilot is our Insights AI Co-Pilot, which helps you optimize the use of your system, figure out those erroneous situations, those needles in a haystack that you just don’t want your patients experiencing. Thank you so much, Ashu.
Ashu Agte, CTO: Thank you, Gui.
