Agentic Patient 8: Why an AI Founder Won't Trust Chatbots With Her Own Health

[00:00:00] Dear listeners, welcome to Faces of Digital Health, a podcast about how healthcare systems around the world adopt technology and a special series called The Agentic Patient, where we explore how patients use AI and what are the best practices to get the most out of AI as a patient. I am your host, Tjasa Zajc.

[00:00:24] In today's episode, you are going to hear from Elena Ikonomowska, CEO and co-founder of Dia Dia Health.

[00:00:37] If the previous episodes were very focused on the best practices and prompts you can use as patients, this particular episode focuses more on the dangers of AI, the way AI works, and why sometimes it might be better if patients didn't use general purpose AI models for healthcare.

[00:01:05] Elena is a serial founder, AI expert, a TEDx speaker and innovator with nearly two decades of experience applying AI for social good. In this episode, she talked about why lab quote-unquote normal ranges differ by genetics and what that means for your blood work.

[00:01:28] She talked about her own story about a dismissed thyroid and pre-diabetes symptoms and what a two and a half year diagnosis journey actually looks like. She also discussed the risk of self-diagnosing from CHEDGPT-style tools and why she believes AI should strengthen, not replace the doctor-patient relationship.

[00:01:53] Enjoy the show and if you haven't yet, check out our newsletter, which you can find at fodh.substack.com. And if you go to our website, facesofdigitalhealth.com, you can find the Agentic Patient series in the sub-page, Agentic Patient, where you have summaries and discussions around the best practices, also in a written format.

[00:02:20] Now let's dive in today's discussion. Elena, hi and thank you so much for joining me here on Faces of Digital Health and actually in a special series that I call the Agentic Patient,

[00:02:43] where I am trying to explore how people use AI, how people even develop their own companies with the hope to just get more clarity around their health that they didn't get through the healthcare system. And you have such a story where your symptoms or your exams that were done in the medical system weren't taken seriously.

[00:03:11] Seriously, I will let you explain how did you decide to focus your expertise in AI and in data science and you applied that to healthcare. Thank you, Tasha. It's a pleasure to be here. It's such an important topic. I've studied machine learning for two decades now. I've been working in the academy and the industry space, developing dynamic systems.

[00:03:36] And initially I started and did some prototyping at Google, joined Reddit, built these kind of systems at Change.org, as well as all of my previous companies. And so I have always believed that AI technology is one of those superpower forces that can help us solve difficult problems in society. Watching people that I love and watching my mom struggle and going from one specialist to another specialist, not finding answers for her health, for her problem.

[00:04:04] Ultimately, I lost my mom in that battle. And then later on, when I started experiencing struggles with my health, as most founders do, because doing this job is incredibly taxing on the body. And they start developing various kind of mild dysfunctions, which can become bigger problems. And so oftentimes those are the things that are dismissed. Doctors usually run the labs. They see normal labs. They can't explain the symptoms.

[00:04:33] And oftentimes what we hear is that it's really in your head. Nothing's wrong. You're just stressing too much. Maybe just don't stress too much and you will be fine. I just, knowing like what happened to my mom, watching that journey, I just couldn't take that answer anymore. And that's when I decided that I'm going to use my skills, everything that I've known so far, you know, all the knowledge that I've gained from building dynamic systems throughout my life.

[00:05:00] That was my PhD, actually, learning from infinite streams, seeing everything as a complex system, that where things are constantly changing and interacting with each other. That perspective enabled me to think about this problem in a similar way. The idea is the first clinical reasoning, causal reasoning AI that can connect the dots across your blood work biomarkers, your full genetics microbiome data potentially, any other diagnostics,

[00:05:28] and then surface the real root drivers of disease, the real reasons for why something is happening in your body. Not just that showing you what is high, but really what is this upstream or downstream from? Like what are the pathways, what are the explanations that show how has this developed and where is it going in the future?

[00:05:49] And then this information we give to clinicians as a clinical decision support tool to identify and finalize the right protocols for you. The AI does prepare the solutions, the protocols that are recommended for your unique biology. The clinicians can revise them and review them and interrogate them and ensure everything is safe and right. And then work with you on a plan that makes sense for you, for your stage where you are in, your ability to commit to changes.

[00:06:19] And in that way we work on helping prevent small problems from becoming bigger, but also resolving complex cases where current traditional medicine doesn't have the answer to resolve, like doesn't have the tools to actually understand the data, the patient information properly. And what types of data do you need to create any recommendations and what type of recommendations are we talking about?

[00:06:48] So does the tool recommend existing therapies? Does it recommend a change in lifestyle? Let's try to maybe explain this on a concrete example of a patient case. Yeah. At the minimum, we require a comprehensive panel and ideally to full genetic sequencing information.

[00:07:13] The AI will work on the panel as well, and it will find relationships between the various systems that exist and that interact in your body and explain things in that way even better than what you'd normally get by just looking at things in isolation. Okay, your iron is low or your vitamin D here is low, but what is the connection between these two things? And what is the connection between those data points and your TSH value? That's not something that doctors are doing today.

[00:07:41] Medicine is designed to work in like specialties, right? So you have a doctor who is specializing in hormones, a doctor who is specializing just in, let's say, your gut or heart and your skin, etc. But everything is interconnected in the body, and you can't look at things like that in isolation. And if you have the idea today, your doctor can, in essence, take that blood work that they've done, upload it into the system,

[00:08:07] and see the relationships between how small changes in a few of those biomarkers, in a few of those lab results, point to dysfunction in a few of the systems as a whole together because they interact with each other by understanding the relationships between these things. If you add into that genetic information, now you're going to be able to understand why these things are like this.

[00:08:32] Even when things are still in normal ranges, it can explain why you may be having the symptoms and point the doctor to deeper testing. Is that what you used when you were trying to analyze what's happening with you? So basically, how long was your patient journey before you got to any answers? And how do you think that your patient journey would be different today if you already had a tool like this?

[00:09:03] My real patient journey. All of my life, I've always struggled with these minor things that seem unimportant, but the inability to lose weight despite working out, despite eating healthy, despite avoiding sugar, despite fasting, despite doing extreme diets. I've always felt cold. And over time, as things got worse and worse, these signs were there when I was even much younger, around 20.

[00:09:28] And doctors could have seen these signs even then that I have likely thyroid issues that will develop further in the future into bigger problems and even pre-diabetes. But no one was seeing these things. No one was learning about these things. And when things started getting really much more difficult and visible, I was getting more and more frustrated with the system going from one specialist to another specialist, from a gynecologist to a chronologist to another doctor,

[00:09:58] and all of them not connecting the dots, the information. So once I started digging into this on my own, I started realizing that the knowledge is there. We know how genes interact with nutrition, with medicine, with stressors. We have all this information. And so I think it took me two years and a half throughout the entirety of building the idea, really,

[00:10:24] to figure out what is wrong with me step by step, getting deeper and deeper. And now I think I feel like in great health. I've been, I feel so empowered. I feel like I don't need doctors anymore. I can manage my own health. Yeah. Hopefully it stays that way for as long as possible. It really is today, for example, if you just look at five years ago,

[00:10:51] we are in such a different world, especially with AI, with generative AI, with large language models, because suddenly it's very easy for every individual to create a multidisciplinary team of clinicians that will look at a particular symptom from their different angles based on their specialties.

[00:11:16] So how do you see the rise of AI in patient journeys? What kind of prompts, for example, would you use if you didn't have your company, but just had chatbots and you would go to them instead of the doctors to inquire about your symptoms? Now that I know what I know after building the idea, I would not do that actually anymore.

[00:11:46] I would not go to the chatbots and ask them questions. Because of two things. Number one, they've gotten to a stage where they look very convincing. And in fact, they're actually doing something which we call fateful hallucinations, which means that they're making small mistakes. They're not following the instructions properly. They're fabricating little details that may not be true based on patterns that they're basically seeing in their training data.

[00:12:14] So it all sounds very confident and plausible, but you don't have the expertise to truly evaluate if that's true or not. And so you wouldn't know pretty much that they're actually guiding you towards something potentially of wrong conclusion that's for you. Your data does not exist in isolation. You really need to allow it to look at all of the information at once to truly identify what's true for you. I would have asked things like, oh, how is it?

[00:12:42] How can my genetics explain the fact that I have this biomarker value, but my doctor says this is normal kind of thing? And then that's what I would have asked. But then in that explanation that it would give you, some of the things could be possible and some of the things could not be possible. It's going to give you a bunch of things all at once. And you're kind of there wondering what we should be going after, what is real, what is not.

[00:13:09] And the worst thing is that you take that to your doctor and then they're wondering themselves what is real and what is not, trying to figure out. It's very easy to spiral. And when it starts flagging, you have a high risk variant for that or for this. Is that really expressed in your body? What's the context? Is that really a risk that is currently real for you or is just something that could be? It's plausible. These models don't have that kind of information. They're not plugged into the real world.

[00:13:38] They're just generating possible ideas and explanations that may or may not be true. And so that's kind of like the danger that around which we've built the whole, our whole new technology, our whole reasoning, causal reasoning framework to avoid those pitfalls for both patients and doctors. So what kind of technology are you using at the moment with your tools?

[00:14:04] And how do you also position them in comparison to large language models based clinical decision support? Yeah, it's a really great question. I love that everyone should be asking any AI in healthcare. How are you deriving your own conclusions for the doctors or for the patients? We basically don't use generative AI for clinical reasoning.

[00:14:27] We've built our own proprietary and patent-pending clinical causal reasoning engine that pretty much traces the thinking process on a structured graph where we can connect the dots between every piece of evidence and ensure that the evidence leads to the final conclusion in a logical way, just like how a human would logically reason across pieces of evidence.

[00:14:53] But we ensure that connection is made from the beginning till the end of that reasoning chain. Sometimes they converge from multiple pathways coming to the same conclusion. So we use this kind of new technology that will find the, let's put it this way, the thoughts, the ideas with the most evidence based on what information we got from your biology, from the clinical research,

[00:15:18] the health related things that we have learned as a society from 310,000 pre-reviewed research papers, 700,000 and more million basically genetic variants, and then rank them, show you the upstream and downstream effects, the connections, the causal basically relationships within things, and order the problems in the clinical priorities such that if you solve the ones at the top of the hierarchy of problems,

[00:15:45] you're going to have the most positive downstream effects on the ones below, which means that you're going to solve multiple problems at once, so that you don't really work on solving symptoms, you're working on solving the root drivers, the causes for these metabolic dysfunctions, hormonal conversions, things that are part of a system-level dysfunction, a disease in progress.

[00:16:09] You talked a lot about how large language models are very specific, they can hallucinate, they can take you in the wrong direction, however, we know that clinicians use them either officially or in shadow mode, so shadow IT or shadow AI, patients use it, some patients solve their issues by digging deeper,

[00:16:39] importing all their radical data to large language models, adding additional context just to get to some insights. So we did see already that it can be useful when used right. You were mentioning that you are looking at data points and papers that are validated, but what happens when suddenly all the reasoning or increasing amount of reasoning in healthcare

[00:17:09] is managed by AI, meaning that we are starting to rely on AI recommendations, and with this I mean large language models, not types of AI, I just mean chatbots, which basically narrows our points of view, kind of the chatbots are in a way potentially taking over our thinking. What I mean by that is,

[00:17:35] how do you think that might potentially also impact the data sets that you are drawing insights from if the thinking is offload to a chatbot to a degree? Where things will go if we continue, if docs will continue using chatbots and not clinical AI? Yeah. Yes, it's so, I think a lot of wrong conclusions will be drawn over time.

[00:18:03] On one side, yes, people are just throwing in their data and there is no proper learning loops where you're basically figuring out whether that was the right choice for this biology, have you really fixed things? You're still looking probably at things in isolation and kind of making broad conclusions that will apply in some cases, will not apply in other places. It's a dangerous game, pretty much, I'm going to say.

[00:18:29] Right now, models are averaging things. They're not really looking at the uniqueness of your biology and what makes sense for your current situation, especially if you look at all the things all together connected and making broad cookie cutter style protocols like we already have. Medicine will not progress forward. We're just going to have potentially more mistrust generated because of the mistakes that will happen

[00:18:59] and trial and error will remain the same or worse. Doctors might get frustrated even further with these tools not being really helpful without the proper guidance of where their limitations and what is not known. People will be making wrong conclusions and I really worry about that kind of a world. If I just go back to the point where you said

[00:19:24] that at the moment some solutions are not taking into account what's your specific situation and what is normal for you, normal can mean different things for different people. So, for example, normal lab results. If you have a particular condition, a lab result that looks abnormal in someone else

[00:19:52] would seem normal in the person with a condition. For example, if my hemoglobin drops because of IBD, then which is the inflammation of the gut, which is seen with bleedings, internal bleedings in essence, doctors would see it as normal in a way that the hemoglobin dropped. Yes, they would treat it, but it would still be perceived differently

[00:20:22] than in a normal person that suddenly has a drop in hemoglobin. So, how can AI be used to redefine what's optimal and what's normal in different people? Yeah, I think this is one of the biggest problems in medicine right now. That we've been, patients' labs are being read through the lens of a really old methodology that was predominantly actually built on male data to begin with,

[00:20:52] and now it's applied universally to women, to different ages, to different ethnicities, to different genetic profiles. But in reality, because of the uniqueness of our biology, we know today that genes are pretty much affecting where your biomarkers kind of converge for multiple decades. So, your genetic profile identifies what is like sort of a predisposition that will last for you for multiple years and explains why that's happening because of the interactions between everything else.

[00:21:20] And so, what the opportunity that we have that I'm like extremely excited about to be a part of and actually bring into this world is to use technology like AI to really understand what is truly optimal or normal for you that would not be normal for someone else because of understanding how the genetics are actually uniquely affecting your enzyme levels activity

[00:21:48] that basically control how much hormone you convert, how much nutrients you convert, how much of that is available to your body to function properly. And so, a TSH level of 2.something might be completely normal for one person, but in you might actually hide a fact that your thyroid is working kind of hard and trying to like produce enough of hormones

[00:22:17] and your body still doesn't have enough of the free T3 hormones, available hormone that manages your energy, it manages your mitochondria and your health and your hormones. And so, your genetics can be used to identify because of the interactions of things what actually is not normal for you and is hiding a dysfunction and another person means completely something else. That's the power that I see. And over time, we would be able, once we continue collecting this, I call it subclinical data,

[00:22:46] which means this is data on people who are not diagnosed as like with the labels that we have today, you have this kind of a problem, we don't have this information because your symptoms are still mild, you've not yet developed enough of comorbidities and problems to call you into a comprehensive testing. And so, we don't really know what's happening to you until the point you get really sick. Now that people are testing more and more, we have available blood work, hundreds of biomarkers for a few hundred dollars,

[00:23:16] you can actually afford to test at least once a year, ideally twice, and you start to create that subclinical data for what is going on in your body across multiple biomarkers before disease fully develops. That will help us identify together with your genetic profile really the right way of looking at your labs and reading what are the ranges that are for your set of genetics, ideally optimal for you. And that will guide doctors

[00:23:44] what they should be shooting for when they're trying to improve your health. So, where should your TSHB should be maybe closer to one instead of two. For you, for another person, it's actually two is fine. Yeah, yeah, yeah. Yeah, you alluded to the fact that with more accessible testing, we have increasing amount of data. And especially when it comes to women's health, the lack of data, the lack of research is one of the big challenges

[00:24:13] to just help women faster in a way. So, how do you, in essence, see this issue? So, do you think that basically new findings are going to come from the fact that patients are testing more on their own and because commercial providers are going to have increasing amounts of data, new findings will be able to come out of those data sets because public research

[00:24:43] doesn't have enough funding to do and gather all the data that is potentially being gathered through direct-to-consumer products for genetic testing, for example. Yeah. I do think that institutions who are working at this forefront of prevention will probably be collecting the best data because consumers are demanding better health. They're needing better solutions today. They are aware that better health is possible with the help of AI

[00:25:12] that even though it hallucinates, it does create awareness of all the things that could be in play and need to be evaluated and tested. And so, we will see a lot more people testing and, in fact, frankly sharing their data because with the interoperability changes that went in play in the United States specifically, anyone can share their data with whomever they want for the purpose of getting better health with whichever doctor. So, the patient owns their data

[00:25:40] and they will be able to move it around. And so, at some point, I believe it will definitely reach research as well. But initially, companies who are working at the forefront of prevention will have probably the best of their assets. Your company is providing a clinical decision support for clinicians. You are also available to patients. So, from that perspective,

[00:26:07] how do you see the changing relationship between clinicians and patients with the changing access to data, with the changing understanding of what the data means with the help of... So, how do you see that relationship might change in the future because of these changing dynamics? Yeah. Yeah, no, I definitely think that the... I'd say a future where doctor-patient relationships

[00:26:36] are actually strengthened by AI and not weakened. That's why we actually have a patient... We are available to patients directly because when doctors and patients become more empowered with information, both sides, however, there is the potential, the best version of the world of the future where the AI handles the computational complexity, avoids the mistakes and the pitfalls, is able to analyze multi-variant... perform multi-variant analysis across all of your biological information

[00:27:04] that doesn't take a huge amount of time and it's basically not posing unrealistic strains on the existing system. Both sides are empowered to make the right choices together and build a partnership towards best care possible. But right now, the reality is that many clinical clinicians, both on both sides, are actually frustrating. patients feel unheard because of just a 15-minute time slot that usually doctors are able to give to them

[00:27:33] and they bring in their data and complexity of symptoms and the clinician feels overwhelmed with all this information. They've not run it through an AI or not sure they can trust the analysis and even this AI that they have is kind of not also reliable. And so it feels really a bit of a... It's a frustrating relationship and situation for both sides. But a technology like ours, I believe, can change that dynamic because now the clinician can walk in with a... Both sides are empowered basically with a tool.

[00:28:03] The clinician walks in with a technology that has already done a comprehensive evidence-based analysis on everything that's going on in that patient's body. The patient is also informed. So now they have more time to do the things that only humans can do, which is listen, show empathy, apply judgment, co-create a plan together that makes sense for the person's life. Because oftentimes where treatments fail is, okay, we know what to do, but the patients are not doing these things.

[00:28:32] And then the reason they're not doing these things is because they don't... It wasn't realistic the way that it was set up for their life. It wasn't well explained the importance of it. It wasn't adjusted. It wasn't working really well initially. So I only believe in a great future, actually. AI will be our best evolutionary partner to handle the complexity of human biology and disease and actually create a new kind of medicine that we don't have today. Yeah, yeah.

[00:29:01] The potential is definitely there. I think there's a lot of excitement on both ends. I know that you mentioned that people need to be cautious when they interact with chatbots, but still they are being used. So from that perspective, if you had advice for people in terms of how to use chatbots that are out there

[00:29:30] or AI more broadly, so AI that is accessible, what would your advice be? What should people definitely not do? What could they potentially do? What comes to mind when you think of all the technology that's out there? They should not be throwing their genetic data into chatbots and other open AI. I would say start there first. You want to know and ask every platform, who has access to my data? What do you do with my data?

[00:30:00] Is my data being used to train the models and how? Would it be sold or shared? Are you HIPAA compliant? Are your business partners HIPAA compliant? Do you have BAA agreements with all of your partners? Genetic data is very sensitive. There's so much power in it, honestly. And I really think everyone should test and they should use it in understanding their health and especially like interpreting their biomarkers at the minimum. So just looking at genetic data on its own is not healthy.

[00:30:27] You don't want to worry about the gazillions of risks that you will find because we have a lot of genetic variants with high-risk variants that have certain potential influence, but it's the interplay of genetics and the environment, the stressors and toxins and disconnect systems that interact with each other that determines what sort of risks you really have, not on their own. When you throw data into chat-GPT style or open AI models, you are allowing that data

[00:30:57] to be potentially leaked into places you don't want it to be leaked. To begin with, insurance companies can use this information to change what you might get as a protection, as insurance, at the rates, how sick or dangerously sick they perceive you are. And you have no control once it goes out where it's going to go and how it's going to be used and what kind of conclusions will be made for you. And you want to be careful about that. Patient privacy for us is really fundamental to everything we do.

[00:31:26] Our business models are not designed around data selling or sharing. If something is free, I always say, you are the product. So think about that. So basically, they're selling you, it's not free. Absolutely. It is, I think, a topic that we will need to talk even more about because at the end of the day, it comes back to when are patients

[00:31:54] willing to share their data. And oftentimes, when you have a big, big medical issue and when you are under stress, when you are anxious or worried, then the guards that you have on for data protection lower to a degree. Or when you ask patients would they share the data

[00:32:24] to help other patients, then they're also more open to data sharing. And especially with AI being very convincing, very pleasing, very... Falsely confident. Yeah. Yeah, exactly. It's also... People can also... I won't say that they are manipulated into sharing data, but it's definitely easier to let them know, to convince them to share the data. So yeah, definitely, I guess privacy is going to be a big thing to discuss

[00:32:54] in the future. Yes, for sure. And I honestly, when I think and worry about the fact that these open models, they're already growing insatisfaction of patients with a care system, which is putting them to share their data more broadly, are going to create a few more other problems. Besides the fact that self-treating inappropriately, people basically, because of this confidence that models show, they will choose to take certain treatments on their own and I've done it myself.

[00:33:23] So I know it's not like wild to think that's going to happen. Another thing that's going to happen is anxiety amplification. You're just going to start looking at things and start worrying too much excessively and because of not really understanding and taking things out of context. We're just going to have more anxiety about our health, which may not be really realistic. And the third most important thing is the equity dimension. I think the people with more resources and more education and more finances will be able to get more out of these tools

[00:33:53] because they'll be not using the free versions. They'll be using versions where things are properly handled and they'll be using doctors who are using AI that is not free. And that will only enlarge the existing disparities between people with resources and without resources. And that's why I really believe that clinicians should be at the center of all of this. They should be involved in using a technology that is safe, clinically validated, that is a powerful reasoning tool

[00:34:23] that gives them more information and a way to learn themselves and a way to safely guide people into, let's put it this way, new territory because we've not tried some of these things on a large scale but we should because there is a huge potential and evidence already that it will work but it has to be done with a clinician at the center of everything and I do think that the human relationship in medicine is not a bug, it's a feature, it's something we need to have. Yeah, yeah, absolutely. So how many clinicians

[00:34:52] do you work with with your solution? Can you tell me a little bit more about how you went from the idea to product development to testing to deployment We had a very unusual path to building this whole product so I, because of my personal health story because I was the patient initially, I was the consumer that was trying to find answers and trying to use this open AI to find solutions for myself, I realized that how bad care is and had spoken

[00:35:22] with many doctors and the central narrative is that you want to create demand on the patient side and you want to basically raise the bar of what quality is and solve these solutions by ensuring that the patients are the most satisfied first. They're really finding the reports, the insights, solutions really helpful and will feel more informed and their concerns will be explained. So we actually went from, initially from

[00:35:52] a direct-to-consumer side of things. We partnered with a few clinicians with who we are working with and we've had done thousands of cases of patients who shared their data with us and we designed the product alongside that sort of collaboration running multiple clinical iterations ensuring that we are surfacing the right kind of things, we are prioritizing things properly and we are actually seeing improvements in results in patients.

[00:36:20] So now that we've validated this, we've reached a 98% of agreement with clinical judgment of the reports and we're seeing in early data 60% less trial and error in resolving symptoms. So we've been able to solve things that patients were trying to do for multiple years with their clinicians. When they came to us and we gave them the report they took them back to their clinicians and they made changes based on those reports and things started immediately changing and

[00:36:50] shifting for them. Things that weren't diagnosed for decades where all of a sudden, wow, I can see that you actually have endometriosis and we need to do these things for you. That was a very unusual process and then we discovered that there's all these things that I've learned which is that people when they start seeing a very complex information around their genetics and their body surface to them they start to spiral, they start to feel more anxious and we also understood that there's so much potential in understanding human health

[00:37:20] that hasn't been done yet and needs to be in the hands of clinicians and so we open it up to everyone now. We've been on the clinical side of things live for a few months, we've had really great adoption, doctors really love the fact that it's transparent and easy to use and so we've been like, we already have about 15 clinics to 20 in process with us, trialing the technology, using it with real patients patients and we're basically

[00:37:48] adding a few every week. Congratulations. Where do you hope to have the biggest impact? Where do I hope? We already are seeing impact. Which area? I'm just asking because you mentioned endometriosis so I assume it's women's health. It's a very horizontal platform. We can solve, we're seeing impact across autoimmune issues, across hormone line balances, across metabolic things, problems which have a strong genetic component, they are

[00:38:18] everywhere. Even fertility is actually an aspect that both in men and women, frankly, is a thing that is on decline and understanding the impact of genetics, of microplastics, of inflammation that has been ongoing is something that's going to help all these problems. Very difficult to resolve, complex cases, as well as the possibility to get ahead of disease for everyone, for men and women, are all areas where this technology actually helps.

[00:38:48] You've been listening to Faces of Digital Health, a proud member of the Health Podcast Network. If you enjoyed the show, do leave a rating or review wherever you get your podcast, subscribe to the show, or follow us on LinkedIn. Additionally, check out our newsletter, you can find it at fodh.substack.com, that's fodh.substack.com. Stay tuned!