Healthcare Data Readiness and AI Adoption: Why 85% of Organizations Aren't Prepared

Chris Hutchins: In a world full of noise, we're tuning in to what matters. Welcome to the Signal Room, a podcast where leadership, ethics, and innovation are buzzwords. They're blueprints. I'm Chris Hutchins, former Chief Data Analytics Officer and host of the show.

Ratnadeep Bhattacharjee: And we're speaking the truth. You're all the right noise in the signal room.

Chris Hutchins: Today I'm delighted to welcome Ratnadeep Bhattacharjee to the Signal Room Podcast. Ratnadeep is the co-founder and managing director of Tech Variable, the leading healthcare IT firm specializing in AI-driven solutions, data integration, and digital transformation. With over a decade of experience as a techno functional consultant, he's at the forefront of making organizations AI ready, particularly in the U.S. healthcare sector, where his team develops innovative tools like SyncMesh, which you'll hear about a little bit today, for seamless data interoperability and predictive analytics. Ratnadeep's thought leadership shines through his podcasts, his articles, and insights on demystifying AI myths, ethical data strategies, and building data mature cultures perfectly aligning with our signal frequencies like the Enabler's Edge, which you'll learn about in coming episodes. Today we'll dive into data readiness for AI adoption, ethical compliance and healthcare tech, and how his servant leadership empowers teams to gain a competitive edge. Ratnadeep, thanks for joining us. Let's unpack the signal shaping the future of data and health. Sure, sure, Chris, let them follow up. As the co-founder and COO of Tech Variable, you've built a reputation for servant leadership, mentoring teams on essential AI skills like Python and Pandas while empowering non-tech stakeholders. How do you cultivate this enabler's edge to help organizations turn data challenges into competitive advantages in healthcare?

Ratnadeep Bhattacharjee: Yeah, uh first of all, thank you for that warm introduction, Chris. Uh you know, when I think about this idea of an enabler's edge, as you rightly put it, right? I don't I don't necessarily see it as a title or a skill set. Uh it's more of a mindset, right? And and for me, that mindset really comes from being a non-technical founder in a very, very technical world, right? So uh when you don't when you don't come from a hardcore coding background, you'll you learn early on that your greatest strength is not in having all the answers, but in asking the right questions. That perspective has shaped, yeah. That perspective has shaped how I lead. I don't I don't walk into a room trying to prove I'm the smartest, I mean, smartest guy or an engineer there. I walk in, you know, curious, eager to understand what challenges people are facing, and then I help connect the dots between technology and real world impact. You know, in in healthcare especially, this has been a gift. Because if you think about it, most of the stakeholders we work with, whether they are care coordinators, you know, stakeholders from the payer systems, executives at provider organizations, they are not really deeply technical either, right? They they care about outcomes, about how quickly data can help them identify, you know, let's say care gaps or how an AI tool AI tool might reduce administrative burden. Being a non-tech founder allows me to naturally speak the language, you know, if you if you know what I mean. I can translate between the technical depth of my team and the tactical needs of the clients, and that bridge building is often where the real magic happens. Now, you know, within our own teams, I do spend time mentoring people, not because I'm the deepest expert, but because I want them to see how different skills connect to outcomes, right? How cleaning up you know messy claims data can improve probably hideous reporting, or how building a good pipeline can make sure a Medicaid patient doesn't slip through the cracks. When people understand the why behind their technical skills, it it kind of sticks, right? And that's the non-tech perspective I bring. I constantly, you know, anchor technical work in in human purpose, right? So for me, cultivating this uh enabler's edge is about you know about listening more than I speak, you know, simplifying where others complicate, probably, and keeping the focus on people. That's what builds trust with teams, clients, and and the entire ecosystem, right?

Chris Hutchins: I love your approach to this. Listening more than you're talking. I couldn't agree with you more. But one of the interesting things I've encountered over the over the years is this interesting tendency that the the technical teams and the project managers seem to want to explain the methodology. We're gonna talk about agile and scrum and a few other things when our stakeholders and our business partners really care about when are we gonna get what we've asked you to get for us. And the goofy things that sometimes we get hung up on are not the things that are important. It really is about what it is that we're we're doing for the for the business and understanding the why so that we can actually connect with some purpose and deliver on what they're looking for. Absolutely. So let's talk a little bit more and get into the readiness aspects of this because what you've just talked about is certainly the very beginnings of that. So for diving into data readiness for AI adoption, in some of your writings and your podcasts, I've heard you talk about the fact that 80 to 85 percent of organizations aren't prepared due to silo data. Can you walk us through a practical data audit process and share how it separates real AI potential from hype?

Ratnadeep Bhattacharjee: No, that's a that's a that's a good question, Chris. I you know, as you rightly put, I often say that the real bottleneck in AI isn't the model or the algorithm, right? It's it's really the data. And the reality, as you rightly put, especially even in healthcare, right? 80 to 85% of organizations simply aren't ready now. The data is siloed, you know, across EHRs, across claims data, you know, even spreadsheets. And unless you solve for that, AI ends up being more hype than help. Now, now I'm a I'm I may not be a deeply technical founder, but that's actually helped me design a very simple structured way about a data audit. I look at it in in four lenses and and you know, I I try to explain it in plain language so that both uh technical and non-technical stakeholders can can engage. The first lens is completeness. You know, do you even have all the data you need? For example, if you're trying to build a care gap analytics model, but you only have claims data and no clinical notes, right? You are already flying half blind. Second is consistency, even if the data is there, is it structured and standardized? I've seen organizations with 20 different ways of recording blood pressure or addresses that don't match across systems. In AI, especially in AI, actually, inconsistent data leads to inconsistent predictions. The third is connectivity. Are your data sources talking to each other? A patient's a patient's lab result might sit in one system, the prescription in another, social determinants in a third. Unless you connect these dots, AI can't really generate a view or the longitudinal view that really drives impact, right? Right. And the obvious one would be compliance, right? In in healthcare, this is non-negotiable. Is the data being used, you know, in line with uh HIPAA, CMS mandates, or interoperability standards like FHIR. Uh AI can't just be accurate, right? It has to be safe, it has to be ethical. Let me give you a quick example, Chris. You know, we we recently we recently worked with a small health plan serving Medicaid communities. They were very excited about AI and wanted to build predictive models for care gap closure, right? But when we did the audit, we found the real issue wasn't the AI, right? It was that it was that their claims data and EHR feeds weren't even talking to each other. You know, members would show up as non-compliant with screenings, you know, simply because the screenings done at community clinics weren't reflected in the claims data. By by you know, by fixing that connectivity and consistency issue, that they immediately improved the accuracy of their quality measures reporting, you know, without without even touching a single AI model, you see.

Chris Hutchins: It's interesting. What you're what you're describing, I think, happens over and over again. And I think your approach is is so important because having done these things before, you have an idea on what might be missing. Oftentimes your business partners might not even have an idea. There's something out there they haven't seen, they don't know about it.

Ratnadeep Bhattacharjee: Absolutely. Absolutely. You know, that's the reason you know only after that foundation was in place did it make sense to even talk about advanced analytics, right? You know, Chris, that's why I believe a data audit is so important, right? It separates the hype, you know. Yes. Hype, like let's deploy a generative AI chatbot tomorrow, you know, from the reality. This is not a reality. We need to first, you know, unify our patient data, ensure quality, and only then layer AI on top, right? Even only if it's really important. The organizations that succeed are the ones that treat data readiness as a journey, you know, not really as a checkbox. They invest, they invest probably in getting the foundations right. And once that foundation is in place, AI AI will move or may move from being a flashy experiment to being a strategic asset, right?

Chris Hutchins: Yeah, I I love how you're how you're framing it. And I think the the deliberate and methodical approach that you're taking is so important for people to hear. So I know you and I have had some exchanges on social media along with some other folks around this hidden cost that's causing really some significant challenges. It's really in this foundational layer where on top of everything, people think about technical debt because we invested in all these fantastic technologies, but there's a layer beneath that people haven't realized that if they haven't spent time securing that foundation and connecting all these silos, the technology is not going to do what they think it's going to do. And essentially, they might have just a really cool tech stack and some additional storage capacity, but no meaningful way to meet the objectives. So this kind of leads into another area that I think you guys have done a phenomenal job at designing meaningful tools, not for the sake of technology, but things that actually have purpose to really drive forward uh the uh objectives that you come to a consensus with uh when you set up set out to work with your clients. Real-world tools. You have your own platform called SyncMesh. It stands out as tech variables platform for healthcare data integration. How does it enable features like a patient 360 view, claims automation like you've mentioned, or how does it make enterprises AI ready? And what's a standout use case study like predicting claims rejections that shows impacts on value-based care?

Ratnadeep Bhattacharjee: Okay, I I'm I'm glad that you uh brought that up, uh Chris, because uh SyncMesh is a great example of how we approach healthcare data integration at tech variable. You know, but let me start off by clarifying something important. You know, we don't position sync mesh as a standalone product in the market, we rather position it as an accelerator, essentially, you know, a set of pre-built frameworks, you know, connectors, parsers, harmonizers that make our that make our consulting and services delivery far more efficient and outcome driven, right? In healthcare, the number one challenge is fragmentation. You you know, you have got patient data scattered across different systems, EHR systems, claims data, you know, pharmacies, even devices, right? And when the data doesn't flow together, it's it's nearly impossible to create what people call the patient 360 view. What syncmesh kind of does is give us a head start in bringing all these pieces together. It accelerates the work of harmonizing data into a common schema, layering in interoperability standards like you know, FHIR and creating a kind of a longitudinal view that's both actionable and compliant. Now, once you have that kind of connected data backbone, a lot of things become possible. And this is where Syncmesh really shines, I believe. It takes care of the integration problem, it gives us a clean, unified, standard-based foundation of healthcare data. On top of that, our teams can build whatever the organization truly needs, right? A data warehouse, data lake, care gap analytics solution, quality measures engine, or even ML models, you know, for cohort creation for personalized intervention, right? In so, in other words, sync mesh doesn't try to be the end solution, clears the path so that higher value use cases can be built faster, you know, more reliably, in a way that organizations genuinely become AI ready. Now, one of the important things to consider here, Chris, is sync mesh isn't the only arrow in our quiver, right? Right. We have built other accelerators to, you know, pre-built data models for stand you know for standing up a healthcare data warehouse quickly, right? And no code ETL frameworks, fast for faster data engineering, even a conversational Gen AI layer that sits on top of uh sits on top of these warehouses and data lakes to make complex data more accessible to clinicians and business users. You know, together these accelerators mean we are not reinventing the wheel every time. You know, we are solving for speed, reliability, and real-world impact. Right. I mean, I mean one I mean a very recent case study would be something that we did for a medical device company. You know, they they they were collecting rich device generated data on patients, but it was siloed. EHR data set in one place, stream data in another. None of it was being combined with the device data, right? So we deployed sync mesh as an integration layer, and then we unified those streams into a single patient record. So I mean I mean that meant that care teams could see not just how the device was performing really, but how it got tied to actual patient outcomes and cost, right? So that is so we could identify patients who who were at higher risk and where interventions are really needed and predict care gaps as well, right? So that's the kind of transformation I'm excited about when you talk about sync mesh. You know, we we don't like to position it as something flashy, but something that really enables to become organizations AI ready in a very, very grounded way.

Chris Hutchins: This is such an important thing for for people to understand for our executive friends that are that are out there and inherit this conversation this morning. These foundational things, they're not the most visible parts of of what we do, but it is absolutely the one of the most critical and essential things that we can do. And when we put these kind of capabilities in the hands of our teams, it puts a significant amount of capability in front of them to accelerate and their accuracy and their what's the word, the the evolution and the way that it improves from the very first day you turn it on, it only gets better every time you use it. I think that's a really important thing for people to understand. There is a price of entry for these types of things that oftentimes is just underappreciated because it is behind the scenes. The visualizations is what people are typically drawn to. And that's it's certainly exciting when you get it right. But this hardwiring that has to be done in the foundational layer, it just can't be underestimated. It's a really, really important thing. Oh yes. In your comments, you mentioned some things that are really important. So let's kind of jump over into the ethical use and compliance side of the world here. You've mentioned HIPAA and FHIR and how these are evolving. How do we ensure things like sync mesh navigate biases and predictive models or privacy and data lakes? I think there's so much that you've already talked about where there's risk of things that just might not be known. They're not there. So how do we bake all of this into our approach, making sure that we are addressing these things ethically and responsibly?

Ratnadeep Bhattacharjee: That's that's such an important question, Chris. Um when we talk about AI in healthcare, the conversation you know often gets swept up in potential, right? Like predictive models, personalization, automation. But I but I believe, you know, we have an equal responsibility to pause and ask, like, are we building this responsibly? Is is it is it compliant? Is it fair? Is it trustworthy? So let's start with compliance. You know, tools like Syncmesh was designed with guardrails, you know, keeping you know HIPAA compliance, interoperability standards like FHIR in mind. These were not just afterthoughts, right? So Syncmesh really what SyncMesh does is ensure that we are integrating EHR data claims data, you know, HL7 feeds into a unified record, and that too very securely, you know. We make sure that whenever we deploy it, we we deploy it with proper access controls, you know, audit trails, de-identification, wherever necessary, right? So by solving these integration problems responsibly, it prevents a lot of these downstream risks that you talked about when the data is used in analytics, data lakes, AI pipelines. That's then then there's this issue of bias, right, in predictive models. In healthcare, bias isn't just a statistical problem, it can literally determine you know who gets scared and who doesn't. So our philosophy is that bias has to be tackled at the data level first. That's where you know accelerators like Syncmesh kind of plays a very uniform very important role. For example, if a model is trained only on claims data, it may it may miss key clinical or you know SDOH insights, right? But when you bring those together, the picture becomes more representative and and the model becomes more fairer, right? Yes. So compliance and bias checking are only half the story. The other half is you know you know leadership, right? And here's here's what I think, right? We need to build AI systems as if they are going to be audited tomorrow. Not just by regulators, but by the very patients whose data is being used. You know, that that mindset would force you to prioritize transparency, you know, document your process, involve diverse voices in the design stage, right? It also means that leaders would have to create cultures where you know teams feel empowered to raise ethical concerns, not just technical ones. Right. So for me, the path to trust trustworthy AI or to trusting AI is about getting the foundations right. Yeah. Compliant data pipelines, vigilant checks across vigilant checks across against bias and a leadership culture that values ethics as much as innovation, right? That's that's how we make sure AI isn't just powerful, but also safe, you know, equitable and worthy of trust.

Chris Hutchins: This is so important in light of some very recent actions taken by the California Assembly just passed some legislation that's requiring a level of transparency in how AI is being used for healthcare providers. This is something that's got to be addressed and enabled from the outset. And I love how you've described your process for auditing because those are things that really do have to be baked in up front while you're putting the foundation in to enable you to be compliant and be transparent. Even the the recent uh decision on Google, I think it was just yesterday, that there is a movement here, and the push is going to be really about this transparency. It's less about the uh it uh addressing whether something is there's some sort of monopoly going on. It's making sure that things are even equitable across technology providers, healthcare providers, but it all comes back to the central mission that we're talking about in healthcare. And it's it's really about the patient and delivering the best possible care. And all the things that you're talking about, your technology, you've designed this already with this adaptability and flexibility so that when these legislative things come up, you and your your clients are in perfect position to be able to be compliant and be ahead of the game. I really love how you've been thinking about this and how you're building. So as a leader, I I've done my little bit of homework and looking at some of the things that I've have found online. There I've read some really interesting comments and heard commentary from others that have worked with you about your leadership style, which I understand is not something you are going to boast about, but I do have to ask you some questions because I think that's a has a really a lot to do with how much success you and your teams are having. What mentorship advice do you give leaders for building compliant, trustworthy AI systems in healthcare?

Ratnadeep Bhattacharjee: So, I mean, as you as you rightly put, I'm I may not be a very, very technical founder. I may not be someone who can kind of recommend or advise someone. But what I would really say is, you know, if you're building uh with AI in mind, you know, just make sure that, you know, we I mean whatever you project yourself as is it is is is different, right? But but what really comes to the fore is how you're building it, right? Because it's not just about you, right? Healthcare is a very, very subtle kind of a domain, right? Because why I why I call it subtle is because it has a lot of intricacies to it. You're dealing with people's lives at the end of the day, right? And and data you have to make sure that whatever data goes into your models, whatever data that you are training the algorithm on, right, is top-notch, right? You have to make sure that the data quality checks are in order, the data integration pieces are in order, and you have to make sure that you know it's not just the data quality, right? It's also the overall quality analytics and you know testing process of your of your entire product should be done in a way that that makes it a product which can be trusted, right? It's not as as I spoke to you about in my previous response as well, right? It's not just about bias and you know compliance, it's also about trusting the product that that goes into the market, right? Success is one thing, but really creating a long lasting and product with a with a good legacy that will only happen if if you have if you have a solid you know base or a foundation of of the data infrastructure, right? So that that's the only thing I would say.

Chris Hutchins: A huge, huge point for people to really grab a hold of trust is currency in the kind of work that you're talking about. And we really have to spend a lot more time thinking about how do we ensure what we're delivering can be trusted. There's always the the unknowns that are part of the biases that are really concerning. What is it that we don't know and how important is it? These are things that face caregivers and clinicians on a daily basis. AI is not going to plug that gap for us. So we have to take the steps and have the flexibility in our approach to make sure that we're always making it as easy as possible to discover what else might be an important factor. So I love that. Focus on the trust, less a little bit less on the technology, not meaning we're going to be sloppy, but the trust is foundational and it's got to be at the top of our priority list. So looking ahead as a guy who's got tremendous vision, as a future builder, what bold trends like modular IT or no code pipelines are you pursuing to ethically scale AI and health? And how does enabling data maturity play into transformative outcomes for patients and providers?

Ratnadeep Bhattacharjee: Chris, you know, when I when I think about the future of AI in healthcare, I try to look past the buzzwords and focus on what might really matter, which is scalability with trust, right? Because if we can't scale responsibly, we we risk building these shiny pilots that never touch patients and patients' lives, let's say in a in a meaningful way. One bold trend I see is the move towards what I call data maturity as a service. You know, today today most organizations still think of data readiness as a one-off project, clean it once, integrate it once, and then move on to AI. But but in reality, data is living, breathing, like constantly changing. You know, payers update policies, you know, providers update systems, patients change care addresses. I believe the future is about treating data maturity as an ongoing capability. Almost, you know, like like like cloud infrastructure. Right. The organizations that win, you know, will be the ones that continuously monitor, reconcile, and improve their data pipelines so that AI has a trustworthy foundation every single day. The second second big shift is around, you know, probably context aware AI. You know, right now, most AI in healthcare is trained to perform a narrow task, you know, predict readmissions, summarize a clinical note, you know, recommend interventions. But but honestly, the real breakthroughs will come when AI systems can understand context across the patient journey. You know, imagine an AI tool that that can't that that just doesn't flag that a patient is high risk, but also understands the social, financial, and clinical context, you know, why they are such high risk and how best to intervene them. You know, that kind of intelligence requires, you know, good understanding of uh you know what kind of data that is there as a foundational layer, right? And and and and and longitudinal records and stuff like that, right? Third trend, yeah. So Chris, the third trend possibly would be AI that reduces clinical burnout. You know, I've been hearing this a lot in amongst healthcare leaders and especially the provider systems. We have seen we I mean we have all seen how technology sometimes create more clicks and more dashboards and more cognitive load, right? I believe the next generation of healthcare AI must focus on ambient, invisible experiences, systems kind of that work in the background, you know, automate the mundane and surface only. What truly matters? That's where our work on, you know, conversational Gen AI layers sometimes I believe is important because you know, instead of training clinicians to adapt to the systems, the system should adapt to them. Finally and and you know, one of the boldest trends of all is actually a cultural one. You know, ethics has to be by design. Right. We c we can't bolt on compliance or fairness at the end. We have to understand that we need to embed ethics into every single stage of AI adoption. So so yeah, these are just my thoughts, initial thoughts, Chris.

Chris Hutchins: Yeah, you're you're you're you're pointing out and and emphasizing again something so important. These are things that have to be baked into our foundational layer as we're starting to even starting out. And if you're already in well into this and struggling, I'd encourage you to take a look at some of these areas and ask some questions. This may be some of the reason that you're struggling and you're not getting the answers you're looking for, which is why Ratnadeep and Tech Variable have been created, is to help to step into these areas to really help you identify where these gaps could be and really start stitching together all of these different silos so that you have the ability to start moving forward in these things. I think another point I want to make sure that our our listeners here, particularly those who are in the decision-making capacity where you've you've got a lot of financial responsibility, we like to think about these types of things as projects. And whilst setting things up to begin with might be a project, what Ratnadeep's been talking about is really building a foundation so that you're building a core competency into the DNA of your organization so that from this day forward, every day you get better information, you get more accurate, your workflows are getting refined, and you're providing even better decision support for your providers, and you're making them able to spend more time with the patient, maybe a few less clicks, less screen time. Giving back time is really something that is so important, and I've I've heard a common thread throughout your comments that it really is enabling that foundational activity to occur and reducing the friction in it. So I I really, really appreciate what you what you've just described and what your your thoughts and approaches are. To wrap up, what's one piece of actionable advice for listeners, perhaps aspiring enablers in data strategy on fostering a culture where AI adoption empowers everyone, not just the experts?

Ratnadeep Bhattacharjee: Oh, okay. Thank you, thank you, Chris. If I had to if I had to leave the audience, you know, with just one piece of actionable advice, it would be this make AI everyone's conversation, uh, not just the experts. Too often AI gets logged away in the data science lab and the rest of the organization feels like passengers on a train. You know, they probably didn't even buy tickets for. You know, that's when adoption fails. A culture shift happens when you invite people in, when the care coordinator, you know, claims processor, even the patient advocate feels like their voice matters in shaping how AI is used, right? You don't need them to understand the math behind a neural network, probably. Right. But you do need to share them, but you do need them to share their you know lived experiences of where the bottlenecks probably are, what frustrates them, what outcomes would really make a difference in their workflow. Because if AI doesn't solve these problems, it doesn't matter how advanced it is, it it won't stick. You know, as a as a non-tech founder, I've seen this firsthand. You know, my my role has often been to demystify the jargon, to to translate the technical into practical. And that's and the beauty is when people understand the why and feel included in the how, they stop being resistant, honestly. They become champions even. So, you know, if if if if if if you're someone who is an aspiring enabler like in in data strategy, you know, start small. You know, bring cross-functional voices probably in the room. Right. Start with simple questions, you know, what part of your day probably feels like a wasted effort. Where does data slow you down instead of helping you? And then and then work with your teams to solve the problem first. When people see that, you know, AI is here to empower them, not to replace them, the culture of adoption starts to take care of itself, right?

Chris Hutchins: So important.

Ratnadeep Bhattacharjee: At the end of the day, a technology will keep evolving. You know, there'll be FHIR, different FHIR standards coming up. Uh, you know, there are more and more versions of FHIR will be coming up. LLM models will keep on more and more advanced LLM models will uh keep on coming. Yes. A culture, you know, the way we listen, include and empower, that's what would make the difference between, you know, hype and real transformation. And and I believe every leader, including me, has the chance to be that enabler, right?

Chris Hutchins: Yes, I love it. I think we've kind of come full circle and we're back to one of the most important things that you set out at the beginning. And it's a lot of listening that we need to do before we start. And I love what you're talking about and growing and having the conversations with the frontline teams that are having these interactions, asking them where the workflow bottlenecks are, where are their pain points, and trying to help in those areas and actually then delivering something that empowers them to operate at the top of their license and take some of the mundane things that are just a nuisance and taking a whole lot of time, take that out of the equation by putting some real technology that automates those types of things that are completely mundane and rule-based, and allowing them to think, allowing them that communication with the patients and with their their colleagues for a much more satisfying experience. And I think this is important for our leaders to hear this as you are leading your organization. These principles are so important because if you're engaging your your teammates and your frontline staff, they're going to learn the approach that you're taking is not only about using AI for the sake of it, and it's not even about the efficiency or replacing people. It's enabling people to do the very things that we're all should be all about in healthcare. Ratnadeep, I can't thank you enough for being my guest this week. I really enjoyed our conversation, and I know we'll continue to have some really interesting conversations in the future as things continue to evolve, and I know you and your team will be on the cutting edge, and I look forward to continuing to follow your successes.

Ratnadeep Bhattacharjee: Sure, sure, Chris. Thank you so much for having me. It is great fun and best of luck with the podcast. Thank you so much.

Chris Hutchins: Be sure and follow Ratnadeep and Tech Variable on LinkedIn and go to techvariable.com to read Ratnadeep's latest white paper. That'll do it for this episode of the Signal Room. If today's conversation sparks something in you, an idea, a challenge, a question, don't keep it to yourself. Join the conversation on LinkedIn or visit us at thesignalroompodcast.com. We're here to amplify the Signals that matter product leadership, ethics, and innovation in healthcare. Until next time, stay tuned, stay curious, and stay human.