Latest / You've Been Heard / 426- AI Breaks Faster Than Broken Processes w/Eric Brosius

Transcript

426- AI Breaks Faster Than Broken Processes w/Eric Brosius · You've Been Heard

• 0:09
  Speaker A: All right, well, it's a pleasure to have you back for another episode of you've Been Hur the space we've built specifically for the IT leaders of today and the rising stars looking to secure their own seat at the leadership table. Our goal is to ditch the corporate scripts and have the kind of rock conversational deep dives it actually takes to lead an organization into tomorrow. Eric, we really appreciate you joining us. Thank you for your time and for being so open to sharing your experiences with our community over at you've been heard.com to our listeners. We're sitting down with a veteran who spent 30 years in the trenches. Eric is currently the VP of IT over at Sun River Health where he effectively acts as the CIO for 10 different entities, each with its own unique risk profile and security posture. A real gem in Eric's philosophy is his rather visceral comparison of IT to a sewer system, something that nobody thinks about or appreciates until it breaks. He's also got a refreshingly blunt take on the AI plateau in healthcare, balancing its life saving potential with the risk of it becoming dangerous crutch for providers. Eric, we've got a lot of ground to cover, from the light switch mentality of leadership to the unsustainable pace of tech since 2015, let alone since 2023. But I'd love to start by handing the floor to you. Could you introduce yourself and to our listeners and share a bit about your 30 year journey to the leadership seat?
• 1:43
  Speaker B: Of course. Thanks Mike for having me. As you noted, I've been in the industry for almost 30 years from when I graduated college. I feel like I've covered most of the industries that are available for you to work in. Retail, point of sale, education, loosely parallel financial. And now I've been in healthcare for the last 15 or so years, so I'm kind of blessed and having that well rounded experience.
• 2:13
  Speaker A: A little bit of diversity there for sure.
• 2:17
  Speaker B: And so like I mentioned, in healthcare, I've been the vice president of IT for Sun River Health for the last 13 years, which is New York State's largest federally qualified health center and we're top five in the nation in patient visits. So ambulatory only, but we're in our own minds, a pretty big fish in the underserved and underprivileged pond of healthcare.
• 2:43
  Speaker A: And that alone has got to be. I can't imagine myself, I've never worked in healthcare and I know some of the compliance that you have to deal with and I can't imagine trying to run 10 different organizations. I can understand using it at scale to offer services, top 10 of them, but trying to maintain and run and deal with 10 different chiefs that are the absolute chiefs of their own domain. That's gotta be an interesting opportunity.
• 3:17
  Speaker B: It's a good challenge for sure. And it's definitely not a complaint when it comes to job security because I'm not worried about that. But it definitely keeps you on your toes. And in some levels there are pieces that are the same, they're consistent. We obviously have the same kind of tenants that try to do the same base security across all of them. Some are nonprofits, some are for profits, all are in healthcare. So at least it's a lot of it becomes a little cookie cutter until you get to 50 or 60% of your tech stack. Then you really dive into the business needs and that's where you really start branching off. I have one client customer company that I deal with that is near high tech certified. So that's one of the highest rated security ratings you can get. Doesn't have to be high tech, but it's like three controls away from high tech. A lot of the other healthcare companies don't have to reach that level of standard. So there's a lot more auditability, a lot more controls, separation of duties for that company where the other ones, it's just hipaa not to diminish the importance around that. So there's definitely unique challenges for each company that we face. Thankfully I have a really good team that handles a bunch of that.
• 4:37
  Speaker A: So I'm curious in that vein, are you able to utilize a single template over all of them, even though you've got the one that's high tech and the other ones that are HIPAA maintaining HIPAA compliance, are you using like the a singular template over all of this or are you having to do individual templates for compliance?
• 5:00
  Speaker B: It's a really great question and like I kind of alluded to and I can't describe it, trying to think of a better way to describe it, but the short answer is yes. So thankfully we're able to take a lot of the CSI baselines and NIST recommendations and apply them across all the organizations. Because every device needs to be encrypted, all the traffic needs to be encrypted. We have to keep retention policies for seven years. So there's a really systematic approach to check off the big boxes. But once we reach that each business specific security needs, as I mentioned before, the high tech kind of takes it to HIPAA on steroids. So you really have to do more controls, you have to have more documentation, you have to prove more evidence to the auditors, otherwise you become non compliant and that leads to a bad position for my customers.
• 5:51
  Speaker A: Yeah, but at least in this case it sounds like that organization is pushing for that. So then that now it's not you trying to champion that for the organization. It's the organization saying we want to be this level so we're enforcing these things on ourselves. So at least they don't have, they
• 6:08
  Speaker B: don't have a choice.
• 6:09
  Speaker A: The leadership team is helping.
• 6:11
  Speaker B: Okay.
• 6:11
  Speaker A: They don't have choice.
• 6:12
  Speaker B: Yeah, no, thankfully. And I apologize for interrupting you, but New York State requires them to meet these controls before they can receive Medicaid data from the state. So personally bidding in it for 30 years, it's a bit overkill. A lot of the CISOs that I talk to about this program, it's a little heavy handed, but it's what the state Department of Health requires that any entity that gets downstream Medicaid data has to meet this floor or it could jeopardize their ability to operate.
• 6:42
  Speaker A: Is this a federal mandate or is it a New York state?
• 6:47
  Speaker B: New York State. And you're going to find a lot of other states are going to be different. Each one's going to be different. California is probably going to be maybe even more strict because they have much deeper privacy policies and laws than New York does. And I'm sure there'll be other states that are a lot looser. It's just New York State is very protective of its Medicaid data are all
• 7:08
  Speaker A: tentative and that's like 50 locations if I remember while I'm reading that correctly. So are all of those inside of the state of New York. So you're dealing with a single state?
• 7:19
  Speaker B: Yes, at this point in time. Sun river, my nonprofit entity that has the 48 clinical locations and then my for profit entity that has clients right now all pretty much are in New York State. But the for profit does have other clients when it comes to certain specific products we offer on the for profit side that aren't in New York state. But it's a different world from the standpoint of not doing their it. We're offering them bi tools or some other solution. So I don't have to live in their security infrastructure. I don't have to worry about that. I have enough to worry about on my own.
• 7:57
  Speaker A: Okay, so I'm a little interested in your statement around the pace of technology and especially going as far back as 2015 because lately the focus has been the pace of technology since 2023, since the introduction of, of the generative, or not the introduction, but the general release, shall we say, of the generative models and people's utilization of that and the way that that's changing. But you're saying since 2015 and having experienced a lot of it right there with you, except for me, transportation and you across multiples, Talk to me, tell me a little more your thoughts on that.
• 8:45
  Speaker C: Your CFO doesn't want 47 Internet bills. We help IT teams consolidate multi location Internet into parent child billing with coterminous contracts while still buying wholesale from the best providers. It gets control, finance gets clarity. No disruption, no sales pitch. Just go to you've been heard.com and answer the seven questions to simplify, streamline and save.
• 9:08
  Speaker B: So it's one of those positive challenges. Sun river has been blessed with the ability to grow. A lot of companies obviously haven't had that same luxury over the last decade. So going back to 2014, 2015, at that point in time, Hudson River Healthcare started to take over operations of the Suffolk County Health center sites out on Long Island. So we took over operations for nine health centers in Suffolk County. And so we had to coordinate with the state or the county. They came in, they ripped out all of their equipment on a Friday. We had the weekend to put in all of our equipment because we had to be operational on Monday. And we did that across nine locations. One weekend. We had to do two because of the towns that they operated in and the. I don't know if it's a federal mandate, but essentially FQHCs can't close. Like we're not allowed to decide. We're going to shut down for renovations for a week because we have to be open to provide patient care to our patients. So it was an interesting challenge going again, that was all 2014, 2015. And I remember talking to people in 2016, I'm like, oh, we got through this. Let's catch our breath, let's try to reorganize a little bit. Let's clean up our processes through 2016. And then towards the end of 2017 we started an M and A activity. And so then that ramped up through that entire process, all of 2018. What that finalized in 2019, we rebranded, changed our name, almost doubled in size when it comes to health centers and FTEs. So again, good challenges to have, but it just hasn't stopped. And I would always tell people, oh man, this isn't sustainable. Like we all need to catch our breath. Well then this Little thing called Covid comes along and that really changed our world. Well, changed the world in general, but it changed our world. We sent 400 employees to go work from home. And they're remote and they stayed remote. They have not come back into the office. So we had to free up office space. It was just all day long, like just take your computer, like just take whatever you have on your desk and go home. I don't have enough laptops. Just you gotta get away from people. And it was a little scary. And then we turned around and started doing vaccine clinics and testing centers. I had technicians out in these locations out in parking lot setting up MiFi's so that laptops could connect to the EMR. And we always, I think in healthcare and IT know we're behind the scenes helping patients save lives. We're helping our doctors do their job, which can save lives. During COVID we actually helped people save lives. Not that we were hands on, but we were right there on the front lines and it, it was hazard pay. A lot of the guys were a little bit taking their own lives at risk.
• 12:08
  Speaker A: I have to admit on to that whole topic right there because having done this as long as you and I have, it's easy for me to start listing out. Okay, if I need a pop up location, I'm going to need this, this, this and this. And you know what, I was completely oblivious to all of that. Even though I lived that life and I'm in technology, I didn't even think about all of the challenges that you and the rest of the medical community had to go through during that time with this. So it's kind of surprising to me to recognize my own blind spot there and. Yeah, exactly. It set themselves at risk, you know, especially we were.
• 12:51
  Speaker B: And you, I'm sure you've seen videos and pictures all over the nation, but we were setting up testing centers and parking lots in parks, in hall space because that's where we could get space where it was wide enough open and you had enough open air so that you weren't within six feet of anybody. And then that turned into all of those turned into vaccine pop ups. So we were all over our service region, which is Hudson Valley, New York City and Long Island, Suffolk county and Long island doing pop up vaccination. So again, here we are. Obviously a little bit less hazardous from the workers perspective. But we were out in parking lots and parks and wherever we could do pop ups because we had to get vaccines and the people as fast as possible.
• 13:36
  Speaker A: So mentioned the population centers in New York State population centers and heavy. I'm out in New Mexico where getting six feet away from anybody dying. Walk outside.
• 13:46
  Speaker B: Yeah, and, and I mean that kind of turns into a little bit PSTD when you really start to think about it. I mean New York City was ground zero, if you know of COVID and the amount of deaths and the scary things that you saw on the news. Refrigeration trucks and just things that nobody thinks about or wants to talk about now. But that's an hour for me, Lane. That all happened. I knew people that were afraid that they were going to close Manhattan down and they weren't going to be able to get off of the island. And so because it was that serious that when it, when the height of 2020. But let's not dwell on that's very depressing. And then. But we have all this stuff happening, but we still have to keep the lights on. We're doing all of this other work, but we're still running 45 health centers. We still have all of these customers and other companies that we have to support, maintain and be there for their security issues and when they have HIPAA breaches and when they have ransomware attacks. So that it's been. And I just keep saying this isn't sustainable, this isn't sustainable. But here I am ten years later, as my mom always liked to say, I'm on the right side of the dirt. So I'm okay. Okay so far. And then sun river has done some more M and a activities in 2022. We're trying to grow the for profit more solutions for Community Health in 2022, 2023, 2024. So it's that positive challenge. This isn't a complaint. Yeah, we might get a little stressed out, we might get a little burnt out, but it's. Nobody is going to be unhappy with growth and healthcare is going to continue to grow. It's shifting for sure. Shifting away from a fee for service, retail healthcare. Come in, get seen, you get paid for the visit to more value based care where you get reimbursed based upon how well you're actually performing and keeping your patients out of the er. So the landscape is shifting but the challenges are still there and we're blessed to be growing and that's exciting.
• 15:47
  Speaker A: So I got another deviation in the conversation a little bit but. And we'll circle back to that pace because that pace has, it's grown exponentially in the last three years at least. But so how your ability to handle these things to get all of These things done that got you that seat at the leadership table. How are you keeping your teams from feeling that burnout of that extended pace and the fact that they're in the middle of a 10 year or a 15, 30 year marathon? Their careers are centered around this. The organizations will take everything we're willing to give and if I'm willing to put in the 60, 80 hours a week, they'll take it. How are you and how are you helping your teams maintain that balance again?
• 16:42
  Speaker B: I think it's one of the areas in my career that I've been very successful in. I have one of the lowest turnover rates in the organization. I have staff that have been here with me for a decade, have lived this battle with me. I have staff that have come from my last company. I have staff that came from the company before that to my last company, to this company. Not that I'm irreplaceable because anybody can do my job at the end of the day, but the strengths that I have seen in my career is the ability. Work hard, play hard, show the team that I'm never going to ask them to do something that I wouldn't or haven't already done. When we converted those Suffolk county sites in 2015, I was down there sweeping floors, breaking down cardboard boxes because I needed them to do their job. And I was able to. Let me get this box out of your way. I mean, I could set a computer up on a desk and plug it in, but at some point back then to joining it on the domain and then the technical pieces, I became our own worst technician and I had to get out of the way. So I just broke down boxes.
• 17:53
  Speaker A: I get that part completely.
• 17:55
  Speaker B: And I think as a philosophy, I'm not a big outsourcer. I think outsourcing, they don't care about your problems like an FTE employee does. Again, there's that loyalty that I've created in the culture. Everybody feels safe and I mandate when they take pto, nobody's allowed to contact you unless it's me. You're not allowed to check your email. I will shut your email off if I have to. So I would like to think that it's all because of me, because everybody has a little bit of an ego at the end of the day. But it's just a lot of proper vetting when we're hiring and then always trying to promote from within. I'd rather grow somebody than trying to go and hire off the street. 1. I'll maybe get a little bit of a discount because I've been able to promote them, I've been able to get them on the job training. And so I might be able to save the organization a little bit of money. But I'm blessed with an exceptional team that understands the job, understands it, understands the 247 nature of it and in healthcare. And so when a database gets corrupted at 4 o' clock in the morning, I've got eight guys on the phone figuring out how to come back from that.
• 19:13
  Speaker A: So the growing of employees versus the outsourcing or bringing in senior people, I too like that philosophy. I like growing them, but especially with the pace and lately the new pace that we've been hitting because it feels to me like we've been in that marathon and now we're having to go from the marathon pace to the sprinting pace and sustain. But so now when wanting to ramp up the capabilities or needing to provide more for the organizations, how are you balancing bringing in somebody new and growing them against being able to drop in a air quotes professional that can hit the ground run?
• 20:00
  Speaker B: So it's been a challenge especially for some of the higher skilled, experienced, truly professional, niche markets developers, data engineers, some of those real high end talent pools. I don't have that internally. You know, I don't have a lot of that where I can say, oh, this desktop person really wants to start sticking their fingers in the data engineering pie and so I can get them some on the job training that is hard to do where if I'm dealing with a help desk or a desktop person that wants to get into networking, it's really easy to have pair him with somebody on the networking team and get some mentoring, some on the job training. I get a little free extra help for the network team because I'm using a help desk person. So we took a guy Friday afternoons, he worked for the network team and now he's on the network team because I was able to promote him. So and the typical IT, help desk, desktop engineer, telecom networking. I have a really good infrastructure group. Yeah. To promote and grow those really specialty pieces. Developer, full stack developers, web developers, front end developers, data engineers. That's a lot harder. I've had a couple people that have come in, oh I wanted to get into. I want to be a dba. Okay. With the dba. And it's not something that you can just jump in and out of because you really have to live in the space. So I haven't had as much success there unfortunately, as much as I would love to see everybody grow because I really feel like if you're not challenging your staff and encouraging them to grow, they're going to become stagnant. When they become stagnant, they're going to get resentful and then you've got a different problem on your hands because then you're hurting morale. Other people have to pick up the ball for them. Those are highly skilled positions. I've had to go out to the market and that's been a challenge in and of itself. I've made some networking contacts over the last year or two that has really helped with a skilled pipeline in. And our last couple of engineers that we've brought in were just terrific. And one of the challenges was we had a little bit of a budget issue and I said, this guy is so good. Take it out of my salary. Like cover the distance out of my salary, because that's how good he is and I don't want to lose him over a small amount of money.
• 22:19
  Speaker A: And hopefully they did the right thing and took care of him while not taking a bite out of you. Correct. Because they just hear that if you're willing to do that for them, then we need to listen. Okay, so, yes, the whole data engineering, the DevOps, all of those people, and I'm finding myself that those people are really core right now to that increase in speed that we're expected to produce. We talked about the tentities and the. You've got to have 10 different chiefs telling you what they expect out of AI today. And there may be some commonality, but I'm betting that there's more disparity than commonality amongst the tentatives.
• 23:06
  Speaker B: Obviously, AI is so prevalent in our industry right now, and I've spoken at a couple conferences about it and I always sort of like, okay, do you want to talk about the technical sides and the good use cases, or do you want to talk about the bad use cases and the moral and the ethical conversation around AI? Depending upon the group, we can kind of go down different roads and I kind of enjoy that challenge. And lately I've really had to put the brakes on AI in healthcare for my organizations because everybody's off doing their own thing. So I have 10 organizations with two to six rogue IT departments because every area, every business unit wants to go do their own thing. And at one point over about six months ago, we just said like, you can't do anything. And we actually blocked some of the loosely major AI sites, otter AI and some locations we had to block chat GPT, like just the very common ones because people were dumping Phi Exfil trading Phi in the chat, GPT or Otter was automatically joining a meeting and if it was a a client conference, that's all Phi. And if you don't have the BAA in place and if that data isn't encrypted in the cloud, you just had a HIPAA breach. But if the meaning has Phi and it's transcribing Phi in Otters cloud and now you just lost Phi. So we've been very reluctant and careful. Even Grammarly the writing tool has some AI functionality built into it and we were really careful with the people that wanted to use it because of the exfiltrating of data. And we had to be really careful that the staff people that are using it will never handle Phi. So they're really executive assistants, they are not clinical staff. These are the little things that if you're not in healthcare you're not thinking about. And there's going to be trade secrets like there's going to be other companies that have their own proprietary information that they don't want getting out into these language models either, whether they're thinking about it or not. Because then you're losing your product. So that's incredibly challenging.
• 25:19
  Speaker A: Yeah. Depending on what you're doing in like in your world you're picking up liability by that happening. In my world I potential of it, but it's not a direct liability. Like, like healthcare, I'll go with due diligence. But you've got a legal mandate. Yeah.
• 25:39
  Speaker B: And that company that I mentioned earlier that has that near high tech requirement, they are probably a year away from touching AI just because of their security silo and they have to have a reportable boundary of their IT infrastructure and where all the Medicaid data lives. And plugging some AI tool in there means you have to redocument everything. You have to get it reapproved by the state before you use it or you're out of compliance. So I just said you're not touching AI right now.
• 26:12
  Speaker A: And even once they do they're you're gonna have to do that one on prem probably you're gonna have to do an isolated case.
• 26:18
  Speaker B: Unless we have essentially either siloed carved out private AI LLM in the cloud with someone's gonna Gemini or ChatGPT or Claude's gonna jump into that business with healthcare where they can guarantee that they're not intermingling data with somebody. Yes, absolutely. Or my for profit company is developing some AI solutions. So I have to push people because that I control in its entirety. So I know its boundaries, so I have the security around that. And okay, governance around AI has been a challenge. We've had to bring again policies together and share them with all the entities because a lot of my nine other entities are smaller. They don't have IT people. That's why they come to me and had to really help them understand. You can't do this because you're just throwing phi out your window. So you really need to dial it down. There are obviously approved use cases in healthcare right now. A lot of benefits to ambient listening and helping providers transcribe visits. We have a lot of AI around our contact center. So what we're able to do in real time coaching and monitoring. Even the IT help desk has some of that sentiment calling and real time coaching integrated to it. I am very hesitant, as you alluded to earlier, of AI becoming a crutch in clinical situations where a lazy provider instead of doing their job is just going to Mike has a runny nose and into chat GPT and it comes back and says oh, Mike's got the flu. When you thought the flu you got a cold or and becoming really lazy. And I read a study the other day that AI is helping detect breast cancer so early that the insurance companies are spending more and more money doing more and more imaging tests to prove that it's right. Because it's so early, it's so small, they can't catch it with their imaging in the human eye, but AI can. So that's a great use case of where AI is advancing healthcare to everybody's benefit. Once the trust is there, once you can believe in what it's telling you. Because at the end of the day ChatGPT themselves said their newest model hallucinates 40% of the time.
• 28:49
  Speaker A: So many people are just taking what it says for granted that personally I think the term hallucination is misleading because to me, and it really took a while for that one to sink in as I was playing and trying to develop something for myself, I'm just trying to make like a little Mike's assistant and as I'm doing that and it suddenly goes off in the left field like I'm setting it up to look at my or to help me with my email management and it starts asking me to log into a completely different email system so that it can access that email that I'm asking it to look at and summarize. I'm like wait a minute, mistakes. It is making mistakes and it is completely off base. Hallucinations do not infer to me, it does not infer the level of mistake that's happening. It's just a generic term that covers any kind of mistake of thinking that it's. Yeah, back to the cold versus the flu. Right. That would be a hallucination or potentially a hallucination, even though it just wasn't given enough context. So that's an interesting.
• 30:04
  Speaker B: What you said is exactly what I said at Enterprise Connect in Las Vegas in March, which is they are using the word hallucination to make it sound better, that it's wrong, that it's inaccurate. And instead of just saying we're wrong 40% of the time, we're saying, oh, it hallucinates 40% of the time because that doesn't sound as bad. So I've been out there pounding what little drum I have that non IT people really understand just how wrong AI is right now. So you always have to take the AI bubble with a grain of salt. Right now I kind of feel like the tech bubble a decade ago, the AI bubble is probably going to burst some point in time because it's going to get us to a point and we're not going to take that next level. So I think there's great use cases to help people move the lower end of their job description into AI and let them focus on what we should be paying our staff for their higher skilled capabilities and have lower hanging fruit be handled by AI because that's a great use case. What I fear is we don't want people at the upper end to again rely on it as a crutch.
• 31:25
  Speaker A: Yeah. And so hearing all of that Ed, like I'm seeing multiple kind of challenges or opportunities in front of us and like a majority of what you're talking about is to me, kind of the level one usage of these generated models and what the public still, or society as a whole is still just trying to get used to, what can it do, where can it be? And so there's that layer of it. And then we have the third parties that Otter Grammarly you take any of your major ERPs or any major system and they're baking it in. And that's gotta be another challenge for you too is any of those systems that are bringing it, baking it in and not necessarily checking with you before suddenly it's inside of your environment and it's potentially leaking some of that data that you were talking about having to secure anyway or pulling it into their model, even though they're promising not to share it or trade in their models off of your data. There's that level. And then you got people like me. Like I said, I was trying to make my own little personal assistant. And I've got teams that are like chomping at the bit, trying to get to the cowork, the deep agent, the advanced things that will help write code and generate applications themselves. And then now I'm actually thinking of like the fourth one that you mentioned earlier, of a pure AI solution, like the detection model, because that is a standalone thing that's being created, utilizing it. And okay, four separate braces, all at the same time. Run them all, let's go. Because every one of those entities or the tentities is still trying to figure out how they're going to leverage it to the benefit of the organization.
• 33:28
  Speaker B: I think we're going to see organizations that have efficiency issues, they have poor workflow, they haven't sat down and really thought out their processes. And let's face it, nobody's perfect. Every place has some level of dysfunction when you layer AI in. And if you're building an AI solution or adding an AI solution on a house of cards, you're just going to have a more exponentially worse process because you don't have a solid foundation. So that's something that I am very conscious of as we think about where we can leverage AI in business. That you really have to make sure that where you're trying to layer it in and the processes that you're looking to for it to do it better be pretty solid because then it's just going to go a little crazy. The other thing that I know a lot of people will talk about and we'll see is Mythos from Anthropic. They gave it to Microsoft before they released it and it found an exploitable bug in Windows 11 like less than an hour is what I have heard. I haven't seen the report, but I heard this from somebody that I trust. And that's scary because the malicious actors are going to take advantage of AI and they'll do it better because they can get around almost every security that we have in place. Now for any organization, any entity, it's not a matter of if, it's a matter of when. At some point somebody's going to get into your environment, be it in healthcare or not, and you're going to have a data exfiltration layer. And what keeps me up at night is that malicious actor out there getting a hold of a strong AI tool like Mythos, and then all of a sudden they have 30 backdoors in your environment and you won't detect them. Your data's already gone and on the dark web or it's 16 months later and now suddenly you're in deep trouble. And that what scares me.
• 35:29
  Speaker A: Well, and I was going to say they're not going to start using them. They're already using Babe. I went to hang out with the guys and one of my guys works at one of the military installations close by me. And he's like, hey, have you seen that latest cdd? And I'm like, no, which one are you talking about? He's like, they've got an exploitable or zero day exploit in L. The Linux kernels that almost every version of Red Hat is fully exploitable. They leveraged AI to do it. Guess what he said, at least for the moment, you still have to have direct access to the machine. And so there's a little bit of a challenge before you can exploit it. But come on. Yeah. If anything, all of this is. Here we go. Oh, now I've got to get the scooter because running at a Sprint isn't going to be fast enough. I need to. That we need to be faster.
• 36:22
  Speaker B: And it's a great analogy.
• 36:25
  Speaker A: And so I'm hearing so much. Or there's a lot of focus on the data side and the data's got to be solid. Well, the processes too. But the thought of the fact that processes. We almost need to stop for a moment and go, wait a minute. This process. Yes. Okay, can we make this process better? Should we make this process better? Hey, can we completely change this process? Because right now when it comes to processes, I. That's an area where, well, this is the way we've always done it. I keep hearing that kind of thing. And layering this in this is a chance for us to take and say, stop. If we were going to redo this or if we were going to approach this completely new today, how would we do it? Let's play that mind game. At least run that scenario at least once. Because our processes are based on human moving something from A to B, whether it's words, technology, things.
• 37:32
  Speaker B: There's a lot there that you said that I completely agree with. If your data is bad, again, not. It's not just the process, but if your data is bad, you're going to end up in a bad spot. And everybody, every organization, whether they're admitting it or not, are in this realm of keeping up with the Joneses is, I got to get on the IT train. It's going. Or the AI train. It's going to help me with this, it's going to help me with that. And if you don't stop and you think about how to deploy an AI solution and how it's going to work, it's probably going to fail. And I've seen metrics somewhere between 30 and 50% of all AI projects over the last couple of years have failed. Their implementation failed, their utilization has failed. And you're investing all of this money into a solution that fails. That's not a good look at the end of the day. And IT leaders don't like to fail, whether we like to admit it or not. That's why we're in the business. We want to keep the lights on proposal standards, we want to keep things running and working, and we don't want to get calls at 3:00 in the morning. So you really.
• 38:39
  Speaker A: We don't want that Stewart backing up.
• 38:42
  Speaker B: Exactly. And so you really need to map out an implementation and you need to have champions that understand the ends and the failure points and the risks, more so than any other project that anybody has probably ever undertaken before.
• 38:57
  Speaker A: Yeah. Well, one of the other guests that I was interviewing and talking to around this topic, he had been challenged to produce 20 AI projects or take 20 AI projects to success throughout the year. And his experience so far in getting to that point where he was given that challenge of 20 was that two out of three fail. So his plan was that he had to get through 60.
• 39:26
  Speaker B: Right.
• 39:26
  Speaker A: And be churning through all 60 of those throughout the year. Meaning what? 60 in 12 months? So that's five a month. Yeah. And he had to go through all of those to be able to get that 20.
• 39:40
  Speaker B: And so I don't have those kind of resources, even if I was given the directive of that equation. I want to see 20 successful AI projects knowing two thirds or half of them are going to fail and you have to do more. I don't have the resources to do that right now. Yeah.
• 39:57
  Speaker A: And so the pace that we're expected to maintain, I keep trying to do my best to use the word opportunities because there are so many opportunities within this. So I'd like to back up into a different portion of the conversation. You were talking about Grammarly, Otter, what methodologies and what kinds of things have you seen in the market that help with that besides just the outbound and inbound network filtering? How else are we stopping this? How else are we controlling this?
• 40:33
  Speaker B: I think the sad answer is you end up getting a little too big brothery than I think any of us really want to Be in being conscious of where your data goes. And in healthcare, we go through a lot of risk assessments, HIPAA risk assessments, based on a lot of NIST standards, and are talking about having to do a data diagram, not just a network topology, but where and what data is going throughout your infrastructure, not just internally, but externally. And that's a kind of a daunting thing. And I know Microsoft has some tools that at least gives you insight. I think it's through their cloud security solution as to what's happening within their footprint. Part of it is you have to pay attention and you have to listen to what your colleagues are doing. Not your IT colleagues, but the business units. The only reason we heard about the OTTER AI issue is somebody said, oh, I was in a meeting, so and so wasn't there, but their OTTER AI recorder joined in their absence. And they were just like, is that okay? And it wasn't like they weren't thinking about exfiltrating Phi, et cetera. They were just thinking like, this is weird. Is this compliant? Is this a privacy issue? I of course, immediately took it to. What meeting was this? Did you guys talk about me Phi? Because you just exfiltrated 10 patients names during that meeting at Database and diagnoses, and that's a problem. So we immediately, I mean that Dave blocked OTTER AI from the environment because it would automatically join meetings. So it's frustrating because we're in this world of where, especially in healthcare, you want to share data with other healthcare partners to do better patient care. Because Mike doesn't see all of the same doctors under the same roof. He goes to a cardiologist, he goes to an oncologist or pulmonologist. Unless you're dealing with some major health center where all they live. There's a sharing of data that is challenging. And healthcare doesn't like to share data. EMRs don't like to share data, but they have to do better patient care.
• 42:41
  Speaker A: Yeah. And I mean back to your example about the breast cancer. More data that those kinds of organizations have to be able to recognize all of the cases to teach the models, the more accurate and less hallucinations are going to be out there. And so that sharing of that data, it's critical for tomorrow. What a dichotomy. Eric, I am enjoying this conversation. I could keep going for a while. I got two questions for you. One is the one I warned you about. 18 months from now, what do you think we're going to be talking about? We as IT leaders, What are we going to be talking about in 18 months that we're not talking about today.
• 43:24
  Speaker B: I honestly believe, and I hope that I'm wrong, we're going to be in the same churning limbo of how we can effectively and efficiently use AI, because I don't think there's going to be a clear blueprint for this is how you implement an AI tool. There's going to be great successes. I've already have great successes with small use cases around contact center or ambient listening from a medical charting perspective. But I don't think the processes and the businesses can keep up with AI. So it's just going to be more and more failed implementations because they're trying to put something in before the entity is prepared to operate as efficiently as some AI things can. So none of us know. I would love to be wrong. I would love to see a greater future in AI in our industry.
• 44:18
  Speaker A: You bring me to a quote and I can't remember whether it was from the Phoenix project or the goal, since the Phoenix project is based off of the goal, which was that optimization anywhere. But the constraint is an illusion and how true that is because AIs can help us optimize all these different things, but if we're not ready for the flow that is created, then the constraint just moves from point A to point B and we're still trying to get to D or E. And so that's an interesting piece. Okay, so at the end of the day, when you've secured your environment and empowered the people, what's the one thing you want to make sure the business and the industry has finally heard from IT leaders such as yourself.
• 45:10
  Speaker B: I bet everybody that you talk to, either through this podcast or in meetings or conferences, not enough non technical companies treat it like a business partner. And I've always said I can't be the tail wagging the dog because I've always worked in healthcare or education or these other industries. I wasn't working for a tech company and all we were pushing out was tech. So you had to work within the framework of, listen, I gotta talk to the doctors. They're the experts, they're the ones that I'm trying to make their jobs better. But they also have to understand we're here to help make their jobs better. We're not against them, we're not trying to fight with them. We don't want their laptops to not work. We want them to be able to work more efficiently so they can go home an hour earlier at the end of the day. And it happened just this morning. Oh, we're deploying X. And I had to go, can I see the contract for X before you deploy it? Because I have about 30 questions that nobody else has thought of. And I think everybody in. In our seats, in our shoes, can tell you that same story ten times over.
• 46:21
  Speaker A: Yeah. I mean, it's really kind of the basis of you've been heard was we started off trying to help teach the current and or the next generation how to get a seat at the table. What did those like you and I who've made it to the leadership, what did we do that helped us get there? How did we get our leadership to hear us? And thus the reason that we're calling it you've been heard. And it still, unfortunately, is so true that. Yeah. Oh, by the way, we're implementing this thing. Hey, wait a minute. This is in my end or in the environment that I'm responsible for. I'm trying to help you guys, and you're. You're doing this and you haven't even told me about it. Let me see the contract. Yeah.
• 47:10
  Speaker B: So I appreciate what you're doing from that perspective.
• 47:12
  Speaker A: Yeah. Appreciate what you're doing and the challenges and opportunities that you're having fun with. And, man, I hope that we can maintain this pace and. And survive it.
• 47:22
  Speaker B: Agreed. Well, I keep my running shoes on and lace tight is the best thing I can do.
• 47:26
  Speaker A: Yeah. All right, well, thank you very much for your time today, Eric. Truly appreciated it and so thankful for you sharing your experience.
• 47:34
  Speaker B: Thank you, Mike. I appreciate.