From Data Silos to AI Gold: Harnessing Your Data for AI-Driven Healthcare Innovation
In today’s competitive healthcare landscape, your organization’s data is a goldmine of untapped potential – but only if you can access and utilize it effectively. This webinar explores the complex challenges of achieving true interoperability in the Health and Life Sciences (HLS) industry, particularly as organizations strive to integrate AI into their operations.
We’ll discuss the historical evolution of interoperability, the critical role of data governance, and the technical and organizational barriers that can hinder progress. Prepare to dive into practical strategies for transforming your data assets into powerful decision-making tools, helping you stay ahead of regulatory demands and position your organization as a leader in the evolving HLS market.
The time for HLS data interoperability is now! Tune in and learn how to unlock its transformative power! Let us know if you have any questions or if you have any requests for our next CAS Come and See Video!
https://youtu.be/VzzTy-Kqs2A?si=py1E926eLUthxauJ
VIDEO TRANSCRIPT:
SPEAKERS
Wendy Cofran
Paul Kukk
Liv Porter
It is a joy to introduce both Wendy Cofran and Paul Kukk! There’s a lot of people in the world. Okay? And then you start looking at people you want to work closely with. And there’s a lot of experts out there, and when you look at the HLS, there are nerds and enthusiasts! It’s great when they’re one in the same.
Wendy is a strategist extraordinaire has a ton of experience in it. And Paul is an amazing software architect so it’s kind of fun getting to work with both of them because they have a different type of brain that comes together in a really cool way, I will say as one. So it’s great to have you guys do this session again. This was one that we got so much good feedback from at Life Sciences Dreamin’ and I know it’s been tweaked and brought up to 2025, because things change that fast.
So it’s great that I have you both together again. So with that, I’m going to hand it over to Wendy and Paul. All right. Liv, do you want to do a quick run through on any tech questions and Q&A Features and some things like that, and then I’ll happily jump in! So we’re gonna have some points throughout this where we would love engagement from you. You’ll see those questions on the screen, and if you don’t mind, throw into the chat your answer.
We’ll be able to see that and sort of gauge the temperature from you all at the end. We’ve dedicated some time to a Q. And A. So at that point, if you have any budding questions, throw them in the chat, and we’ll get to discussing them. Fabulous. Thanks. Liv. I appreciate that. And if there’s any technical issues during it. I, we are both supported by by Liv and Sofee, two amazing folks in the background. So put it in the chat, and I promise you it’s it’s being monitored. So I wanna welcome everybody to today’s session.
Again, I’m Wendy Cofran, the CEO of a consulting, firm Novosity as well as a former CIO. And then just all around health tech enthusiasts. And I’m so excited to bring you the virtual version of our session, I should say, from the can’t miss event of the year Life Sciences Dreamin’. So if you weren’t able to attend, I highly recommend that you put that on your calendars for next year. It is such a great opportunity to meet with your peers and hear from experts, and just get a chance to stay on top of everything that’s going on in the industry.
So with that, I am so fortunate today to be joined by my good friend Paul Kukk. As Liv mentioned, Paul is the chief technology officer at Copper Hill consulting. He’s also the author of your single source of truth, and absolutely one of the most passionate data experts that I know.
So, Paul, welcome to our virtual stage! Absolute pleasure to be here with you, Wendy, and very excited to talk about this again. To to your point, like we had great feedback from the last session, a lot of interactivity between us and the crowd, so very, very excited to be here again to talk about something I love, data!
So why keep Paul in containers, enthusiasm and engage here with all of you. So, that being said, let’s take a look at what today’s journey looks like for us. We’re gonna cover data strategy foundations the all important balancing, governance and innovation.
And you can’t have governance and innovation without, of course, regulatory compliance and advantages. We’re gonna look at how to overcome some common barriers both technology based and some internal politics and culture barriers. We’re gonna look at cultural transformation for data success, and the important role that plays.
And then of course, we’re gonna, we’re gonna finalize with how a good strategy for AI implementation and standardization. So before we get started quick question for everybody, and if you could, in, in the chat session which you could just type in whichever bullet point hits you, which aspect of healthcare data are you most interested in learning about today?
And while you’re all popping that in there, I’m gonna jump in, Paul, and we’re gonna, we’re gonna get started while we see that coming in on the chat. Excellent. All right. So Paul, we know the whole world is talking about AI data, the new gold, the new oil. Pick your valuable element. Ethics and bias is a huge topic that comes up but as you and I both know in order to turn this data into value and gold for the business, you have to have a strategy.
And it is not just an IT thing. So let’s start with the company and industry side of harnessing and enabling a company’s data because many organizations struggle to realize the full value of their data. So what are some of the practical first steps that executives can take to start transforming their data assets into actionable and valuable decision making tools for their company?
Yeah, absolutely. I think that one of the, probably the largest step that a company can do I is start a data governance initiative. This initiative will help bring valuable insights into your business and how your data is, how your data’s doing, how valuable your data is and then also the biggest thing, security around your data.
I think when you build this foundation, right, thinking about foundations for a house, if you’re. Foundation of your house is not strong, the rest of the house is gonna come down. And that’s what you have to think about your data strategy moving forward. When you start thinking about data governance, I always want you to keep these questions in mind in your head.
Is your data available? Is your data usable? Is your data pure? Is your data secure? And is your data of the highest quality to make data-driven decisions from? So it’s, so Paul, it’s interesting because, in the in the chat, the que to the answers to the question, interoperability came up as the foremost, there’s a couple of AI implementations, but it’s interesting that everybody immediately goes to interoperability.
So when you look at this journey and this journey map in order to get to interoperability, can you talk a little bit about the importance of the foundation and the assessment? Because you can’t connect what you don’t know, where it is. So, again, and I hear that just having come back from the Vive Health Tech Conference everybody’s talking about interoperability, but explain why it’s important.
These first few steps of this journey are important. Yeah, that it’s part of the inventory process of your systems and of your data is part of data governance, right? It is. It’s how those systems, the connection methods, right? How those systems can be connected to, if they can be connected to, or do you have to think outside the box?
Is there reporting that can be delivered to an some type of external SFTP server and then delivering data that way, right? That, that before you can even get to interoperability the inventories, the system, the data inventory, right? Because that connection methods, the data quality, and then the format of that data, right?
We all know not only do systems talk over different protocols, but also systems talk over different standards. HL seven, right? Fire SOAP XML is it JSON, right? Without that inventory of systems to go to the interopability. First, you’re going to do a lot of rework and plus to see quicker, ROI attack those systems, right?
Once you have that inventory, you’ll have higher trusted systems that you believe not only in the data quality, but have rest APIs have that interopability forethought of how other systems can connect, and then that’s the prioritization of your interopability. Allow that inventory and that data confirmation to drive what systems should I connect first so you can see value quicker.
All right. I think, you look at this and again it’s truly a balancing act and. There is always the tension between governance and speed. So a lot of the things that, that you and I have talked about and certainly I see it all the time out in the marketplace and I want you to share.
So with our audience the traditional approaches have been, governance, implementation, innovation with long timelines. So can you talk a little bit about a phased approach towards governance that you recommend and then, how do you identify and where do you begin?
Especially when you think about your trusted data systems and, for the life sciences industry and healthcare at large that trusted data system is obvi is, something that is impacting people’s lives. So there’s a large amount of risk around those data systems. So can you talk a little bit about where to begin and that phased approach?
Absolutely. I have this methodology of what I call is the what framework questions. What data drives action, what data can be combined? What data is duplicated and what data can be enhanced. Now, there are a lot more what questions that you can bring, but it’s asking those questions of yourself and of the system owners, right, will give you that value.
So once you ask those what questions, now we’ll start with the trusted systems, right? And you bring up HLS, right? HL S has many transactional systems. They have EMRs, they have ERPs, they have CRMs, right? We know from experience that EMRs, that’s usually the dirtiest data. So what we wanna start with is some of the cleanest data, right?
To see that value, to bring that data, not only to get the systems talking back and forth, but bringing that data into a data link, right? So you can now start to report and get data-driven actionable decisions from this data. With implementing. So now that you’ve identified these systems and you’ve brung up a scorecard.
In my book, I do have a method of scoring these systems. You now can implement that one system. Start with always one system, right? Because the big thing with Interopability, I think a lot of the times companies fail, is that they try to implement too much in the beginning and not keep it small in a phased approach.
Start with one system. Is that data clean? Well, it’s, be honest with yourself. It’s 80%, 90%. Alright, well clean it up. Before you want to start moving that data, you wanna make sure you’re not moving dirty data from point A to point B, because then that just compounds your issues. That’s why the phased approach is so important because if you think about that, Wendy, right?
If you did that for four systems, you try to do phase one with four systems, especially with dirty data, the amount of cost it would take is exponential. Well, and I think that’s one of the things I’d really like to, like you to touch on is, the traditional approach has been, again, those long timelines the decision making, granted it leads to high accuracy, but the delivery time is much slower than an innovative approach.
And I think a lot of companies clearly would want to shorten the timeline and see. The ability to move more quickly, but, talk about how to balance those two things with the, with governance and innovation and how do define those clear goals and objectives, but also making sure that, you’re not moving so fast that you’re leaving yourself subject to critical errors and mistakes and things that can cause sometimes irreparable harm.
Yeah. I’m gonna use an example a client of ours that we took this approach. We set up a data governance. We ran in parallel systems to get systems talking. I think what we did was we identified the scorecard is the most important thing. So we identified that their CRM is their most trusted source because you know what they did?
They put all the validation they needed when they were collecting the data. They did that work upfront to ensure that the cleanest data at the UI level was being put into the transactional database. Huge first step, big. So we took that as approach. So as we were building out the data governance, we connected the CRM to their EHR.
Now, it was one direction because they only trusted the data in the CRM, but not in their EMR. So as we cleaned up the data in the EMR, as we say, we got the patient data cleaned up, right? The demographic, the appointments. We then started to open up one stream at a time to feed back into the CRM and then ultimately into a warehouse.
So as we were introducing, we identified the CRM, who’s your champion? Keep EMR, who’s your champion, right? We found a champion for each one of those transactional systems, brought them together, they communicated and started the data governance. As we started to implement it took us probably three to four weeks, which is incredible speed for interopability and moving data around while we were building out the data governance plan and executing on that data governance plan.
How do you see with, I see it a lot when I talk to folks no matter what industry you’re in I’ve got my legacy systems, and again, I would throw your ERP systems into that. Clearly your EHR, the, all of the critical systems are really HR systems that keep your, your business running and how do you think about this when you have these systems.
Are not just legacy and technology, but they’re legacy and the massive amounts of data and historical data. When you think about moving this data to to the next phase for your company, how do you see companies doing it? Are they just starting picking a brand new data lake, data warehouse and then deciding that’s gonna be my clean future because this is such an enormous undertaking?
Or do you see them, how do you see them refining their practices so that they can innovate going forward at a pace that allows the business to scale at the pace that technology, quite frankly, is changing today? Absolutely. That’s great question. And because you need to keep your business running, you need to keep making money, you keep need to, in the HLS, you need to keep patient care enhancing.
So typically when we have a data architecture, right? Data architecture is the most important thing, right? That foundation. Typically what we’ll do is put an object data store there to be able to bring that data into an object data store, cleanse it within there, and then push it off to either a warehouse or the next destination, and also feedback that into the source.
So we’re actually doing three things. We’re creating availability of that data. We’re actually purifying or creating the highest quality of that data through an ODS an object data store. And we’re also feeding that, that pure data back into the systems, whether it’s a data lake. Back into the EMR or back into a destination system for Interopability.
So that is probably some of the best ways. Now we’ve done other phases and approaches, but I, to me and ODS and that data architecture underneath is probably the most pivotal way to keep the business going and also the data flow. I love that. I just popped a question in the chat for the audience.
Again just because our poll data we were having trouble with the polling. And I think as you reflect on that and the balance, let’s jump into some of the real challenges with regulatory compliance. Paul, it’s clearly a major concern. When we think about, clinical trials and drug discovery and precision medicine how can organizations as how can they leverage that improved data interop and governance so that they can meet the current requirements, but also keep an eye on what happens in this compliance and the regulatory changes that are happening so quickly.
It was the, I will say this, it was the biggest conversation piece at the Vive conference is there’s a lot of changes with this new administration and there’s a lot of uncertainty. So when you think about this regulatory compliance, what are the concerns that the industry is gonna continue to face?
Yeah. I get that, I get this statement a lot is it’s like, Paul, I can’t afford data governance, master data management, right? I can’t afford it. I’m like, well, can you afford not to do it? Right? If you think about, if you’re out of co non, if you’re non-compliant, it’s usually 2.71 times the multiplier, right?
Of not being compliant. Again, a phased approach can limit the cost of data governance and master data management, increasing that foundation. It doesn’t have to be an all or nothing. That data governance will also help you with policy enforcement, right? Whether it’s HIPAA regulations, SOC two that you have to follow.
That is your guideline, that’s your Bible. And to get through all the different types of compliance issues that your company must face. So again, it’s not that. Can you afford to do it? Phased approach is, can you afford not to do it right? A, a as data sc data hacking right now, data mining, people trying to get to other people’s data right now is probably at the highest point right now that it’s ever been.
Data is gold, right? And even the hackers want the data. Even these bad companies want the data. Everybody wants your data. Look at the marketing data now that’s being collected on everybody. It’s just astronomical. Can you afford not to do it? That’s always the question you should ask yourself. So the other thing I want you to touch on too is the importance of standards in this ability to react and also again, keep an eye on your progress as you go through this.
So, can you spend a minute or two talking about how to future-proof through standards in the organization. And then I have a. Great question from the chat that I wanna bring into your follow up answer here. Absolutely. I think that we stay with JSON, right? We stay with the the industry standards moving away from XML, moving away from, proprietary formats.
I, I think that just increases the complexity, right? Think of the 2016, right? The Cures Act. Right? Interopability was supposed to be done by now and we’re nowhere closer. HL seven Fire is beautiful. Fire is a great standard. Adapt those files, adapt the common standards that are within your industry.
Bring them in-house and they make the transactions. Of interoperability so much more easier across your business. And that’s across your business, not department, mental. Right? Because we know, we’ve seen departments have one set of standards in the, in an organization and another department has a different set of standards and they’re actually talking about the same.
Same data. Same data, right? It could be patient data, it could be user data, customer data, but they’re passing that data in different formats. Agree. And that’s where data governance and master data management comes in again. Right? Because of that policy to drive those standards, not only will those standards help define your data better and have people across your organization understand the data better.
But when changes happen across regulatory, you’ll be able to react to those changes a lot faster and become compliant a lot quicker. So data standards is an absolute must across the business, not just, do you see the major tech players in the industry adopting that as they’re building their technologies?
Whether it’s, the AWS’s, the ServiceNow’s, the Salesforce, the pick your major tech player. Are they understanding that customers are it would be a good best practice or a standard to, to follow the model that you just said? Or are they just Hey, we’re out here leading the way and innovating and you guys catch up and figure it all out on the backend?
Which of those do you see? I see a lot more of do what we are doing and not helping the business for, especially in the health market. I think this is the, probably the, you have HL seven, which is supposed to be a standard, but we have EMRs that have butchered HL seven, right?
They taking the segments and the sequences and completely change them. And then now you can’t talk to that system because you’re not talking their HL seven instead of following the HL seven.org standards. Now, fire is a great way of doing things. I haven’t seen anyone butcher fire yet, but I don’t see there’s enough adaptability yet in fire.
There’s no one pushing, there’s not a government body pushing hard enough for these companies to maintain those standards, at least in the HL seven HLS marketplace. Well, I will say, I’ll give you some good news. It was, again, a hot topic of conversation at VI and almost every leader from those companies and more, also said fire is the way to go with the JSON to the fire. So the grievance is I do think culturally that’s happening, and I don’t know that people are I think you’re right. They’re getting tired of, there has not really been a strong enough push from a federal level. I, so I think I, it sounds and felt like a lot more people are moving towards, we’ve agreed we’re gonna move fast with this, the fire and the JSON.
This is good because it opens up opportunities for you. So with that, I wanna hit this question in the chat because it brings us right to the barriers and Don brought up a biggest challenge was gaining stakeholder buy-in for investments into either data warehouse and fun fundamental infrastructure.
So, let’s spend a few moments talking about these barriers. You’ve touched on a few of them, but, let’s jump in. We’ve got ’em all here between the, let’s start with the legacy systems and then jump into how to get the stakeholder buy-in, how to talk about that value upwards to the CFO suite and the people that are, holding the budget strings.
Absolutely a, and legacy systems, I think are some of the hardest ways to go, right? Is, can you come creative ways without migrating off one system to another system, right? It’s that creative ways, right? We, what we call Copper Hill, we call brain trust, right? We have a problem. We bring our people together and we drive to solutions.
Now, sometimes with legacy systems, you can’t get the data out, right? Or it’s very difficult. Then take that phrase approach, right? Say you have 6, 4, 4 to six legacy systems that you want to drive from. Well start with the one that has the highest value. Move that one, right? Take that phased approach.
Again, I think you hear that common theme phased approach. It’s not an all or nothing, right? Which one’s gonna give you the highest value? Phase that out, move it out to the cloud. Find a competitor that’s doing the same thing. Then temporize and migrate that data. Okay. Now the next one, the next barrier with department silos.
I’ve hit this a lot personally. It’s actually drives me more nuts than the legacy systems because this is political. It’s in company politics, it’s people within the same company not wanting to share their data with other companies because it makes them feel powerful. We call them data misers, right?
That one has to come from the top down, right? Especially when you pick that executive sponsor that runs your data governance. He has to have that passion, that excitement to and tell the company This is trusted. This is trusted across the company to allow our businesses, our business to make better, better informed decisions.
That’s right. I agree with you on that one, Paul. I think that’s one of the. The biggest things that I see especially when I talk to companies about their strategy I think the other way to, to get buy-in is across the different business units in your org. So I do spend some time speaking with, product strategy, sales strategy, marketing strategy go to market strategy.
And what’s interesting is everybody has goals around how to get out there. And yet when you talk to them about, well, where’s the data you’re relying on? And it’s that commonality. So I do think I. You definitely have to have executive sponsorship, but the other way to do that is seed and grow through the organization.
And as you find major decision makers or influencers within your own company, that’s another way to start the energy building up so that by the time it gets there, these conversations have already happened at the executive level. So that when you look at getting that buy-in, it’s more of a conversation of sharing the common systems that everybody’s talking about.
And you’re the unifying voice to say, well, here’s the challenge with, we all wanna use this, the, these sources, these key sources of information, but here’s our challenge and here’s what we need to do in order to make this data more valuable for each of you. Keep it clean and governed and so on and so forth.
So, when we talk about, which brings us to the interoperability. Yeah. And, what are some of the, what’s the balance between building, buying, partnering? I hear that frequently. I know you do. So can we yeah. A little bit on the bill buyer line approach. Right. I will always tell any company, you should never build a middleware.
There are too many out there to pick from that we’ll do it. It, if this is not what your company does. Do not do it. This is one of the things that you should invest in. Buy your middleware for interopability. It will allow you to pivot a lot quicker. Not only when you add systems, but we all know people change CRMs, they change ERPs.
You don’t have to manage a huge custom code base, which cost that cost alone. Will will drive you nuts. Middleware is definitely not only the best solution for Interopability. I think to me it is the only solution because it is something that you should not build. So that’s an interesting point.
And actually this just something that came into the chat, but, as somebody that is CTO of a middleware company. Couple of questions that you know, one, what do you look for? Because just there’s the reason to go with a middleware company. There’s a lot out there. How do you validate that?
It’s a good one to go with. And then the question from the chat is, what are some of the best bi-directional middlewares that are out there and, how do we’ll call them the non-enterprise level. So the commercial s and b firms that maybe can’t afford some of the tools. I think it was mentioned in the chat, MuleSoft and things like that.
So, first how do you look for and validate a good middleware partner? And then two, how do you do that at the non-enterprise level? We’ll say to scale and grow your company without, it consuming all of your budget. Great question. I think you have to take it as you do anything else, business requirements.
Ask yourself those questions, right? Is do I have the in-house talent, do I want to manage the middleware or is that something I can outsource? Right? Those two questions alone will drive what middleware you should pick. Your middleware should handle all the standard data formats. Your middleware should already have prebuilt connectors, right?
So that low code, right? Not no code. I have not yet done a integration where there’s no code. I’m gonna keep saying that. And I have actually never done an integration that has been a Daly. Same. I’ve integrated who hundreds of systems. I probably done so many Salesforce to ERPs like a NetSuite.
You would think you would find, I have not come across anyone that had been identical, but asking yourself those questions. Right. What’s your volumes of messages going in and out of the system, right? Is it, if it’s not high volume, then you know, pick this middleware now. ’cause you have middlewares that charge by messages.
You have middleware that doesn’t charge by messages, it charged by connections. So you can pass as much as your data as you want, but you don’t get charged for all the messages. You only get charged from point A to point B. Right? Those two connectors. I think that’s a really important point and Samuel who put that question in the chat we’re not gonna recommend anybody commercially here that breaks some rules.
But the important piece really, I think Paul is, I’ve seen this change with different vendors, is, they didn’t charge for the interoperability messages passing back and forth, and then they realized, hey, it’s an opportunity for us, for our revenue generator. So that would absolutely be the caution when you’re looking at evaluating middleware and their knowledge of how much that can affect your revenue and cost streams.
I think that’s, yeah and one other thing too it’s like how fast do I need that data going, right? Do I need it really real time? Do I really need it real time or does batching solve my solution? And then that really does too, drive your selections of different middleware. Yeah.
And I think that real time question also can help with some of the buy-in questions is to get those different leaders and business units saying, do you really need this in real time? And of course they all say yes, but they’re, you really have to educate yourself on, what drives value for the business.
And a lot of my time is spent in that strategy role saying. Tell me, you, let’s validate that you really need it as fast as you need it. Yeah. Because there’s so much happening in interactions that everybody thinks they need it now, but there’s a reality that you don’t. And once you really start to shave that down it helps tremendously on the data that you can surface in the data that you need to action on.
Correct. So, I know we’re I wanna just be mindful of time, but I do wanna hit the two. We’ve talked a little bit about cultural. So the real one that I don’t wanna leave this screen without is security. It is the be all and end all of certainly our life sciences industry. All regulated industries.
Quite frankly, this is the number one concern. It was, it’s the number one topic that I’ve come across at least in the last, year and a half. So can we talk a little bit about the barrier of security? Yeah, I think right now I think, and data governance will bring this out, right?
Is that if there’s any transactional system and we’re gonna point to your legacy systems up top at the top again, right? That if data is no longer encrypted at rest, you need to look for a different system. That is probably one of the biggest no GOs right now that I’m seeing. As well as, interopability, again, middleware selection, make sure encryption from point A to point B.
But as far as data governance concerned and security, that is probably the biggest concern that I always have and it’s everything that keeps me up at night is because new hacks are coming out, new ways to get into data new ways to listen to how data’s being transported across, right?
Continuous education. Right. I still think, and it is still out there, social hacking is still one of the biggest ways how people get data. Continuous education of your employees twice a year on, on, on security concerns, social hacking. Right. Getting phishing emails, not to click on things.
I think it’s some of the easiest and cheapest way of putting security not behind you, but not always fine because those are the people that are usually are at the front line something and it’s gotta be a cultural thing. Security’s a cultural thing. Data is a cultural thing. It’s not a it thing, it’s not a business analytics data scientist thing.
It is a business cultural thing. Yeah. I agree with you and I think some of the other, some of that strategies, I think, for improvement is really transparency in communication. And I find that to be a major challenge commu. I can’t over overstate how important communication and transparency is in an organization because it also helps set clear expectations which then leads into, policy development.
And those are the important things that establish trust and both internally and externally. And I think that’s something that companies big or small need to remember is that your trust from the outside world is your biggest asset. Once that’s broken, especially in regulated industries, that really puts you in a very inopportune space.
So I was talking to a person a couple of days ago and I asked them that question. I said, what would a data breach do to your brand? How would that, and he’s it would destroy us. And I’m like, there’s where and, investment in data is one of the biggest things.
And I think if you get the executives to think about that, right? It’s your brand and how that infects dollars coming in. I think that’s when they start to really get to get it and know it. Agreed. I could not agree with you more. I think it’s so important. So that being said I do now wanna reengage with our audience here and again, big or small, what does this look like?
What cultural strategies do you see have been most effective in your organization? And I guess conversely to that, the unasked question is, which ones haven’t have been there or there aren’t any that are effective in my, so I’m leaving this up to the audience. I’m putting it putting it all on them, Paul, but I’m sure we can we can help them along as they put pop stuff in the chat.
Let’s move to wrapping this up into how do you build your AI strategy and the foundation that’s gonna help you be successful with all of the tools that you’ve just heard us use. Now, how do we actually turn that into something actionable to our AI strategy and our data strategy and that foundation found?
Yeah, again, it is beautiful. Slide. It all starts with the data. I’ve been seeing it across the market now that rushing into ai, now companies are being held right? They could be sued now for what comes out of ai for their business. I think the best way to approach it is what is your goal?
What do you want? Right? Taking it one step at a time, because I’ve seen companies just dumping data into ai and what are you doing? What do you want from that? So having a clear objective and a clear goal is the first MA main thing. And maybe it’s not ai, maybe it’s going into ML first, right? Going into machine learning and bringing back context on how the data, how it categories, maybe scores, right?
Enhancing, taking those steps, right, and saying, oh, I am bringing value from this data. Let me keep going in that direction. Because now if you’re not get seeing that value, you can pivot a lot more and you just didn’t dump a whole bunch of data. Into God knows where and where it’s being stored. Right? So I think clear defined goals of what your AI initiative is for a business is very important.
And that, to your point, Wendy, before, is communicating that goal business-wide. Because what you don’t wanna do is have employees going out to AI using this. What data are they pushing in? Are they allowed to push in this data? Right? Are you giving away ip? Are they giving away P-H-P-I-I information? Right?
Communicating your AI initiatives globally to your business is very important. So you don’t have. The rogues out there. Yeah. I’d also like you to touch on some of the ml ai and the training needs of those models and how it applies to this data the foundation.
And again, we’ve, you’ve touched a little bit too on the fire implementation challenges. Again, from the audience, some of the smaller and commercial sized customers are still doing things on paper. So when we think about the models and how to train them in this foundational part of the process.
Talk a little bit about that for our audience and what absolutely. Well, you can do ML without ai. You can’t do AI without ml. So I always say if you can go ML first, right? Learn from that. You don’t need those reasoning models. And a lot of ml, a lot of those mls are already out there.
We use AWS medical comprehend a lot as machine learning to enhance our clients’ already data sets. Right. And also to, with that enhancements, it’s increasing patient care, right? They’ll be able to find things within ICD tens pics and all those different codes. Before we get into ai, that’s where your, that’s where the big beast is, right?
That’s where your modeling, your reasoning modeling has gotta be defined. It’s gotta be tweaked. So I always encourage everyone to go ml, once you get a good handle on ml, go to AI and then start enhancing that, right? And there’s some great tools out there. We’re using Lex Lex Amazon, Lex with bedrock, right?
So we have a, an agent in Lex, which is a nice agent that, that includes ml. And then you can layer bedrock behind that, which is an ai, which you have to define reasoning models to. So again, we took a phased approach to this. We brought Lex in first. Got that working beautifully, be able to respond to questions learned.
We fed it, probably five, six years worth of data to help that learning process. But then we brought Bedrock in behind it, which is a powerhouse, AWS product that really embeds ai, but that takes a lot more time and energy because you’re rerunning those models. You’re having those models constantly learn and you’re constantly tweaking those models.
And then what do you what do you think about some of the the data that’s being made available that are, again, clinical trial specific or medical field specific? How do you see tying in or using those models that are de-identified data that is available out there? Do you recommend.
Using that data before you train yours? Or how do you know, what do you see that process, Paul, that’s a great question. So we’re gonna go back to my Lex thing. We did de-identify that data going into Lex. So I, I think that’s important because. Really to get substance from your AI or ml, you really don’t need to pass in personal information in the beginning.
You can use age, race, ethnicity, all those common category items within HLS to really bring back value. Whether it’s diagnosing or you are at risk for a heart attack, right? You don’t need to know the persons the age, the all that insulin overweight, right? All that can be done in a step-by-step process.
Again, Wendy and I, you and I believe in a phased approach. It brings a lot of value. The ROI could be seen very quickly. That sparks enthusiasm, especially from the top down, and it allows you to keep working that project and showing more and more value. I love that. Let’s look to I’m gonna jump then to the top of the pyramid.
So wanna bring it home on what are the outcomes of using this strategy from the foundation applications to consider. And then what does, where does that bring us? Well, the outcomes are tremendous. We are seeing some great results in claims in insurance data what insurance company owes for what what patients owe, right?
Because we all know claims. That’s a huge amounts of data, right? Insurance data is astronomical. Why not use the power of ML and ai right? To read through that, to bring back and find those anomalies within that data set, right? Because unfortunately, I’m not up 24 hours and seven days a week. But my AI is and we’re seeing huge improvements in billing.
In that aspects. As far as patient care, yeah. We’re seeing a lot of improvement there. Especially with setting appointments, right? We built an application that used AI that was able to detect testing. And then if you tested positive for Covid, it would actually automate and create an encounter in the EMR, which then created an appointment, sent the appointment back to the patient, and then notified the doctor if it was a negative Right.
It would still create the encounter. Because the doctor does want to know that patient thought he or she had covid or whatever test it could be done with any test now that they have and still allows that, right? So patient history, all that is still maintained. But ai we fed that about, I think about 45,000 images and that AI was amazing and the client was extremely happy, right?
Because now they’re not having a, to man the calls, they’re not having to man chats and all that stuff. And that was the power of AI to increase patient care, to get them to see a provider if they tested positive quickly. Yeah, I think another outcome that is. A is right in line with every conversation I hear in the life sciences and healthcare industry is AI as it sits next to humans we are nowhere near where AI should be on its own.
But really the importance of, in that personalized care, some of the outcomes we’re starting to see, it’s not there yet. Definitely not on the provider side. With identifying and giving more time and information, putting it in the hands of the human the decision maker. We see it with clinical trials massive amounts of data structured and unstructured.
And I know you and I didn’t really touch as much on unstructured data, so I, I wanna bring that in because that is still a challenge for the life sciences and healthcare industries. That we have to tie into this, but, how important is this as you are growing stronger with your AI strategy, your data strategy, and your foundation to consider the validations and progress where humans need to stay in the biases that are associated with that?
Absolutely. We see a lot of that too Wendy in the medical device area. Right? Because to your point of others we’re they’re streaming tons of data, right? AI and ML can see, okay, maybe, an insulin pump, right? It’s bringing back, well, you still want a human to validate, before you say, okay, up the, up, up the glucose level, inform that injection.
Right. I don’t, we are definitely not there. But I think there’s still ways that AI and ML can use, can be used to improve the provider, right? So the provider’s not looking through, megs and terabytes worth of data to bring that quicker data to them so they can provide patient care quicker in with greater action.
That’s a great point. I do see a lot of that in the med tech. Space with, and I think this is the huge opportunity for the AG agent AI where, you know, when a device goes down or it seems faulty odds are that data has already been generated for something to have noticed an anomaly long before a person.
It’s a little bit easy if a med machine isn’t distributing, there’s alarm that goes off on it, but likely there was signals ahead of time. And I think the ability for letting the AI and agentic AI come in over top of that and start to really get ahead of, and that pro and let us be proactive about solving those problems and getting in front of it.
And I think there’s, I see that huge opportunity. And of course it affects the patient care, but the reality is I. The tech all technology thinks about healthcare in four segments, provider, payer, med tech, and pharma. The reality is the industry doesn’t feel that way necessarily. Providers need pharma.
They need the med tech, obviously, yes, they need their payers, but the reality of the way the business flows in your data exchanges and the opportunity for AI to increase that collaboration is absolutely I think we’re right at the cusp of seeing AI and data being able to really transform the way this industry works.
And I got a question for you, Wendy. Oh. And I was thinking about this late last night. I was in data cream. If we just helped companies, healthcare companies be one or 2% better, what would that do? Healthcare just one, 2%, right? Make them one to 2% better with their data. It’s in the billions.
That I can tell you. I was just looking that up for for a conversation that I had last week. It is in, it’s over 2 billion. Is the opportunity, especially in workflows. I think I see those things as they’re probably the stuff that isn’t talked about the most. I how fast can you turn a room over?
How much, how efficient are you with your inventory control your lab inventory control? So it is astronomical the impact that ai agentic ai and really streamlining your data. The opportunity and the value to that is along the cost savings. I see it, it’s gonna be talked about as cost savings, but the, I think the real opportunity is if I get better at reducing my costs and I get more efficient in my operations, I’m gonna see massive revenue gains.
I, it’s always, if I could go back to our other slide, that balancing act, Yeah. Do you, which, which comes first, that chicken or the egg of, do I I’ve gotta get my house in order before I can create that efficiency, but one I do. So, what is the tolerance, of my organization to be to be where, put the investment, see the growth, put the investment, see the growth.
And I think that goes back to that innovative approach of multiple streams in your innovation rather than that, that more traditional, lengthy, yeah, it’s gonna be right at the end, but it’s gonna take so long that it might not affect it. Correct. And with that being said, we’ve got we’ve got about seven minutes left.
So, I do have a couple of questions that came in. How are let’s see, where’d it go? This will follow right up to what we just asked. So many of us are under pressure to show quick ROI from the AI investments. So what governance metrics, Paul, would you recommend to track that we track to demonstrate progress while building this foundation?
Well, I think that, again, it all comes back to reporting. What is important to the business. Let’s take a look at let’s look at patient care real quick. If if we’re seeing so many patients a day how can we. May 1st is every provider being utilized, right? So now we can track utilization, right?
Utilization of that. Now, if we know that there’s, each doctor can see, so many 32 hours of patients in a day. Right now it’s all about the numbers. So, but taking one number one, K, PI, right? Identifying that and then working it towards that. So utilization is always the one that tops of mind, because that’s always our client is when it comes to providers, are providers being utilized the best they can be?
Are they taking too much time, filling out paperwork? Well, if that’s paperwork, use ai, right? How can we use AI to help them? Right now you have that measurable, now they’re not spending an hour every evening, filling out paperwork. They’re using some type of automated system to help them fill out that paperwork, push that through that is measurable.
Yeah, I think there’s a, there’s another one too, is again, across the different industries what is the most impactful system? And, everybody automatically, if you’re a provider, they go, oh it’s the EMR, if it’s payers, it’s their claims system. It’s for MedTech it’s inventory and ERP systems.
I think the thing that I see is look at your business and see what is the most impactful system. And I’m gonna give a different answer to. To what I just said, because I’ll take the for clinical trials, right? The ability to get the right patient into the trial and get them through the trial.
Astronomical savings and efficiency. When I look at the healthcare provider side, again with that med tech comparison, bed management is monstrous. If my beds are empty or not utilized correctly then I have massive revenue losses there. And then the one that’s harder is, how fast am I moving off of my legacy system?
And, which legacy systems do I, again, it may not be those main core business systems, but what it might be is that I have. Eight to 10 other bolt-on legacy systems that I really need to get rid of. Yeah, so your ROI metrics and your could, could be those investments. So don’t just think about just those major points.
Look at places within the operations of the business that moving your systems or getting your data in healthy. Giving it the healthy foundation that we just talked about those metrics will get you future buy-in from the company that helps build. Yep. We have a strategy and it is gonna be, and so your ROI can show deprecation of systems, it can show better data.
It can show that your models are training. So, I wanna leave everybody with that. Paul, hold on. Wendy, you just brought a tear to my eye because all the data architecture stuff that we’ve always talked about, you’re hit it right on the point. Right? It’s that data foundation. That’s huge.
Awesome. I gotta wipe this tear now. Okay. Well thank you for that, Paul. I always do. Okay, Liv, I’m gonna turn it back over to and I just wanna thank everybody for their participation in the web chat. It was great to have the energy in the web chat. I can’t thank you enough and I can’t thank cloud science, cloud adoption for having us back on.
I don’t know how they keep putting up with us, Paul, but you are. But it’s been great. Yeah. Thank you both. Thank you everyone for attending. Participating next month we’ll be having a debate and we’ll be talk, talking about sometimes we automate stupid stuff and how much automation is too much and a debate between sales ops and sales within a regulated industry.
So we’d love to see you there. Thank you both, and looking forward to the next one with you guys!