Enterprise AI Journey: Agentic AI, Generative AI and Data Foundations in Healthcare

Chris Hutchins: Today on the Signal Room, I'm joined by Gary Cao, a longtime friend of mine and executive leader who has been navigating AI adoption from the inside of an enterprise. Gary brings a perspective shaped not by theory, but accountability, aligning AI ambition with operational realities, managing risk while driving innovation, and helping organizations move from experimentation to enterprise scale. As a chief data and analytics and AI officer and serial founder of multiple internal startups focused on data and analytics across industries, Gary will share his experience and lessons with a focus on AI strategic roadmaps and how we can potentially integrate traditional analytical AI with the emerging and rapidly evolving generative AI capabilities in the enterprise context. Gary, welcome to the Signal Room.

Gary Cao: Thank you, Chris. Happy to be here.

Chris Hutchins: I am happy to be having this kind of a conversation, let a few folks in on it. We've had some nice conversations previously and I always learn something when we talk. So thank you again for joining me. I want to kind of start with some things that are really near and dear to your heart. I know you've been talking about these things a bit lately and with all of hype and the buzz going around, I thought this was a really great topic. So when you are talking to an organization and they're saying they're on an AI journey, What does that usually mean in practice?

Gary Cao: Generally, most companies when they say they are on an AI journey, they mean they have general very vague intention and also some type of tool evaluation projects, use cases in mind. but they don't have a holistic systematic way of designing a framework or roadmap for the next three to five years. It's not just next month, next quarter, even next year. It's really, I would say minimum one to three years, if not three to five years.

Chris Hutchins: Right. So how do they go about really assessing where they are then? I mean, it seems like it's not an orchestrated strategy by what you're describing. How would a leader be assessing where they fall on the AI maturity scale, do think?

Gary Cao: Many companies are still, I would say, maybe in the traditional journey of data analytics, maybe in a new phase of using generative AI, but in many people's mind, those two things are disjointed. They are not properly organically connected. And when I talk to midsize companies, CEOs and board members, I ask them questions, saying, what do you want to see in three to five years in terms of the future state of your company, in terms of strategy, in terms of your strategic priorities?

Chris Hutchins: Right.

Gary Cao: From there, I will see your strategic plan and then from there, translate that into things that we can plan. combining four things together. One is technology, is the foundation. Without that, you cannot do anything. Second one is data, is raw material. The third thing is algorithm patterns analytics. The fourth thing is the business adoption and execution of the recommendation from analytics, pattern detection and algorithm side. So that way we can see the real outcome or results of delivering the real value. So those four things come together. create a virtual cycle to truly make this capability development journey sustainable and truly delivering value that are expected by the board as well as the CEO and the owners and investors.

Chris Hutchins: Right. Right, you mentioned something very, very briefly, but it's always a challenge. Data is always one of the biggest barriers that I've run into, and I'm sure you've seen this before too. It's like people have a different understanding of the quality and completeness of their data than actually is reality, which... kind of leads to another area I want to touch on. Maybe you can pick up on this data component of it because it's a pretty significant lift, but there's a gap usually between the ambition and the execution. And I think the elements that you described probably fit into that gap in various ways. maybe talk a little bit about some of the gaps that you see and how do you approach that?

Gary Cao: great. I recently talked with a group of people, practitioners in large companies in technology, data analytics and digital transformation in Cleveland area. we met at the Case Western Reserve University, Weatherhead School of Management. We talked about enterprise AI journey. And consistent with my professional framework based on past 30 years at eight companies across industries, I identified four things to kind of as food for thought. Number one is business strategy. Without that, nothing is really meaningful. Number two is innovation and diversity of ideas, analytics and pattern detection and kind of like algorithm driven recommendations decision making. Number three is data management and number four is computing or technology or tools infrastructure. all of those four things business strategy is the most difficult topic for discussion. And innovation and diversity of ideas and analytics is less discussed. More discussed, generally on top of people's mind, is technology, tools, infrastructure, and those kind of things. And data management or data quality, data lineage, data knowledge, generally is below surface and also is less, I would say, undervalued, under discussed. So all of those four things are critical and are required to make this journey less frustrating, more successful, more productive. going back, business strategy, analytics culture, data management, and then technology. So those four things have to be there. And technology is mostly visible and discussed. Data is below the surface, sometimes out of sight, out of mind. Analytics is less discussed, still under...

Chris Hutchins: Right.

Gary Cao: valued or maybe underappreciated. And business strategy is generally also difficult because people say, how do I connect business strategy with technology and problem and use case and data management? Because data management just longer term requires much longer, let's say patience or deliberate effort.

Chris Hutchins: Right. So do you think that there's a disparity between the investment in the data foundation versus what the ambition is? think that I've seen it more often than not. There's a hidden cost that's kind of tucked underneath the technology contracts, and it's underappreciated in terms of the level of effort when it comes to getting the data fit for use. What's your experience on that?

Gary Cao: Yeah, data is... It's not as tangible as technology or tools. It's sometimes not even visible. But that's the raw material. The disparity here is that people try to invest for the long term. They are generally also very knowledgeable about, for example, data, for example, technology. But the gap here is not about investment in data management, but rather connecting data technology with business problems, and then focus on use cases, and focus on short-term value creation, and then have that connection, traction. That's the very interesting integration of all those four things together. Many companies are very strong in technology and sometimes also strong in data. Business, of course, is foundational. But the connection point, the weakest spot, is the analytics part.

Chris Hutchins: Right.

Gary Cao: like using technology and data to focus on a problem to solve it and then deliver value with supporting and servicing the business decision makers. So that connection generally is not easily established or maintained or supported.

Chris Hutchins: Yeah, it's an interesting thing. mean, I've been playing around with my own models. I'm sure you have too. But the quality of the inputs, the props and things like that. They have everything to do with what kind of an output that you're gonna generate. There's a lot of variability as well. If you don't mind, just talk a little bit about how you see generative AI kind of fitting in at an enterprise level. Like it is part of an analytics foundations. I don't know that we're really clear on where the guardrails are, but I know you spent some time on that. I'd love to hear your thoughts.

Gary Cao: Yeah, it's confusing to many people. Everybody says... about AI, but in terms of definition, people have different definitions. And in the common discussion or casual dialogue, people mention AI casually and there is something like AI slop or AI wash. In my mind, because I've been in this data science machine learning space for 30 years for my entire career, I believe AI's definition is pretty broad. There are three layers.

Chris Hutchins: Right.

Gary Cao: First layer is the traditional data analytics and machine learning, predictive modeling, statistics, mathematics, economics. that's of course further including dashboard and business intelligence. And then the second layer that's kind of like there but people just don't necessarily recognize they're there or see it is the natural language processing, visual and audio.

Chris Hutchins: Mm-hmm.

Gary Cao: and image processing. The third layer which is more visible in front of the users as well as consumers is generative AI, large language model, agentic systems and all the other hot and very dominant type of topics right now. For example, chips, example, the language model, for example, the visual video creation, those kind of things. Those things coming together, there is a clear line of evolution for the past 20, 30, 40 years. And if people don't understand the evolution, they believe AI is something magically emerging in the past few years, then that's necessarily, that's missing the entire history. So if you look at history, then you will demystify AI and say, it's magic. No, it's not. This algorithm is statistics, economics, it's mathematics. So if you look at that way, and when people say, OK, I want you from the board of directors, from CEOs, from CFOs, I want you to use AI, or generative AI, what they really mean is that, OK, I want to see results. I want to see the kind of like

Chris Hutchins: Yes. Mm-hmm.

Gary Cao: what we're doing with the new hot topics. That's not healthy in my mind. In reality, you should look at the entire history and then make sure we start, let's say, first crawl before we stand up and walk. Walk before we start running.

Chris Hutchins: Right.

Gary Cao: That's the capabilities. Of course, I'm not saying you have to do sequentially. Sometimes you can go back and forth, but then iteratively without the right data foundation, without understanding the data management side.

Chris Hutchins: Right.

Gary Cao: anything that's kind like generative AI, generative content, for example, customer service, for example, meeting summary, for example, new idea generation. Those things are helpful and useful, but they are not going to fundamentally change the decision process and culture and the product, let's say workflow.

Chris Hutchins: Yeah, I think that gets to probably a point that maybe we don't talk about enough for people to really wrap their head around what this actually is about. But I know when we've talked, you've talked about a tension that kind of sits between enterprise systems and probabilistic AI models. There's... It's not magic. I figured that much out. I don't know that I understand everything. talk a little bit about what that tension is between the enterprise systems and probabilistic AI models.

Gary Cao: I worked in healthcare industry, also worked in financial service industry. I'm using the simple, let's say, perspectives on functions. For example, financial service banking industries, there's always tension between marketing and risk management. Marketing means that you acquire new customers, you cross-sale to customers. with other products while they have some other products. marketing is generally very flexible in terms of requirements. They don't have any high level of accuracy in terms of precision. So as long as you have the probabilistic decision-making process, that's good enough.

Chris Hutchins: Right.

Gary Cao: In risk management, think the tolerance for lack of accuracy is much lower. Of course, it's still probabilistic. For example, you have a model. You can say, OK, these group of accounts have higher level, higher probability of going to delinquent or something like that. So that's still a probability. But their requirements is higher and it requires more compliance and more calculation, more economic, scientific, statistics analysis rather than

Chris Hutchins: Yes. Right.

Gary Cao: rather than tolerance of the wider range of ballpark number. Similarly, in healthcare, clinical decision making or diagnosis or treatment, they have higher level of, let's say, requirements for accuracy and for higher level of compliance and emotional value there. If you look at administration and, let's say, scheduling or...

Chris Hutchins: Right. Right.

Gary Cao: say logistics supply chain, those type of things are slightly, I say, more tolerant in terms of lack of accuracy or precision. So that's the healthy tension between those different areas for potentially applying analytics. So generative AI can be very helpful for different use cases. I have a list of things that's really valuable. For example, technical assistance, troubleshooting,

Chris Hutchins: Right.

Gary Cao: content creation and editing, for example personal professional support, those things are very top use cases for generative AI. But if you go back to say okay I want to look at the preventive let's say equipment and machinery maintenance

Chris Hutchins: Great.

Gary Cao: those things will require a little more statistical analysis and at that moment it goes beyond language models. It goes to more mathematics, statistics, then so it requires higher level of specialized expertise and a simple language model may not handle that because of course we have agentic systems that can potentially blend in.

Chris Hutchins: Right.

Gary Cao: processes and languages and also quantitative numbers. But we are not there yet. So we cannot blindly trust. Language model can help us make better decisions in traditional data analytics can easily solve. One great advancement here is that in the past,

Chris Hutchins: Right.

Gary Cao: Building a data science model may take a few hours, or a few days, or sometimes a weeks, sometimes even months, depending on the complexity of the product. But right now, if you have the right data and have high confidence in data, and also you have a process workflow defined, that model algorithm development can be as fast as a few seconds or a few minutes.

Chris Hutchins: Right.

Gary Cao: So that enables a lot of data scientists to be much more productive. But does that reduce the workload and attention to data quality, data relevance, and data usefulness? All those things still have to be there. Otherwise, it's garbage in, garbage out. And also, it goes beyond language. It's truly quantitative. The other thing is that for our current data management,

Chris Hutchins: Right.

Gary Cao: Most of enterprises are still using relational databases, Data mart, data warehouse, sometimes also they have data lake house, data lakes. But they are tabular, maybe relational data. That's different from the more commonly used.

Chris Hutchins: Great.

Gary Cao: data feed into generative models. Generative model, imagine that it's really more unstructured language, visual, audio, log files, right? So the difference is there. then generative AI greatly expanded the capabilities or maybe the ability for data scientists to use to expand or speed up their workflow and to

Chris Hutchins: Great.

Gary Cao: expand their scope of potential impact.

Chris Hutchins: Right. It occurs to me that what we're talking about is an acceleration of learning that typically would be stretched out over a much longer period of time. Just the pace that it's generating new recommendations or outputs just begs the question, do we need a different kind of governance for this generative AI approach to things that's different than just what we typically are doing for analytics?

Gary Cao: I think it's more complex and there will be some similarities and consistencies, but there will also be something different, maybe going beyond the traditional data governance. Data governance has been would say struggling for the past few decades because it's high cost and also it's very hard for people to prioritize master data versus the non master data, a little bit more production data, operational data. So it's impossible to boil the ocean. How do you balance the 80-20 rules in terms of saying 80 % of the...

Chris Hutchins: Right.

Gary Cao: impact actually is deliverable created with using the 20 % of the volume of the data. So that could potentially be a great space. Genitive AI can help in terms of saying, okay, let's look at the distribution and focus only on things that matter and also look at the outliers and look at the things that's of the norm, then we can correct that. So the governance there is probably useful. But the language part of the leakage part, many people don't realize, okay, if we do this, things already are out of the door. It's not necessarily in front of people's, let's say, computer. That will require discipline, it will require, let's say, culture and education and training and Because people have to realize we want to minimize risks, minimize the chance of making major mistakes that cannot be reversed. That has to be there first. then many others, of like smaller potential problems, mistakes, if we can recover from them, yeah, we'll just recover and gradually we learn and then each time we'll minimize the chance of making the same mistake in the future.

Chris Hutchins: Great.

Gary Cao: So you don't have to worry too much about it. But the big mistakes that you want to avoid, that has to be really on the high priority list for many top executives, including the boards.

Chris Hutchins: Yeah. Yeah, it's just a different dynamic to be having the conversations, I think, in an executive level, getting into the differences between deterministic versus probabilistic thinking. How should executives be thinking about risk tolerance? Because as you were just talking about, are just certain areas, there's just a higher threshold that has to be met for something to be acceptable. So how should they be thinking about that? And where's that margin where you know that you've got to make sure that you're hitting a certain level of accuracy before anyone's going to be comfortable that can trust it?

Gary Cao: Yeah, I think the business judgment is still important, right? So I always go back to the fundamental elements of the business, running a business. So first of all, you have to have a customer. And second of all, you have to have some type of service and product that can deliver value to the customer.

Chris Hutchins: Yes.

Gary Cao: And then you have financial performance, have operations, have everything else. From there, it's still, right now, human judgment. I would say, generally, most people are very comfortable with mechanical deterministic models. So if 1 plus 1 is 2, OK, done. But if you say 1 plus 1 is maybe 1.9, maybe 2.1,

Chris Hutchins: Yep. Yep. Right.

Gary Cao: Then people say, I don't understand that. Why do you say that? So in models, probabilistic data science or machine learning models, it's all about patterns. It's all about, let's say, judgment. So I remember some statisticians said, all of the models are wrong. Some models are useful.

Chris Hutchins: Yes. Right.

Gary Cao: I always believe that. So if all of them are wrong, then it means nothing is perfect, deterministic. So it's really realistic. So it's a mindset shift for leaders to say, OK, how do you balance the probabilistic versus deterministic? In reality, most business leaders make decisions based on judgment anyway.

Chris Hutchins: Right. Yes.

Gary Cao: For example, how do you know to invest this much dollar amount into this new product going to this new territory or market? It's all judge. So they should be comfortable. They are already comfortable making decisions based on limited information, especially for example, in business, in military, in different things. It's all the same decision making process. So as long as leaders remember the

Chris Hutchins: Right. Right.

Gary Cao: business foundational principles, then just use that. then again, generative AI is another, could be another tool, could be another enabler, could be another, for example, people say kind of like a silicon, maybe a new digital species. Okay, that might be too far in a philosophical way, but from a business perspective,

Chris Hutchins: Right.

Gary Cao: most enterprises are still in the early stage of the journey for data analytics, let alone generative AI. And what I've heard is that some CFOs are saying, saying to data analytics leaders, sorry.

Chris Hutchins: Right.

Gary Cao: I in the past few years you pushed for me to invest in data analytics. I always pushed back and said, what's ROI? Now we have even a higher level of pressure from all the places. So we could have invested more in data analytics two or three years ago, earlier. Now it's still not too late, but still we should have been, could have been much more proactive. So I think some CFOs are all realizing that, okay, until we push to the next

Chris Hutchins: Right.

Gary Cao: level, then we realize, okay, we are a little bit behind. Even though they think ROI is critical as required, but ROI is not the only decision criterion you have to use. You also have use judgment. You have to use, let's say, intangible values. You have to use culture factors. You have to use strategic value of the different projects.

Chris Hutchins: Right, you mentioned something that I think is a really important thing to dial in on just a little bit. it's really that we're talking about a thought process that has to be, we have to instantiate it better and make sure that we're not bypassing some things because. When you're talking about patterns that may exist in the data, facts that exist in the data, it leads you to predict things within some range of acceptable percentage points. But in reality, sometimes what's available to us in history, while true, is no longer the direction we're going because we've learned something. And so I think the governance process has got to be a little bit different than what we're accustomed to, I think, for that reason. You can pick almost any kind of scenario that you can imagine. But one of the best ones I've heard was just this disparity in salaries between men and women historically in certain fields. And that historical data would predict certain things and tell you where it's headed based on what it sees in the history. That's where the judgment stuff is so much more important. And we have to make sure that we're designing our governance processes to appropriately pause and create space for those judgment calls to happen so we don't get ourselves into a spot that is taking on more risk than we really ought to be able to take on.

Gary Cao: Yes, that's right. That's right. The one other factor that I've seen many companies may not have spent enough time or investment is the workforce upskilling. So it's about mindset shift, it's about ownership, it's about driving change. That's actually ultimately beneficial for the company as well as for the employees. So that will require some type of lifetime learning, lifelong learning attitude. So if you don't train them, they will stay. If you train them, they will leave. Okay, which one is better for you? So training and education and up-skilling workforce is great. But there are some other potential

Chris Hutchins: Yes. Right.

Gary Cao: about, okay, if the AI is really reducing the workforce, why would the workforce be welcoming the adoption of AI? So that's another philosophical discussion. It's about company culture, it's about why companies exist.

Chris Hutchins: I think that's a great point and I love that you're bringing up human judgment and wrapping everything around the support of human beings, which I love that because it's as exciting to talk about the technologies. I mean, it's interesting. I know people that think like we do probably enjoy some of that technical jargon and investigating and learning new things, maybe more than others. But I think you're making some really important distinctions where this is not a technology thing. This is really a whole nother level of support that we're trying to put a wrap around the delivery of good health care. but even beyond the healthcare sector. Same thing is true. This has all really got to support the lives of human beings and make their lives better in ways that we are mitigating as much risk as we possibly can. We're never gonna bat a thousand, but if we can get to 80 or 90%, when we've never been able to hit 50%, you know. That's not perfection, but it's certainly an improvement. And I think that's the balance to be figured out. And leadership has a really important task to try to figure some of these things out because the pressures for cost reduction in containment and healthcare, I don't anticipate really changing a whole lot. In fact, there's an expectation that we're gonna be more efficient. And what does that actually mean? It really depends.

Gary Cao: maybe reducing the waste, maybe optimizing the resource allocation, or maybe a better alignment of timing or geographic locations, so kind of a assignment. So I also see multiple dimensions of disparities in terms of adopting

Chris Hutchins: Right.

Gary Cao: data analytics and generative AI. One is industry spectrum, right? Some industries are much more advanced. For example, the digital native companies versus manufacturing versus all the industries, right? So they may be much earlier in their evolutionary stage. So that's the industry difference. The second one is probably a geographical difference. So East Coast, West Coast versus South versus Midwest, there may be geographical differences because of

Chris Hutchins: Yes.

Gary Cao: industry exposure and talent and other factors. The third thing is within each company, the difference or the disparity between or the gaps between technical people versus business people. Business people, think, okay, don't tell me the details. I want to get this done. But then the technical people say, okay, I know the technical aspect, but I don't know how to

Chris Hutchins: Right. Yes.

Gary Cao: effectively communicate with the business people. And then so many companies could benefit in the future or right now by identifying people and developing people who can fluently speak business language as well as technology language and then data language and analytics language, all those together. So it's a long-term

Chris Hutchins: Yes. Bye.

Gary Cao: challenge or maybe long-term phenomenon but eventually it will get better and when we are in the journey we're making changes the progress can be slow right now if we look at them but if you look back five three year five year look back at today we'll see we have traveled a long distance so it's really about believing in the vision and believing in the direction and then truly effort into it

Chris Hutchins: Yeah, I think there's an interesting term that I've been thinking about lot lately in terms of what it, the kind of role that you're describing that you've been in. I've spent a number of years in it too. But the ability to do that translation between the business and the technical team, that can't be overstated how important that is. It's such a big part of what you've done in your own engagements as an entrepreneur, but also as a major player in enterprise systems. That's just such an important role. that doesn't really matter what the title is. You just need to have people that are really good at understanding both in bridging that communication gap that often is existing because at some point you're gonna have to be held accountable for what you're delivering on. Which kind of brings me to one, just one thing on the business side. I'd like to have you weigh in a little bit. Before we pivot, I want you to put on your special glasses and look into crystal ballings and then tell me what's coming. How should boards and enterprises be evaluating the potential ROI or lack thereof? Will all the things that are coming at them? Because you and I spent way too many hours fishing through literally hundreds of emails sometimes in a single day with solutions providers covering a wide range of things that may or may not be viable. But even to do that assessment is challenging. How would you say that boards need to be thinking about that?

Gary Cao: At the of the day, it's really about judgment, about experience, about let's say, calculation of pros and cons and then benefits and cost. I would say the ROI discussion can have two elements. One is value, the other one is strategy. or growth. actually wrote an article in the past few months and saying the board and CEOs, when they look at AI investment, should balance the two approaches. One is value. The other one is growth. So value is more very specific, measurable metric in terms of ROI project based, specific use case based, but growth is more strategic value in terms of even hard to quantify benefits. So, but you should not, we should not

Chris Hutchins: All

Gary Cao: discount or maybe deprioritize the strategic value of other AI investment. In my calculation of potentially a scored card design, there are three components in this benefit ROI. One is direct impact on incremental revenue or profitability. Second one is cost avoidance or cost reduction.

Chris Hutchins: Right. Yes.

Gary Cao: Those are measurable numbers. And sometimes you will say, how about number of hours saved? That's efficiency. That's another shift. The third thing is benefits that are hard to measure in dollar values, even quantity, numbered numbers. For example, customers' net promoter score improvements. So those three things are

Chris Hutchins: Right. Yes.

Gary Cao: the decision makers can assign different weights on those three things and then come up with the score. So for example, one thing that has low or no impact on revenue, but has some light impact on time saved or maybe cost reduction or cost avoidance, but huge benefit in let's say, qualifiable but not quantifiable, let's say soft benefit. Then,

Chris Hutchins: Great.

Gary Cao: decision maker should say, yeah, I know, but this is something we should do, even though it's not short term ROI number driven. So CFO has a lot of great input into the discussion, but generally CEO is the decision maker or board will make a decision. So we should not necessarily let the one metric

Chris Hutchins: Yes.

Gary Cao: just dominate the entire discussion. it's really about balancing, it's about making sound decisions. It's a different perspective, different ways on different factors for the decision making.

Chris Hutchins: Yes. Yeah, there's always a difficult task sometimes. If you're looking at things just from how they contribute to a margin, whether it's with cost reduction or increased revenue, that obviously is the top of mind in most conversations. But then there's this other factor that probably can be a whole lot more disruptive if we don't get it right is when we're dealing with some pretty difficult shortages for staffing for nurses and clinicians, particularly in the general medicine, general and short medicine space. know, I love that you kind of highlight that there's more than one thing that you got to look at. The numbers are important, of course, but those numbers are guaranteed to get worse if you can't staff the clinicians that you need.

Gary Cao: That's why I always believe that everything has its life cycle. So one industry, one location, one product, one company, one person has this life cycle. So there is a chance that the board of directors or the CEO could reinvent a company, a culture, a business model to start a new life cycle curve. But that's not easy, right? So if you don't reinvent yourself, then it's easier to just take the ride and then eventually the life cycle will be growth and then stagnation and then decline. So how do you bend the curve from decline to be another new growth? So there will always be some type of dynamics in terms of conflict in terms of

Chris Hutchins: Right.

Gary Cao: friction, different tension, so it's always there. It's just a matter of the company's life cycle and also the vision and the core value.

Chris Hutchins: Right. Well, this has gone like super fast. I I always enjoy our conversations. If I can get you just kind of look out a little ways into the future or the next three to five years. Two things. What do you think some of the most important leadership skills and capabilities are going to matter the most? And how do you see things shaking out in terms of the organizations that are going to really come through this in a really, really positive manner, what are the characteristics of those companies and the leadership that it's going to take to do that?

Gary Cao: I think the most important trait for a leader or influencer is to solve problems, deliver value, and also provide service. So if you are delivering valuable service to the right customer, to the right stakeholder, to the right audience, you will be in a good position. To do so, I'm coming back to the potential traits of the organization.

Chris Hutchins: Right.

Gary Cao: it's the way to deal with ambiguity versus specific details. it's kind of trade off between specific ROI versus intangible benefits. How do you measure that? Another important trade for the successful organizations would be workflow optimization, architectural design. So it's not about, you have a rigid process, okay, let it go. If it's not broken, don't fix it. But based on the market demand, based on the customer needs, how do you make this workflow process steady, but also continuously evolve and improve? That way we will prevent this more.

Chris Hutchins: Great.

Gary Cao: Agile approach so when things change you can also change or be adaptive. be adaptive that's the one very important trait for the future companies.

Chris Hutchins: I really appreciate that. think the concept of disruptive innovation has been around for as long as time. But the challenges are going to only get a little bit bigger now because we're advancing at a pace that none of us have ever experienced in our lifetimes. And it's just staggering how fast the advancements are coming.

Gary Cao: easy to say, very hard to do. Yeah, I would say there are two or three things I can leave with the audience today. one is that go back to the fundamental basics. Don't forget about the core value and principles and guardrails. Number two is balance probabilistic versus deterministic approach. So you need both. It's not one or the other. And the third thing is for data analytics professionals, how do we take advantage of the advancement of technology and the environment? A lot of visibility, right? A lot of awareness about data analytics. How do we pivot from generative AI to the entire spectrum of data analytics? And of course, including generative AI.

Chris Hutchins: Right.

Gary Cao: authentic stuff on robotic process automation, digital process automation. To focus on two aspects, one is data management. The other one is explainable algorithm based recommendations. Huge opportunities exist for data management. How do we streamline this entire data management process? Is time consuming expensive? It's also not very accurate. But then,

Chris Hutchins: Yes.

Gary Cao: So how do you streamline that using, let's say, generative AI or using the genetic process? The second one is explainable algorithm-based recommendations for business decision making. That will deliver value as a last mile value delivery.

Chris Hutchins: Right. Well, Gary, mean, as always, really great conversation. I always learn things when we talk. thank you so much for coming on the show and sharing your experiences and your knowledge and giving us a glimpse of what we should be expecting over the next three to five years. It sounds like we've got plenty of work to do.

Gary Cao: Thank very much, Chris. Happy to be here and I always support you in a way that I can. It's also benefiting the data analytics professional as a community.

Chris Hutchins: Thank you so much. I really appreciate you coming on the show. I look forward to having you back to the audience. Watch for the show notes. They'll provide some information so you can definitely reach out to Gary. He's got some really great experience. if we have questions, you need some help, I know he can help you. So thanks again. And that'll do it for this episode of the Signal Room.