What does a transforming supply chain look like in the age of artificial intelligence? How can stakeholders quell fears of displacement and affirm enhancement of the workforce when it comes to AI integration? What does it take to implement AI into an organization’s operations? Joanne Friedman, Ph.D., CEO and Principal of Smart Manufacturing at Connektedminds Inc. joins us in this Executive Perspectives episode to answer these questions and more as she tackles all-things AI in the supply chain industry.
This interview was edited and formatted for clarity.
Tyler Fussner, Managing Editor, Supply Chain Connect
Joanne, thank you for joining us today.
Joanne Friedman, Ph. D., CEO and Principal, Smart Manufacturing, Connektedminds Inc.
It's a pleasure to be here, Tyler.
Fussner 0:26
And could you please introduce yourself to our audience and tell us a little bit about your supply chain industry relationship?
Dr. Friedman 0:34
Certainly, I'm Dr. Joanne Friedman. I am the CEO and principal for smart manufacturing at Connektedminds, which is an advisory services company focused on value creation. Our clients are primarily manufacturers and their ecosystems of supply chain partners, value chain partners and technology partners.
My relationship to supply chain comes from about three decades of being in manufacturing and not only working with supply chain executives from the IT perspective, but also throughout organizations related to manufacturing by industry segment whether that's in high-tech electronics, on the process side in pharmaceuticals or life sciences. I've pretty much covered all the bases of leading IT organizations as a CIO and then later on covering them as a Gartner analyst.
Fussner 1:33
Well, I'm so glad you're able to sit down with us and speak today, because I am really excited to talk to you about artificial intelligence and how it is transforming the supply chain industry. And of course, that is a huge and broad question but maybe we can kind of break that down into different sectors. Having so much experience in the manufacturing sectors, I'd love to ask you, how is AI transforming manufacturing today?
Dr. Friedman 2:04
It's a really good question. A lot of people seem to be very focused on the idea of generative AI, and the ability to interact with AI in the form of a conversation; to raise very specific questions, which are called prompts, and then ask a generative AI, a large language model, to come back with an answer to help inform them. And what I find very interesting about that is AI has been around for a very long time. It's been used in a number of different ways, whether it's on your smartphone saying, “Hey, Google,” or “Hey, Siri,” or whatever, that kind of thing. And it's very interesting, because my smartphone just reacted to the fact that I said, “Hey, Google,” trying to record what I'm saying to you, which is ironic and interesting. But AI is predominant in our smartphones, in our use of robotics, our use of certain types of programming that are not based on large language models. And it's almost as if we forgot that AI is pretty much ubiquitous in everything that we do. It's only this big bunch of noise, if you will, hyperbole in a lot of sense and over-promising in a lot of sense, that people jumped on this, “Oh, I need to have generative AI in my factory,” or “I need to find a way to make this work,” not realizing that they've had that capability for a very long time. They just haven't really thought about it in that way and tried to enable it.
But whether it's industrial robotics, or the new form of cobots that are out there (think about Spot from Boston Dynamics) or other things, AI has been in the factory for a very long time. And what's happening now is it's starting to permeate in two different ways, in the equipment that's being used, and also in the business processes. As they're changing through digital transformation, or industry 4.0, companies are looking to use AI capabilities to bolster those processes. There's a lot of different places in manufacturing and in supply chain, and even most recently, putting AI into the customer side of the digital thread and trying to leverage generative AI to make the buying process from the customer side more efficient and more effective and more homed in to what people really want.
Fussner 4:47
I think it's interesting, right? Like you touched on, I think AI has been put in a new spotlight as of late and it's definitely interesting where people are approaching the technology from different lens, a different perspective, with the massive surge in generative AI. But there's not just one avenue to implement this technology into your processes, right?
Dr. Friedman 5:11
Well, generative AI has captured the world's attention because it's the first time that we can directly interact with it without realizing that we've been talking to our phones for, you know, 11 years or 12 years, maybe or longer, and that we've interacted with other forms of AI for many, many years, probably the past 25 years.
That being said, because what I would call the HMI (the human machine interface) has now become very sophisticated, it's captured our creativity, there are people who are very worried about it taking jobs away, and replacing people. As I said before, there's a lot of hyperbole out there. But the idea is, especially in supply chain and manufacturing, if you think about predictive analytics, for example, for the supply chain, a lot of the same prediction algorithms that have been used in sophisticated programs that you've bought or started to build yourself, have been around for a while. So, there's probably 100 different forms of AI from linear regression to all the different algorithms that can be written and permutations and combinations.
This one branch has captured our attention, just the same way as AR and VR captured our attention two or three years ago. And they're still not quite mainstream. But we're starting to think about them a little bit more, we're starting to look at, “Could I conceive of a supply chain marketplace for widgets and parts that I need to buy? Where my little avatar floats into the metaverse and says, ‘Hey, Tyler, I need to buy 10 million widgets. What have you got for me?’” Right? That's something that we are only now starting to really think about the advantage of doing that, if in fact there is an advantage.
With generative AI, it could definitely help the frontline workforce in some ways, because we can give them a voice capability that they didn't have before. Instead of looking up information on a screen to try and find the root cause for a problem, you would be able to talk to it. That makes the process faster. That makes life easier for the individual on the front line. They can do their job better. That's an additive value, not one that will take their job away. And that's part of why I was really interested to have this conversation is because workforce displacement is something that I'm very passionate about in my practice. And I don't want people to think that all of a sudden their jobs are going to disappear tomorrow, because it's not going to happen.
Fussner 8:02
I think it's really one of the biggest questions tied to the entire AI conversation, right? Every stakeholder who is thinking about integrating AI into their practices and into their workforce, they're all faced with that exact question: Are my employees going to be displaced? Are they going to be worried about displacement? But how can stakeholders really quell the fears of displacement and rather reaffirm some notions of enhancement when it comes to AI’s impact on the workforce?
Dr. Friedman 8:36
I think there are companies out there, particularly in the visual inspection space—think about video cameras on a production line. There's two schools of thought about that. One is, “I'm going to study how people move so that I can help them be better suited to the role that they're playing in the organization.” There's another school of thought that says, “That's violating people's privacy, don't do it.” And so that's one of the kinds of questions that stakeholders should be asking themselves.
The other is, “Where is the value being added to their ultimate goal?” which is faster time to value—not faster time to product, but faster time to value overall. Being able to produce products with less waste, no more rework, greener products, lower carbon or those products that would capture carbon. There's a lot of issues that need to be thought about. I would say from a stakeholder perspective, as a stakeholder in different domains, because we have more than one set of stakeholders, you have to look at AI as a tool. It is not something that is going to take 200 people out of your workforce. It is a tool to make them more productive, because that adds value to the organization. And they can then achieve the time to value that they're struggling to get to.
AI can play a role in situations where there's an extreme shortage of labor, because you can use it to upskill people from one role to another. It doesn't mean you walk a bunch of people out the door and then just put in robotics, because even the most sophisticated robotics will still need a human in the loop. So, look at different types of AI and where they add value, as opposed to just thinking that you're going to replace people on the line. And from a stakeholder perspective in IT, you have the issues of security to deal with, governance, and how back-office systems are going to have to change.
As a supply chain stakeholder right now, you might be using predictive analytics, but how far down the road does that predictive analytic take you? Because there you're looking at scheduling materials in, keeping your inventory levels going, looking at price, all of those kinds of features, but it doesn't necessarily mean that you're going to be out of a job because you have innate institutional knowledge. If you've been doing your job for more than a couple of years, you're already at a point of, I wouldn't say being irreplaceable, because no one is, but rather that that knowledge needs to be captured by the organization to help them gain the benefit of not only that knowledge, but the benefit of the value you bring to the organization in other ways.
So that's two sets of stakeholders. If I look at manufacturers and frontline workers on a production line, they know how to fix a problem better than any AI because they've had the experience. Period—full stop. So, we have to consider multiple lenses when we're talking about these things, for each of the different stakeholders. Rather than a cost-reducing, more effective way of doing something, this should be a value-add way of doing something.
Fussner 12:26
You bring up the notion of AI being a tool to put in the hands of the workforce to reach that value-time as efficiently and quickly as possible. I came from the vehicle maintenance space. And I think seeing that [concept] in action was with the latest integration of AR-enhanced goggles that vehicle technicians can wear that in real time can bring up schematics; they can connect to tech support; they can help guide different connection points. In that vein, I think it really drives home the same attention-capture or the ability to get a grasp on what AI can do the same way generative AI has captured our attention as of late. It's realizing that this isn't a displacement technology, it really is a tool that will be able to come in that you can use in your everyday practices to help expedite efficiency.
Dr. Friedman 13:33
Absolutely. And we also have to consider the degrees of AI that we're going to use. To your point, you wouldn't use a sledgehammer to put a picture on the wall. You'd use a tiny, little gentle tapping hammer to do it. Because these are tools. And, as someone I know very well once said to me, “Use the right tool for the right job.” Don't just blanket across. To your point about AR, fantastic technology. So is VR technology. But the one thing that they can't replace is touch and feel. No matter how good you are at using the goggles and pulling up schematics, or in virtual reality training someone to do a particular type of repair on a particular type of equipment, that touch sensation of knowing, “Oh yeah, I'm touching the right object, and right is tight and left his loose.” You can’t teach that in AR and VR. These are enablers. But they're not the tail that should wag the dog. And they should never be viewed that way.
Technology is only as good as what it's being used for. And part of what we do within the firm is teach people, especially executives, that if they're leading digital transformation or industry 4.0 or anything related to manufacturing that is a modernization or transformation issue, that they have to look at it from the point of view of: Don't do digital. Be digital. Because if you're not looking at both sides of the coin, the cost savings and the process efficiencies that you're going to gain, as well as how you're going to generate new revenue and new growth, you're always going to be off kilter, and you'll never see the ROI that you actually expect. It's a tool. AI is a tool, not the be all and end all. And sorry for bursting a few bubbles out there. But I'd rather be factual and authentic than, “Oh, yes! It's going to take over the world! And 65% of the population will never have a job again! And blah, blah, blah.” A) It's fear mongering, which I really detest. And B) Even the most sophisticated forms of technology are not in the workplace yet, and they won't be for another two to three years. So, before we get all bent out of shape, let's look at this from a very pragmatic point of view.
Fussner 16:17
And if we are taking that approach, and looking at what AI can do for us today, what are some of the major issues that are facing supply chains in general that AI is really primed to step in and help solve?
Dr. Friedman 16:33
I would say, in a broad category, because there are so many different issues… We have global supply chains that are broken. And they're either broken because of logistics issues, shortage of parts, shortage of raw materials. And there's a whole notion that if you look at the individual buckets of problems, there are solutions to each of them. But in the aggregate, everything is broken. There is that perspective out there.
I would say that for something like better scheduling and better receipt of goods for not just in time processing, but just in case processing, AI would be very good for that. Because historically, predictive analytics are based on history (a circular statement, but intentional). Because what people did, what companies did, was they went back and looked at a historical perspective, and said, “Okay, based on the numbers and based on the data that I was presented with, I predict that I will need X in Y date and Y timeframe.” But what they don't take into consideration is that in the real world of real time, things happen. You can't always get from here to there. So, where's your contingency?
Where I find AI is very helpful in planning “what ifs.” Assume you are a procurement genius, as a prompt, Now, tell me what are the percentages of success I'm going to have if I order 10 million widgets from my friend, Tyler, at a price of half a cent per widget to be delivered within four weeks? The AI may come back and say, “Well, historically, he was always able to deliver these to you. But because the price of gasoline has skyrocketed, or because cargo containers are at a deficit right now, or the cost of those are too high to send across the ocean, he's now delivering about seven or eight days late. Can you live with that?”
Then you go through a chain of thought exercise of saying, “Well, if this, then that.” And you begin to put your thought process in the prompting for the AI to give you the best possible answer to your question. And the places where AI tends to be really good is quick computation. Long “If this, then that” type statements. This is what algorithms are all about. So, depending on the type of AI that you're using, you might find that the predictive analytic that you put in place three years ago is not so accurate anymore. And you have to allow for the context to have changed between three years ago and today; between six months ago, and today. That's where I think it starts to add value. And this is a form of advanced analytics, not your typical analytics, that goes back and allows you to add a stream of context to get a better qualified answer. And that's where I think companies can make a lot of progress in a shorter period of time then throughout what I don't refer to as a journey of digital transformation, but rather as a lifecycle of industry 4.0 or digital transformation. Because it's iterative. And it will always be iterative for two reasons: Context changes, as I just described, and technology changes even more rapidly today than it did at any time in our past. Every week there's a new iteration of a form of technology. Keeping up with that is IT’s job or the CIO’s job to a certain extent, but it's also looking at what's coming down the pike in the future. Is there a way to get ahead of the curve? That's where the focus should be.
Fussner 20:51
I couldn't agree more. I was ready to ask you what stakeholders should be looking out for. Say they are trying to integrate more AI into their business practices; they've gone over the logistics of making this happen; they've assured their workforce; everyone's ready to start putting these tools into their practices. What do they want to keep an eye out for? Is there anything that they need to be ready to adapt to so that they are relevant and prepared for the future of AI integration?
Dr. Friedman 21:22
I would say caveat emptor, first of all. Buyer beware. Because one, the AI will only be as good as your data is clean. Machine models need to be trained, that's machine learning. If you have dirty data, it's going to be a situation of garbage in, garbage out. That's number one. Number two, look at where you are in your lifecycle of industry 4.0 or digital transformation. Make sure you still have the correct alignment because there's a thing called drift. And the drift is that at the earliest part of digital transformation, or industry 4.0, you have a vision, you have certain objectives that you want to meet or business outcomes that you're trying to achieve. Over the course of time, as you get through the planning stage, which could be much longer than people normally take and should be much longer, they get very gung-ho, “Okay, we're going to do this, and we're going to get some quick wins. And then we're going to go down the road.” If you are out of alignment, in other words, if circumstances have caused drift and a change, business circumstances have changed what that vision for transformation was to something more realistic, you have to be able to account for that.
The second part that you have to look out for in terms of alignments and timing is human-centered design. If you're not using human-centered design as part of your overarching methodology for transformation, then AI is not going to work well for you. Because you'll be creating tools that people will not use. And you'll get either resistance to change, which is a whole change management issue and very difficult to deal with, or, alternatively, you'll have certain people who will immediately grasp not only the technology, but the concept of using AI, and then a whole slew of other people who it makes no sense to. And it's not because they're not digitally native or afraid of technology, it is because it's not relevant to what they're actually doing.
You have to be a big picture thinker when it comes to not only implementing AI, but even choosing the AI. Where are you in that lifecycle of transformation? What has changed? How good is your data? How clean is it? How ready to be input as machine learning [is it]? And what part of your data do you want to make sure it's kept extremely private? You don't want IP leakage where your intellectual property leaks out into the model for AI or any part of the AI you're going to use where it could be viewed by either a third party or customers, where that is a trade secret to you, and you don't want it exposed in any way.
And the last issue is the fact that AI still can hallucinate. This is not a joke. If you've ever used ChatGPT or some other large language model, sometimes the answers you get back are really strange.
The other part of this that you must take into consideration as well is: Are you going to train, take a derivative if you will, of a large language model, use your own corporate data, train up machine learning, go through machine learning and create your own generative AI or other form of AI method and use it? If so, I hope you have a lot of talent in-house, not because it's available to you, but rather because these are difficult issues. Data engineers, data scientists, all that professional capability is not readily available, number one. Number two, it's expensive. And number three, you may not need as much of it as you think even though you're being led down that path, whether it's through systems integrators, or VARs, or technology shops, big vendors, who you get may not necessarily be the most appropriate person. You want people with industry experience who understand not only the language, or the lingua franca of the industry, but understand how to translate the data that you're giving them in machine learning into actionable insight for a specific outcome.
Fussner 26:20
I'm curious, it sounds like there's only a very select few organizations that are going to have the circumstances and the resources to be able to build that in-house. Is it more common today for those looking to integrate AI and develop AI into the practices, are they seeking out third parties? How relevant is that? Or are most people going to try to bootstrap it themselves?
Dr. Friedman 26:48
The largest organizations already are onboard. They've reached out to big consulting houses. The mid-tier is starting to look at this a little bit more carefully. But their budgets are more limited, which is actually a good thing. Because that forces them to focus on cleaning their data, looking at not bare minimums, because I don't want to sound like they have no money to do this, but rather to say, focusing their attention on what they really need today and what they can future-proof for tomorrow. And that's where they're starting to look at bootstrapping their own.
It's not as difficult to do as you may think. It's just that you need a lot of planning. And you need a very careful thought to go into it. Because the first thing you would clearly look at is, “What am I trying to achieve here?” Is it improving my time to data? Is it improving my time to decision? Is it improving my overall time to value? Where are my values being created? What are my levers from my customers, or my suppliers, or even my upstream and downstream supply chain? Those are the kinds of questions that you want them to focus on. And then as far as the bootstrapping is concerned, you can get large language models; you can get access to them. There are certain organizations in segments of manufacturing that are looking at creating what are called small language models—those that are specifically designed to meet the needs of a segment. Some of that could be supply chain-oriented or value chain-oriented. And some of it could be specific to process manufacturing versus discrete manufacturing.
There are different industry groups, and I am affiliated with many organizations, that are looking at doing this, where they're looking to translate either standards and specifications so that small- and mid-tier organizations can start using the standards and embedding those into AI for best practice. There are others that are looking at the ontologies for specific segments of manufacturing industry; automotive would be one, pharmaceutical might be another. It really may come down to being patient for six months. Let the hype die down a little bit, not into a trough of disillusionment as my former employer Gartner likes to say, but rather to say, let the market shake out. There'll be a lot of consolidation. There are about 500 startups a week in AI right now. Making choices is difficult. If you are a little bit patient and wait for things to level off over the next six months, you'll be in a better position to make a decision. Concentrate on getting your data cleaned up first in that period of time.
Fussner 30:02
It really does all come down to the data. And it sounds like if you are ready to move forward with this, you need a very forward-thinking, future-focused leadership that is able to, like you said, plan out the long game here and make sure you're hitting your benchmarks and taking the right steps.
I'm also curious, we mentioned the different segments that are out there. Are you seeing any select verticals that are adopting AI at much higher rates than others, say in the electronics space, or automotive or Mil/Aero? And on the flip side of that, do you see any verticals that are in dire need of AI integration that may be lagging behind in their adoption rates?
Dr. Friedman 30:51
I definitely see electronics, automotive and the ecosystems around those two, as well as manufacturers of fleets. When you talk about automotive, I think car, in actuality, it's trucks. That industry is adopting AI quite quickly. As is the electronics manufacturing industry because they're looking at trying to relieve the shortages and seeing how they can use anything from robotic process automation to natural language processing to machine learning and a variety of different areas of technology to help them do that. On the other side of the coin, raw materials manufacturers, in other words, goods that are then either assembled or used in the remanufacture—they're lagging.
Fussner 31:44
Yeah, and I feel like they've got a spotlight on them that they haven't had in a long time. Since coming out of the COVID pandemic, we heard so much about raw material shortages. I'm sure that they're reassessing all of the processes at this stage.
Dr. Friedman 31:59
I think part of the issue is also (and it's not one that I want in any way, shape, or form to be interpreted as politicization) people are in many organizations in many segments of manufacturing or supply chain are looking at onshoring or reshoring. They're literally moving production to other parts of the world. And they're doing it either for cost efficiencies or they're doing it because of regulatory requirements. But that's something that shouldn't be not discussed. Because that tends to influence whether or not industries are ahead of the adoption curve or well behind it. So, we have to be careful about which manufacturing sectors we're talking about.
We still have a tremendous amount of goods being produced in Asia. In apparel, it seems to be concentrating in Asia, India, also in Thailand. In any country where there was production of threads and yarns or fabrics is now in the apparel business. And there's a lot of companies in Sri Lanka, for example, that are way ahead of the U.S. in some respects when it comes to apparel manufacturing. So, when I talk about insuring, reshoring, onshoring, certain goods are going in one direction, certain goods are coming back in the opposite direction. And that has a huge impact on how technology is adopted. Because, as I said at the outset, context matters. And we live in a time of fluctuations, constant change, good, bad or indifferent. The world moves forward on a daily basis. And you have to be able to keep up with that. So, in that set of contexts with respect to whether it's RPA or NLP or machine learning, any of those technologies being adopted have to be viewed as tools in the right context, at the right time, in the right context.
Fussner 34:14
Joanne, I have a question that may almost seem redundant at this point after our conversation. I know we covered a ton of benefits; we covered the challenges; we covered the areas that AI is ready to really overhaul processes; we talked about how it can enhance things. If you could sum it all up, why does the supply chain industry need to become AI integrated?
Dr. Friedman 34:45
Recent studies, those by Deloitte, our findings, many other companies’ findings, look at two things: 65% of supply chain professionals believe AI will mitigate risk and impact risk management over the next five years. I think over the next three years, personally. McKenzie, for example, felt that AI-driven forecasting can reduce errors by up to 50%. I disagree with that. I disagree because unless your data is yesterday, or in the last hour, there's no way that that's going to happen. I think it might be 30%. Because, again, if we look at even generative AI, all the news around ChatGPT, it stopped learning in 2021. We're now in 2023, almost 2024; how current is that data? So, whereas some others believe that there will be improved accuracy, unless you're feeding in real time data, constantly, it's never going to be up to date. Therefore, how can it be predictive, or reduce errors by such a huge percentage? What it may do is help cost reduction, help risk mitigation, improve some degree of accuracy, but certainly not to that extent, unless you're pulling in feeds from everything from the commodity markets, real time trading, to the weather forecast, you will never have all that data in one place.
It's now 11:45 when we're speaking; you just won't—it's impossible to do. The best you can say is, well at 8:30 this morning I got my latest feed from weather.com or weather.ai, or whatever it is going to be called in the future, to say by 2:30 in the afternoon it may start to rain and the impact that then that may have on crops or raw materials or logistics and transportation. I think that there's a tremendous amount of benefits. But I also think that we have to be reasonable about our expectations, quell the fears of those and reaffirm that—Yes, you still need human beings to do jobs, but you can help them do them better with AI.
Fussner 37:29
Joanne, I'm also curious, say an organization has been going through this digital transformation, this transformation lifecycle, for quite some time and now all of a sudden, they're faced with the conversation of AI and bringing in AI: What should they do?
Dr. Friedman 37:45
That's a really terrific question. And I get it a lot, from clients, from peers, about what companies really should be doing. I alluded to some of this before about where you are in the lifecycle. One point of note in particular, according to Big 5 consulting, 35% of the value of your digital transformation evaporates during implementation.
If you're at that implementation stage, I would say, take a pause; take a beat; look at the other technologies that you've put in to see where AI will fit based upon two things: what is the outcome you are trying to achieve? What is the state of your data? If it is telling you or your transformation leadership is telling you that, “Oh, yes, everything is 100% clean” where your data is concerned, and it all lives in the cloud, well, I would say take a step back and decide where you want to put that machine learning and how valuable it is going to be with data going to the cloud. And that has a whole range of technical questions around it like latency, network connectivity, et cetera, et cetera. You can't just stick AI in, or a generative AI in, without considering the technological background around it.
The other side of that is, again, going back to garbage in, garbage out, your data needs to be absolutely clean. You want to be looking at the metadata as well as the data; dark data as well, which is basically data that has been sitting in a system for a fair period of time, may have cobwebs on it that needs to be dusted off and looked at whether or not it's relevant. And I would look at the contextual data. Because any outcome that you want to achieve from AI has to be reverse engineered. Here's the outcome. What did it take me to get there? Where's the data now? How can I make it cleaner before I start the process?
Because implementation tends to drop the value of transformation ROI by such a huge percentage, you want to make sure two things. You're aligned. In other words, your vision and your execution are completely in alignment. If you have any question about the quality of your data, or you want to go back and revisit that before you Institute AI, it's worth spending the time and maybe even a few dollars to do. The third point is if you're all the way down the road, you're five years in, you've already pretty much finished your transformation and you're now using it, I would say, take a pause; look at the process side, not the technology side, to see where AI adds value to the processes. Whether they're frontline workforce-oriented processes, supply chain oriented-processes, logistics oriented-processes—that's where you may find tremendous value.
Fussner 41:05
It sounds like first things first, pause. Take a reassessment of everything in-house, especially your data, with your goal in mind and then really start to put that data under a magnifying glass and make sure you have the data needed to really execute on your AI aspirations.
Dr. Friedman 41:24
Yes. And the other two things that I would add to that is, to your point of do you have enough data, is do you have too much data? Because if we look at the cost of cloud, which is exorbitant. In many, many situations, we're storing data that we don't really need. We need to rein ourselves in to train machine learning, to train models with machine learning, that are relevant and going to be relevant going forward. Not always go to ‘historian’ and stay there. Because all that does is make you reinvent the wheel more often and create technical debt. So yes, it is about clean data. But again, focus. It is the focus on what you actually need versus everything. Too many companies put everything into a cloud storage and then they're paying a huge amount of costs to get at that data. Maybe it's time for a different perspective.
