Riding the Tsunami: Why Expanding into Unmet Demand Makes Our Jobs Safe

Rob Collie: Hello friends and welcome back from the Thanksgiving break. Today I'm going to tell you why we're super lucky to be working specifically in the data space during this AI revolution, why our industry is mathematically less likely to replace its workers with AI than other industries as long as we play our cards right.

But first, let me ask you, is part of you going, "Oh, no. Another episode about AI, Rob?" [00:00:30] Well, if so, I want to validate that feeling because I think it's natural to be experiencing a sense of AI fatigue, hearing about it all the time. You're not wrong to be worn out by it. It has all the factors that make it exhausting to hear about. First, it's buzzy and manipulative. As I've mentioned before, I hear multiple times it's the greatest gift to marketers and in influences alike, maybe ever. It inflicts FOBO, the fear of becoming obsolete, [00:01:00] and there's no gift to marketers and influencers greater than the ability to easily create fear in their audience.

AI is also hard to define and quantify. So the messaging around it is just loaded with misinformation and dramatic ill-informed takes. We all have an intuitive sense of when we're probably being lied to. So our spidey sense is basically screaming all the time these days. But again, because it's so mysterious and also fast moving, it's hard to spot which [00:01:30] messages are lies and which ones are truths. So we tend to want to shut down the incoming flow. Tuning it out is in some sense the most rational move, and there's a potent mix of risk and powerlessness here.

On one hand, there's this dark fear in the deepest recesses of our brains telling us that our careers might be at risk. And on the other hand, what happens in this space is kind of out of our control and out of our depth. The people at OpenAI and elsewhere [00:02:00] who are building tools like ChatGPT and Claude and Midjourney, those are professionals who've trained in a very, very different discipline than ones that we're used to. Some of us data people could have come up as AI researchers if our careers had started at different times and places for sure, but we didn't.

The audience for this podcast, the people who are listening, for most of us, our peak exposure to technology is basically Power BI, whether as practitioners or as business leaders, that's the tool, that's where [00:02:30] the pinnacle tool that we've all been exposed to. The breakthroughs that are happening seemingly every day in AI, those are happening in rooms we aren't in and rooms we wouldn't necessarily understand even if we were there. It's out of our hands and it's a dark art. So again, all of this combines to make the instinctive and rational move to be, tune it out. I get it. I really do.

So knowing all of that, why do I persist? When I put my podcast [00:03:00] hat on, well, I'm a content creator. I'm supposed to optimize for what my audience wants to hear, and I do suspect that a lot of people listening would probably prefer to instead hear more stories about Power BI's success. Which, by the way, those stories are happening all the time in our business today, it hasn't stopped. So if growing our podcast audience is an explicit goal, why am I not optimizing for that in our topic selection? Why, indeed. Well, here's why. Because to do anything else [00:03:30] would be both irresponsible and dishonest.

I could instead be producing episodes saying, the most important thing for you to be learning today is combining the treat-as function with calculation groups and bookmarks to produce clever Rubik's cubes and visual forecasting. And I could be taking selfies of myself making YouTube thumbnail face to promote said episodes. And those episodes would do well, they'd outperform these for sure. But things like that are not the most important thing for you to be thinking about today, and it is [00:04:00] definitely not the most important thing for me to be thinking about.

At Thanksgiving last week I reconnected with an old friend who really distilled it down in an eye blink. "Wow, Rob, you have 70 employees today. That's 70 families that you're partially responsible for feeding. How cool and how terrifying." Yeah. If I sit back and succumb to AI fatigue, if I give in to the sense of powerlessness and noise, if I give in to self-insulating denial, [00:04:30] not only am I putting my own career at risk, I'm putting many others at risk as well. And for the people tuning into this podcast, well, I feel a flavor of that same responsibility to you too. This podcast only works for me as an ongoing weekly responsibility if it feels authentic for me. As soon as I have to start faking things, I lose the energy to do them.

Which is one reason why you will never ever see an image of me on the internet making the aforementioned [00:05:00] YouTube thumbnail face. I know that shit works. I know it boosts clicks. I don't understand why it does, but I accept that it does. And nope, never going to do it. Okay, that's two mentions of YouTube thumbnail face in one podcast, that's enough.

Circling back. Folks, our careers are at risk if we ignore this trend. These generative AI systems are getting really good at writing code. The surest way for any of us to get replaced is to pretend that it's not [00:05:30] possible. And on the flip side, the surest way to also miss out on the opportunities being created is to again, seek the comfort of thinking that it's not going to amount to much or that there's nothing we can do about it. It is uncomfortable to be paying attention to this stuff. It has to be. But even with my eyes open, I'm on net an optimist about all of this. Heck, I'm an optimist here probably because my eyes are open rather than in spite of that.

So let me tell you some good news, [00:06:00] an epiphany I've had about why our industry is kind of special. This all started with a Facebook conversation with a doctor I know from my hockey league. Let me read you the post, his original post, that started it all out. So here it is, "The company that I work for is developing AI to summarize medical records into a concise history of present illness. I typically read hundreds of pages of medical records daily in order to extract the information that matters. Soon physicians [00:06:30] may not be needed except to confirm that the algorithms have chosen a treatment course wisely."

What a great post. If you've been tuning in for a while here that should sound familiar. This idea of the expert no longer executing all of the dredge work. Instead, the expert operating as like conductor, troubleshooter and referee of a set of tools which are performing the drudgery parts at our behest and leaving us the experts [00:07:00] for the most important thinking. So this doctor, his name is Tim, is very much on the right track here, I think. But of course then a bunch of his doctor friends descended into the thread explaining why they "Aren't worried in the least about this," and giving their reasons why they're not worried. And those reasons were all bad.

They were bad reasons, and Tim did an excellent job puncturing their excuses. "What about patients who speak languages other [00:07:30] than English?" One doctor said. "Well, the robots speak every language," replied Tim. "All right. Well, what about patient histories that are poorly written and/or contain incorrect information. Garbage in, garbage out," said another. Tim replied to that one and said, "Well, it's going to be trained to absorb that and still reach a course of action, and a level of confidence, just like we are. And it's going to be trained on, the work of good doctors with good outcomes more than bad doctors with bad outcomes. Just like us human doctors are if [00:08:00] we're paying attention."

And the overall tone of everything Tim was saying in this conversation was, this can't happen soon enough. He wasn't hiding behind skin-deep self-serving dismissals like his friends, but even with his eyes open, he is looking forward to this future. Less drudge work, more time doing the truly expert stuff, more time to do the right things for patients. Sign me up yesterday is his attitude. [00:08:30] This is a good mindset.

Now, given that I've been thinking about this kind of stuff a lot, and I thought that his head was in the right place and he was playing the eyes wide open game, I decided he could handle me testing the fences a little bit. So I weighed in and said something like, "Hey, it's all fun in games until someone realizes you can handle 3X the caseload." And folks, they will realize that, especially in the United States. Sadly there are a few industries as financialized as medicine has become. Big hospitals and similar medical [00:09:00] institutions are increasingly short-staffed, which is bad for patients and bad for medical professionals alike, but it sure juices margins. And the funding that's behind these AI startups in the medical field is positively banking on these tools being used to increase margins even further. These tools are being fast-tracked precisely because they promise to cut down on one of the most expensive labor costs around, physicians.

There's nothing really [00:09:30] exceptional about the medical field here though really this is a story as old as time. Paul Bunyan versus the steam-powered tree-chopper. John Henry versus the steam-powered rock drill. Now notably, Paul Bunyan lost that competition and then basically retired into the wilderness. John Henry won his competition only to die on the spot from exhaustion. These stories, these legends, they reflect the truth. Machines [00:10:00] have always been replacing labor, but historically we haven't associated that with knowledge worker jobs. That's changed today, and it should make all of us more sympathetic with our blue-collar brethren. It wasn't some special status that was protecting white-collar jobs, it was just that the right tech wasn't here yet. Yeah, if labor can be replaced, it will be. You better believe it.

I recently read a first-hand account of more than [00:10:30] a dozen people at a local news station losing their jobs as editors and producers of the local news show because that station is piloting an AI suite that makes that whole process something that can be done by one person and it's the pilot project. They get the bugs worked out of this and then they're going to roll it out to all of the other stations around the country. I'm not happy about that. That doesn't make me happy at all. These [00:11:00] skilled and creative people are now overnight out of a job. Not just out of a job, out of a career. I don't like that. I don't celebrate it, but we have to see it. We can't tune it out.

Circling back to Tim, I did catch him one time doing the self-comfort thing, I think, when he said, "Fine, if I can handle 3X the patients, I'll just get paid 3X the salary." Nope, that will definitely not happen. You know what will happen instead [00:11:30] if suddenly every doctor can handle 3X the caseload, two-thirds of doctors will get laid off, that's what will happen. And the one-third who keep their jobs, they will get paid less, not more, than what they're paid today. Because all of those doctors who are now out of work and are being paid zero, they would gladly take their jobs back for considerably less than what they were being paid before. The business in this case will hold increased leverage against the labor force.

Wait, wait, wait, wait. I told you I was coming [00:12:00] to you today with a reason for optimism. So far it sounds anything but. So let's repeat that previous thought. Two-thirds of doctors will get laid off and the one-third who remain will take a pay cut unless new doors open. Imagine for example, that there are patients who need a doctor like Tim but don't have access to one. What if, say, there were two patients not getting help today for every one patient who is? [00:12:30] What if, in other words, the medical industry and Tim's specialty is meeting only a third of the demand, which currently exists? In that scenario, a 3X efficiency boost doesn't translate so obviously into a dramatic loss of demand for physicians. The addressable market expands in light of this tech breakthrough. And not only are there more people getting the care they need, but no one loses their job. Now we're talking about an everyone wins scenario.

[00:13:00] But is it true for Tim's field? Is there a significant reservoir of unmet demand? I don't know. Worldwide there absolutely is for sure, because not everyone is fortunate enough to live in a developed economy like the US. But will that help Tim and his colleagues? Will they become the, quote-unquote, offshore resource for patients in developing countries? Or is there a sufficient unmet demand here in the US thanks to the semi-heartless [00:13:30] and short-sighted nature of operating something like healthcare as if it were an investment banking operation, pricing people out of it? And or maybe a 3X improvement in care will fetch a higher price point for an elite subset of patients and providers. I'll leave that to Tim to figure out. I'm glad he's asking these questions and glad that he's on this journey.

But while I don't confidently know the answer for the medical industry, I do know the answer for our industry. [00:14:00] In the data industry there is tremendous unmet demand today. I'd say that our industry is at most addressing like 5% of the actual demand that even exists today. Harvesting actual information from the noise of disparate and transactional business systems, optimizing workflows by filling the manual and labor-intensive gaps between those systems. As a society we're still so early in that game. In our industry [00:14:30] for every patient who is getting care, there are 19 who aren't.

And folks, we've already seen this movie. Many of you listening today weren't even around for the old style of BI, the pre-Power BI era. Back then the BI industry wasn't even addressing 1% of demand, maybe not even 1/10th of a percent of actual demand was getting met. So today's 5% represents a massive increase, and that's because a lot of pointless labor-intensive [00:15:00] drudgery got removed from the process when Power BI came along. And guess what? At the advent of Power BI, the dawn of Power BI, we had the exact same social dynamics. The old guard denying that the new way was going to work, denying that it was going to be faster, denying that people could learn it without having studied the old ways for decades. The same kind of self-serving denial as Tim's doctor friend saying ridiculous things like what if the patient doesn't speak [00:15:30] English?

And guess what? All of those established traditional BI experts, they are just fine today. They did not get replaced. Not all of them really embraced the new up-tempo methodologies, but they all definitely adopted the new tools. Along the way the data solutions industry expanded tremendously in terms of its reach. Most of P3's customers today would never have afforded to get in the game when it was a slow, expensive ivory tower style of game. [00:16:00] So we're lucky, very lucky to be working in an industry where demand remains so incredibly elastic. Not all industries are that fortunate. Take a look around. It's kind of a stimulating exercise to go one by one through other industries and ask yourself if they are addressing most of the existing demand or only a fraction of it today.

And even if you're lucky to work in one of those latter industries like the data industry that's only addressing today a small fraction of the overall demand, you [00:16:30] still have to change with the times and you have to accept the bad and the good. For example, do you know why demand is so elastic in our industry? Why today we're reaching 50 times the market that the industry used to reach? It's because the price came down. The price of a project came down. It didn't stay the same as it did before, else we'd still be stuck at fractions of a single percent. It would be nice if the market grew 50X [00:17:00] while price stayed level. But that isn't how it works, that's not how it's going to work. The price came down and that opened the market up to a much broader reach of customers. And the overall size in terms of dollars of the marketplace did grow tremendously as a result, but the individual price point had to come down for that to happen.

Power BI made labor in the data industry far more efficient. When that happened [00:17:30] the data industry, because it was so bottlenecked and reaching such a tiny fraction of its addressable market, had room to expand into that unmet demand. Generative AI tools that help us write code, that help us write formulas, that help us generate dashboards, whatever, represent another leap forward in labor efficiency in our market. In some ways it's a very different kind of revolution. No one's ever seen tools like generative AI tools before. But in other ways in [00:18:00] terms of its economic impact, it's kind of more of the same, another efficiency multiplier. And fortunately for us, still plenty of room to grow into unmet demand.

But if you want to thrive during a tech-assisted market expansion, you have to be willing to change. For example, in the last round with Power BI, we had to build a completely new kind of consulting firm from scratch and frankly discover over time how it needed [00:18:30] to work in order to pull off getting the most out of this new tool, how to pull off addressing the market that wasn't previously being served. It wasn't easy. But I'm confident we can do it again with this new wave of AI-assisted development. We surfed a 50X transition already, so bring it on. We have a lot more patients-