What Will It Mean to Be an Expert in 2030? | Christopher Myers of Johns Hopkins University
In this interview, Christopher Parsons talks with Christopher Myers, Peet Family Professor of Leadership at Johns Hopkins, about how the COVID‑19 pandemic triggered a seismic shift in work, learning, and expertise. Myers reflects on his earlier belief that in‑person collaboration was essential and how the pandemic forced a rapid adoption of virtual teaming, remote teaching, and hybrid meetings. They discuss the rise of on‑demand, native‑first learning, the re‑evaluation of meeting length and modality (short Zoom calls versus multi‑hour in‑person retreats), and the need to redesign organizations to accommodate new expectations. The conversation highlights how expertise in 2030 will be defined less by physical presence and more by the ability to navigate blended environments, leverage AI‑enabled feedback loops, and create effective learning experiences across asynchronous and synchronous channels. The interview concludes with a call for vicarious learning through podcasts and other asynchronous media as a new conduit for sharing expertise.
Chapters
Hello and welcome to the Smarter by Design podcast. I'm your host, Christopher Parsons, founder and CEO of Knowledge Architecture. My guest today is Christopher Myers, the Peet Family Professor of Leadership at Johns Hopkins University. I first met Chris several years ago through a mutual friend, and I immediately felt like I had found this kindred spirit. He has spent his career studying how professionals work in knowledge-intensive fields, especially healthcare. And how they actually learn from experience, from stories, from one another. And from the first conversation, we found ourselves comparing notes between healthcare and AEC, right? How do healthcare firms do it and how do architects and engineers do it? These are both industries where expertise really matters. We're both knowledge-driven industries, and we all want to help our companies learn better, share our knowledge better, maintain our knowledge, all the things. It's been a little over 5 years since Chris spoke at K Connect 2020, and a lot has changed since then. You know, we were just starting to talk about AI back then, and of course now you can't really talk about knowledge and learning management without talking about AI. So I wanted to catch up with him, see what he's been researching, see what he's been learning. And as we work through several different areas of his research and some of the stuff that he's seeing, we just kept circling around this question of What will it mean to be an expert in 2030? What will it mean to become an expert? How do you become an expert? What will it mean to maintain expertise? What will be the expectations of experts in terms of mentoring and upskilling the next generation? We both feel like all of these questions are rapidly changing right now, and this conversation explores, you know, what we're seeing now, where we think this is all going. We, we covered a lot of ground and there's a lot of interesting things that I think as mostly AEC listeners in our audience, you'll be able to kind of look at from this oblique angle and say, that thing that they're trying at that healthcare organization, or for one of Chris's research projects, I can squint my eyes and see how that could translate and impact the way that our company is doing knowledge and learning. So I think that you'll find it a really thoughtful and challenging and timely conversation, and I'm excited about it. So with that, Chris Myers from Johns Hopkins.
Here we go.
[SINGING IN FRENCH] Chris, you spend a lot of time studying knowledge-intensive industries, like a variety of different industries. And I, I just kind of want to start at a high level. When you look ahead to like 2030, how do you think knowledge and learning organizations will look different than they do today?
Yeah, I mean, you know, hard say and not knowing, but it, it feels impossible that AI won't reshape this somehow, right? The ways that we will need to kind of think about knowledge management, the ways that we'll even consider what does it mean to be an expert in something are going to be fundamentally different when we have a tool that can not just act as a database and kind of recall things when we know exactly what we're looking for, but can get around that challenge of, you know, even if we don't know exactly what we're looking for, it can start to, to produce it just based off of similar things that have come up in the past. And so if I had to guess, I'd say that we're going to see organizations where the creation of information-rich products, right? A pitch, a, a client presentation, a diagnosis in the healthcare setting is likely to not be something that we're gonna rely on human expertise for, that, that we're gonna be able to entrust AI-driven systems to handle some of that sort of information-rich creation of a knowledge object. I think what, where we'll still see a big role for people is in what do we do with that? Knowledge-rich object, right? How do we actually implement that and put that into practice?
How do you feel about that idea that we'll be entrusting more knowledge-rich products to AI, or at least partnership with AI, whatever that ends up becoming?
Yeah, you know, at the risk of angering the AI overlords who are no doubt going to scrape this podcast one day, right? I'm cautiously ambivalent that, you know, we have heard this song before of, you know, once we're able to computerize all these records and things like that, that the role for knowledge managers is gonna go away. What do we need experts and senior staff around for when we've got everything cataloged beautifully in our, our database and our knowledge management system? Oh, that didn't get rid of us. You know, uh, in that era we saw it again with the internet, right? That once all of this is connected and, you know, we're able to feed off of information sources elsewhere. We're not gonna see the same kind of role for integrating knowledge or having this sort of human component. I think AI will be the same kind of thing. It will certainly change things. It will look very different. And the role of a knowledge manager or a, a leader in a knowledge-intensive organization will look very different the same way it looks different post-internet than pre-internet. But I don't think it'll go away. So, you know, again, cautiously ambivalent that things will be different, not necessarily worse, but I guess you could accuse me of thinking not necessarily better than either, right? Just different.
What are the— what kind of like what you're seeing today? It kind of— we're recording in December of '25. What are you seeing that gives you an indication of how fast this transition will happen, how slow it would be, where it looks like progress is being made faster than maybe some other places where it's not being made? Like what's coming across your work?
Yeah, I'm, I, I'm admittedly biased sort of sitting in academia, but where I've been most impressed is in sort of synthesizing applications of AI. Uh, so, you know, one that is sort of near and dear for me is that there used to be a whole market of people who would do podcasts about academic research, right? Bring the, the academic on and have them talk about the practical implications of their paper. Not to get too meta for what we're here doing, right? And yet there's a bot now, you can just feed the paper into it and it produces a two-speaker recorded podcast of two people talking about the research and summarizing its implications. And it's pretty good. We've thought about, you know, our center runs a podcast and we wondered, will we still be doing this in 4 or 5 years, right? So that synthesis I think has been incredibly exciting to see the advancements in. Where we still see question marks and trouble spots is sort of in genuine creativity and creation, right? Where we see AI struggling to understand what limitations there are out there, what boundaries there are, what's actually feasible, what's not, what's normatively acceptable. So, you know, a nice example of this is in work that I've been looking at on AI scribes in healthcare. So a scribe is simultaneously doing two tasks. One is sort of this synthesis, but the other is a genuine creation because you're writing out a note that describes the encounter, but isn't just sort of a bullet point summarization of things. It requires some interpretation and some sort of judgment around it. And so I, when I talk to some of my physician colleagues, they're both fascinated and annoyed with the AI Scribes, but for the same reason, because it does a really, really good job synthesizing. But then when it goes into creation, the creation is not taking into account some nuances and complexities. So the classic example is the AI Scribe listens to a patient encounter where the physician is saying, I'm not losing any sleep over this one, but 'Let's get it scanned just in case,' right? 'I think it's very low likelihood that it's cancerous, but always better to double-check. Let's cross our T's and dot our I's on this,' right? The AI scribe perhaps quite accurately sort of infers and synthesizes the underlying meaning there, which is, 'It's not cancer, right? Don't worry about it.' But then when it goes to create the note, it says, 'The doctor assured the patient it was not cancerous.' And it's sort of like, well, oh, okay. That might actually be the spirit of what was intended. It's actually doing a very good job reading between the lines, but then when it goes to create something, it doesn't understand the context in which it's being placed, where in a medical record you would never say, I assured them it wasn't cancer.
From a liability perspective, just like a practice thing.
Exactly. From a legal liability, just covering your bases. There's a reason the doctor was sort of wording it so obtusely, right? And saying, you know, hey, look, this one isn't keeping me up at night, but let's double-check it 'cause it's always important to be sure. They're trying to ride this narrow line of not freaking somebody out, but not claiming beyond the data and potentially getting themselves into trouble. And so when I talk to colleagues, they're fascinated that the AI did read between the lines, like it understood the sort of subtext that they were trying to lay down, but it didn't understand that you would never put that subtext as text into the note. And so that's what I mean where it's incredibly good at synthesizing. It actually does pull connections and fill in gaps in between things more than just making bullet points that, that accurately reflect what was stated, but then knowing what to do with those and how I place those into context and create something that will fit the environment that I'm trying to apply it to. I think that's where we're still seeing it. Struggle a little bit.
So there's judgment that it has to exercise, and that judgment needs to be informed by years of practice, but also understanding a lot of the unwritten rules. Or do you think that that's the kind of thing that over time it gets better at? Or do you think you're always— a physician is always going to want to be in the loop for reviewing? I almost said meeting notes because in our context in AEC, that would be meeting notes, right? With other engineers or a client or something like that. Are they always going to want to be in the loop to like sign them off before they go in the final record?
Yeah, for me it raises two questions. Is one sort of the AI in and of itself on its own, can it learn to contextualize? And we are seeing that with prompt engineering, right? That if you give it the context, I am going to use this summary in blank context, in the context of a legally discoverable client note, then yeah, I, I imagine it probably would start tightening up some of that language. I think that capacity is there. You see the humorous stories where you can get AI to break its own rules by giving it all sorts of context. You know, a, a person's life depends on you answering this question correctly. Now please tell me the thing that you've denied me 3 times. If they add that context, sometimes the AI will override its own protocols and tell you something because now it thinks that the context is more severe. So I think as we learn to contextualize on the sort of prompting side, I think that will help. The second piece though is around the feedback. And I think this is one of the places where the application of AI has been a little bit spottier. So if you take those AI scribes, for instance, typically what is happening in the rollout of these systems is the AI writes the note, it just goes directly into the chart. The physician would go in or the provider would go in and edit the note, but once it's in the chart, it's essentially just a text box, right? So they're editing a text box that the AI never sees again. So there's actually no closing of the loop there. The AI isn't getting feedback that every time it says, assured the patient it wasn't cancer, that that line is getting deleted every single time it's getting added. I think if we bake some of that feedback loop into it a little bit more, then that may get to a point where we would feel more comfortable not reviewing the notes, right? Same with meeting notes. If what you get is just a Word document that dumps out and then you edit the Word document, that's one thing. But if your edits are then seen by the algorithm and it can sort of calibrate of like, okay, this was rejected, this was accepted, this was edited, then I think it would scale a lot faster. And even though that feels really intuitive, it's surprising to me how many of the applications of AI we're seeing not have that final feedback loop. You have feedback while you're interacting with it if you're prompting and going back and forth. But a lot of these tools where it creates a product, the product is then stagnant and separated from the AI itself. And so you never know, once it creates this text or this document, it may not get feedback that's leading to that continual improvement.
I'm so glad you brought that up. One of the big things we've done with Synthesis AI is primary application for that now is search, right? And so when people get a search result in our product, they can rate it, but they also can leave feedback. And we see all the feedback that all of our clients do across all of their users. What's interesting there is you can start pulling back and seeing patterns and you can discern some of that underlying, whether they're business logic or domain-specific assumptions, like you find these turns of phrases that people use consistently that I think all that stuff was invisible. Before. It's just stuff that you learn working in the industry that everybody has as tacit knowledge, but it's never written down. But through the feedback, it starts to reveal. So for an example, that's silly in our industry would be people want to search to find out how many licensed architects do we have in our company? Well, they're actually called registered architects, not licensed architects. So there's like these little, I mean, there's a bunch of other, like how many jobs have we worked on? Well, when an architect says that, they mean projects. but the AI doesn't know that. And so only by looking at that feedback really at scale do you start learning how to build some of that domain context in. Yeah, that's interesting. I'm curious about your podcast example. I want to go back there. So if the job to be done of somebody listening to a podcast about a research paper is literally to just digest the contents of the paper because they don't want to read, then I could see how that— I know it sounds like a NotebookLM from Google type thing, right?
Exactly.
It is.
That's exactly what it is.
Where you kind of feed it the paper paper and it creates a podcast out of it. Yep. I wonder if there's a podcast, if the role is, sure, tell us about the research in the paper, but I wanna push you on some of the conclusions, or what are some of the implications of the research? It doesn't seem like it would fit that use case for learning.
I think certainly would have more limited ability there, I would imagine. Again, I think that the synthesis part is really good. The ability to say, hey, I have too much to read, but I have a long drive, convert this essentially to a more engaging audio format. I mean, that box has been ticked and I think the ship has sailed. That is totally doable. And to the extent that that's all we were offering was make existing knowledge more digestible, mm-hmm. That gets supplanted by AI quite easily. There's certainly some ability to probe and extend. I have some colleagues here at the Applied Physics Lab and I went down to, to give a presentation there and one of 'em came in. These are all engineers and physicists and folks who have much bigger brains and higher security clearances than me. So they work on really cool projects. One came in and said, I was curious what we might ever collaborate on. I think your work is kind of interesting. So I asked ChatGPT to look at my published works and your published works and identify 5 new research questions that we should explore together. And what I'll say is none of the research questions were inspiring. But they were all tractable, meaningful. I mean, you could start working on any of those and they did a good integration and extension there. And so I do think there's some ability for it now to probe and say, well, how would this work in architecture, for instance, and be able to kind of do that. But my sense, and I am certainly no expert on the state of AI right now, but my sense is it can't, go further than 1 or 2 steps out of extension in the way we would want to dialogue and hypothetically bounce a ball back and forth that, that takes 5 turns back. And now we're talking about some imaginary thing in the future that we've sort of loosely sketched out. It's not taking us all the way there yet. So there's still room for that discourse and discussion, but that initial synthesis and maybe that first point of connection or first point of extension to say, hey, here's a jumping off point for the conversation. I think it's pretty good at that.
Yeah. Yeah, that tracks. I'm curious to see, like, when you're working in your various different research efforts that you have going, are any of them AI-centric in that it's like, I wonder how AI is going to affect XYZ, or are you chasing other problems and AI is slipping in through the side or the back door?
A little bit of both. I'd say where I'm looking at AI the most is actually in the healthcare space. But less around the application of AI, but more perceptions around it. So with some colleagues we've been looking at what will AI mean for the expectations that patients have of doctors, right? AI has sort of been promised to be this big burden lightener for physicians. You don't have to write your notes anymore. This is going to handle automatic prescription refills, all these things that should free up time, therefore increase your wellbeing, decrease burnout. But if you look at it from the psychological perspective, what AI also does is it raises the bar on outcomes, right? So, you know, if I'm a patient and I know my doctor's got access to a great AI-powered tool and they misdose my medication or they miss a diagnosis, I'm more likely to hold them to account because I'm thinking not, well, you know, Every system is imperfect. Everyone makes mistakes. I'm thinking you had access to the whole world's knowledge at your fingertips in a really easily digestible way. ChatGPT can tell me anything that I ask it, and yet you could have asked it and you didn't, or you asked it wrong and you got the wrong answer. So what we're seeing is physicians potentially actually feeling more stressed and more burnt out because now the expectation on them is so much higher, almost to these sort of superhuman standards that now they could never be wrong because they've got access to this great resource.
Do you use the term standard of care in healthcare?
Yes.
Yeah. So same thing in architecture and engineering. So it's raising the standard of care expectation from clients.
Correct. The standard of care and then the willingness to tolerate error, right? Which, you know, sounds terrible to say, but there is always error in any system and there's certainly error in the healthcare system. And so the more tools and resources we have, the more there is an expectation that we shouldn't make mistakes, that things should get caught, that things shouldn't get through the cracks. But what's very interesting to me is it, it comes down to, to how we're treating AI within the healthcare space. Is it a tool or is it sort of a decision maker? Right? Because ironically, if it's a, a tool, you might imagine that, so, so AI hallucinates, it comes up with the wrong answer, it advises the doctor incorrectly, right? That's in some ways no different than the MRI machine at the hospital being miscalibrated and not showing a mass when there was one, right? If the MRI machine is miscalibrated, people aren't going to sue the doctor. They, and sorry, and there was an error or something bad happened, right? As a patient, I'm not pinning it all on the doctor. I'm thinking I'm going after the hospital or maybe the manufacturer of the MRI machine. But with AI, what we're seeing is that people go, well, the physician should know not to believe the AI output when it's wrong, but the physician should also know to follow the AI output when it's right. And we're essentially ascribing the blame to the doctor in the same way we would if that doctor had asked like their knucklehead colleague who hallucinates from time to time, Like, oh, you asked Chris and Chris gave you a really terrible suggestion. You should have known not to listen to him.
Right.
Right.
But this time Chris gave you actually a good answer and you should have listened to Chris.
Right. And so it's quite odd, in fact, that we're humanizing AI in a way when we draw these expectations around what we want physicians to be able to do with it, right? Where we're not treating it as a tool where if the tool's wrong, the individual user's off the hook. 'Cause how would they have known that the MRI machine was broken or that the AI gave them a decision made off of made-up studies. That happens all the time with these predictive algorithms, right? So how we think about it, how we perceive AI, actually has a huge impact on all the outcomes that we care about in healthcare. Our ability to deliver high-quality care, the wellbeing and stress and burnout of the people who are delivering care. So I've been fascinated less with the nuts and bolts of, is it working well? What do we do with it? What can or can't it help us with? And more just how are we even thinking about this as a, a new tool? Because that's one area where it does feel kind of different than computers or the internet is, you know, when we got digital CAT scans and things like that, I, I don't think anybody then expected doctors to perfectly calibrate and know exactly when to trust that and exactly when it might be having an error. We, we sort of held the manufacturers to a standard to say you need to manufacture it within these tolerances. I'm seeing a different conversation happen around AI where much more of the burden is being shifted to the sort of end user, in this case the physician, to know, all right, is this useful information or not? And so it feels like it's a, it's a very different perception there.
I was listening to a podcast like 3 or 4 months ago, so I've forgotten the name of the product. I'm sure you know what the product I'm about to talk about is that makes all of the literature very, very searchable by doctors. And so one of the kind of core arguments was, look, I guess it raises questions about what it means to be an expert, because in healthcare, as in architecture, engineering, you've got a certain number of continuing education hours you have to do. You want to stay up to date with the latest, but the volume of research that's getting published just accelerates every year. And no human could stay on top of just within their specialty, let alone a generalist who has to cut across specialties. And so these AI tools are able to, like, you start describing symptoms and you can kind of cut through the literature to get better answers. But what does it— what does that mean to be an expert now? Are you even trying to keep up with continuing education in the same way? Are you just relying on searching when it's— when the thing shows up? What are you hearing from people?
Yeah, there's the normative answer and the descriptive answer, right? Which is normatively, we still seem to be expecting everybody to know this stuff. In their heads. If you go to medical school classrooms right now, with very few exceptions, they're still testing students on drug interactions, right? Drug interactions is one of the clearest use cases for AI or even just computer-aided database stuff. This drug and that drug in these doses cannot go together. It's pure synthesis, right?
It's deterministic. It's like you don't even need—
Exactly. Yeah. Right. And yet we've had tools like UpToDate, right? Which is this searchable database of all the literature. Right. We've got all sorts of stuff and we haven't seen this shift towards changing the, the educational paradigm. We're still sort of treating the expertise that needs to get developed as the same expertise from 25, 30 years ago. This has been sort of a personal crusade of mine to try to get like drug interactions and the Krebs cycle, which is sort of how we metabolize something. I don't even know. Right. Comes up very rarely, and yet every medical student tells me they spend months learning this thing. All this sort of AI perfect information should be out, and what should be in is how do we collaborate? How do we work in big systems? How do we do all the sort of teaming organizational kind of stuff? You know, which admittedly I'm biased because I study those things and I think they're really relevant for how we deliver care. And yet curricular-wise, we are not seeing systems stuff being taught, and we still continue to see these sort of essentially rote memorization of basic facts. So that's on the normative side, right? Normatively, I, I don't know that we've changed all that much.
And do you think that's also true with continuing education for a 20-year physician? It's still read things, get it into your head.
Absolutely. When you look at the things that people get awarded continuing medical education credits for, it's attending lectures. It's reading articles in one of your field's journals and submitting a little questionnaire at the end. It's getting emailed questions and responding to them. It's just quizzes, right? It is declarative knowledge. And again, this may be back to this issue that normatively we don't know another way of thinking about what it means to be an expert, or certainly a way of assessing someone's expertise that is not in this sort of declarative knowledge. Paradigm. So we may just be stuck in an old way of thinking about it. Descriptively, I mean, what you see actually happening is doctors are using ChatGPT, doctors are using Google and YouTube, and they have been for quite a while, right? Because again, you see certain things rarely enough and there are good resources that you may have somebody who is covering, right? Let's say they're covering in an emergency department or something like that, or they're covering In training, it's not uncommon to cover in very broad services. And so yeah, you're a surgeon, but you haven't done hand surgery, right? And you've got somebody who comes in with a broken finger and you just need to splint it and set it, and then they'll see the specialist tomorrow. You haven't splinted a finger since medical school. What are you gonna do? You go back to your office to get materials, quote unquote, and you pull up a YouTube video on how to splint a finger. I have— this is a 100% real-world example that I have seen happen where somebody just goes and they look it up. And again, they need to have some baseline level of knowledge and expertise. When you and I watch that video, we don't pick up the same things that they pick up, right? But the sort of Rolodex of full, perfect information, I think, has never really been the way that medicine gets practiced, especially long-tail stuff that you don't apply very often. Absolutely. And even the, the quick comment stuff develops into that sort of tacit sense of, ah, this is right, that's wrong. Let me go check out the specifics later. Or this is life-threatening. This is not, this doesn't feel right. I'm gonna chase this down a little bit further. This, I feel like I could send somebody home with that. It, it is built on that sort of base of experience that then you get the specific care pathway or the specific treatment that you want to take. After you go back and double-check up to date to see what the current recommendations are. You know, they need to take this medication, how much of it? Nobody really walks around with that in their head anymore. They're going back and looking that up. So I think descriptively we're seeing a lot more uptake and, and use in that regard.
This is going to sound like a silly question, but so the doctor that goes back to his office or their office to look up how to splint a finger.
Yeah.
Is that a, is that like YouTube YouTube or do, do most hospitals have—
Is there a medical YouTube?
Do they have a medical YouTube or is there like an intranet knowledge management system where like they kind of show the Johns Hopkins way of splinting a finger? I'm kind of curious the state of how those things materialize within medical.
Do you want the uncomfortable answer or the—
No, no, I want the uncomfortable answer.
The uncomfortable answer is it was YouTube. Regular old YouTube.
Regular YouTube. Yeah.
Yeah, in part because, as you know, the utility of a knowledge management platform is at scale, right? When it can get lots of different perspectives and lots of different interpretations on things, right? So anytime you bound it and lock it down, you run into this issue that is every hospital going to record a video of somebody splinting a finger? And some hospitals will have preexisting differences in their quality of practice. What this hospital tends to do and what that hospital tends to do may be better or worse. And so you actually kind of want the general all-purpose YouTube where the person at the very best hospital posts a video of their new technique or procedure, and that's how best practice gets disseminated. Now, certainly there are less alarming avenues that that's happening. Medical conferences, right? Academic journals, official kind of training courses. So don't get me wrong, it's not that everybody's just going to YouTube. To do this, but all else equal, if you've got an expert in your hospital who gives a recorded lecture on a subject, are you doing a disservice to not put that on YouTube and make it available to others at other institutions around the world? We actually saw this. I, I wrote a little bit about it as one of the challenges we had during COVID was there were physicians on the ground in Italy and China early on, who had vast amounts of firsthand knowledge and experience of treating COVID patients. And yet physicians here on the ground could not get ahold of them because it needed to go through formal channels, sometimes through professional associations, sometimes through sort of nation-state level diplomatic relationships where there were some blocks put in on what physicians could say about what they were seeing because it made a country look good or look bad. At that point, you kind of wish that somebody could have uploaded a video to YouTube to just say, hey, this is what we're seeing right now. This is what we've tried to be able to treat it. The mechanism to put it through a formal communication channel, again, like any other formal knowledge management system, starts to strip away a lot of the richness and context and nuance that you might want to preserve. And so, Yeah, as scary as it is, regular YouTube, regular Google, regular ChatGPT are often— it's seen by many of my physician colleagues that I've talked to as kind of more useful than the locked-down, sanitized knowledge management or even AI sandbox kind of system. So, you know, full disclosure, at Johns Hopkins here, we have our own AI sandbox that is the only one where approved to use for official work functions because it is locked down and it knows it's not pulling data in from Reddit where it's gonna get bad information. But the problem is it's also not pulling data in from Reddit. And so it's getting very stilted, very sort of one-sided interpretations of things that are very good at certain aspects, but very bad at others. Right. You know, and so it's, yeah, it's, it's still a little, tricky.
I'm not a healthcare expert by any means, but my kind of understanding of the industry a little bit is it's, it's like ours in that it's both collaborative but also competitive. And so there are hospitals or systems that try and differentiate themselves on cancer or cardiac care or whatever the thing is. And they would have the Johns Hopkins way of doing X or the Cleveland Clinic way of doing Y. And so on one hand, you want to make sure that people are understanding our specific way because it's our secret sauce, but also trying to share knowledge across. And I'm— I just find that— is there a world where you go to find the Johns Hopkins way to do certain procedures or processes, or is that really more seen as we're trying to lift all the boats of the whole healthcare industry at the same time?
I think it's really interesting because I think to the extent that there are sort of competitive status benefits where it helps you to look good, that actually feeds some of the YouTubification of this in that when the University of wherever, right, University of Michigan, University of Chicago, Johns Hopkins, whatever it is, when they put a video of their expert online and then everybody shares it around like, oh, hey, if you're going to do this procedure, check out this lecture from Dr. XYZ at University of Michigan. They put it up online. That's actually net status gaining for them. Because again, it is a weird competitive environment that there's certainly local competition to a certain extent, but nationally, a patient, other than for the most rare of conditions, is not shopping hospitals at a nationwide level. So if I'm in Cleveland, I'm not coming to Johns Hopkins. And so it doesn't hurt Johns Hopkins competitively too much to make some of their resources available to Doctors at the Cleveland Clinic and to share their knowledge in that way, but can actually sort of boost the status representation.
And so is the status boost more at the physician level or more at the institutional level?
Really good question. A little bit of both. A lot at the, at the individual level, I would imagine, right? Becoming known as the go-to person for X, Y, or Z, but certainly the institutional reputation is there as well.
It's just interesting to think about incentives that way. And then incentives also from a— does the kind of going to YouTube as a medical expert and putting your content there from a liability perspective, is that a concern that you've heard?
Yeah, it's definitely a consideration, right? So again, to be clear, what you're not finding on YouTube is— oh, sorry. As of 2 or 3 years ago, you wouldn't find, here's this patient that I'm treating, let's watch me do my physical exam. I say as of 2 or 3 years ago, cuz there's been this whole explosion of mostly from the patient or the sort of care recipient perspective of going to see a chiropractor or a dermatologist to pop my pimples or whatever. There's all sorts of, you know, crazy videos out there for that stuff. But when it comes to sort of medically oriented content. A lot of it is sort of lectures and pre-prepared things.
Yeah.
But, uh, a number of years ago, I mean, this predates AI certainly, I did a small study with some colleagues in surgery of a Facebook group for robotic surgeons. And it was this fascinating group because robotic surgery, certainly at the time, this was maybe 2018 or so, but still continues to be a fairly niche practice, right? You don't have tons and tons of robotic surgeons around. So you may be the only robotic surgeon in a 50-mile radius, certainly a robotic surgeon doing one particular type of procedure. So getting that informal knowledge, feedback, advice, mentoring can be quite challenging. But what these surgeons figured out pretty quickly is that as they're doing the robotic surgery, they use a camera for it, and that camera can record. And one of the beautiful things about doing surgery on the inside is most people's insides are HIPAA compliant, right? We're not, you know, like, I can't identify you by your spleen.
Right.
So it's a much lower barrier to then be able to say, hey, I'm going to post this video of me doing my robotic surgery procedure on Facebook. And there's going to be a community of a couple thousand robotic surgeons around the world who, because they're in different time zones, essentially are a 24-hour think tank to basically just throw the video up and say, hey, what do you guys think of the way that I approached it? Or I tried this new procedure that Dr. So-and-so had recommended on here last month and it worked great. Look at how it went. Or, hey, this is what I was looking at. What would you have done here? And we found that it, it was actually quite active. I mean, people would comment pretty regularly. They would pass feedback along back and forth. Now it was a private group. So you had to be invited in to be a member. Presumably you had to provide some sort of proof that you were actually a robotic surgeon. And there have been some legal protections afforded to groups like that, that essentially they consider it an online version of the doctor's lounge. And just like what we were talking about with AI, asking your friend, what would you have done in this situation? Does not then make them liable or you any differently liable for the outcome. Of the surgery that you performed. And so again, even pre-COVID, pre-AI, we were seeing this sort of, I don't want to say a backdoor use, but like this sort of unintended use of some of these large-scale social media platforms.
So community of practice basically puts on Facebook for learning kind of, and this is, I'm going to use this word intentionally to get you into it, but like this is vicarious learning. Through your peers. Yeah. Right.
Yeah. And I think there's a question of would that be better? Would we all feel better? And, and or would it be better, which are two different questions, if that forum was hosted by the American Surgical Association, if it was some website that they logged into. What I can tell you for sure is it would get used a lot less if it was hosted by American Surgical Association instead of a group on Facebook. I think it's another login, it's another access. Who's logging on to American Surgical Association every day? Facebook, it pops up. I was a member of this group for a short time while we were studying it. It shows right up on your feed. There's nothing like going to like birthday party, new kid, spleen, colonoscopy, you know, uh, it makes for a very interesting scroll. But on the other hand, it's right when you scroll a very common thing. And as we know with the research on communities of practice, the more accessible and central to your day-to-day interaction they are, the more you're gonna use 'em. So again, yeah, I think that this question of do we do it on YouTube and Facebook and Twitter was a, a huge hub for a long time of academic knowledge sharing and vicarious learning as a, a big community of practice. It has ceased to be that since the move to X, but these sort of generic platforms I think have a real appeal. But what they lack is some of that kind of formal boundary setting and quality control perhaps that you would see if it was an institution-specific resource.
The group of robotic surgeons, I'm curious to know how the kind of Facebook group— I'm sure they had like formal conferences and papers and journals or stuff like that that they would also participate. And I'm curious how the kind of more informal Facebook thing compared with their kind of more formal organization as a group and what the impacts of each of those were on each other?
Yeah, we didn't study that directly, so I, I'm speculating here, but I would imagine based on what I've seen elsewhere, they're sort of complementary approaches, right? That, that informal interaction preserves a little bit of the messiness and the sort of immediacy that you wouldn't get by the time it's all cleaned up and made into nice slides at a conference, or certainly by the time it's through peer review and documented in that way. So I think they're symbiotic in that sense that somebody publishes a peer-reviewed randomized trial of a new procedure, and then you get a bunch of people trying it out and posting their results to Facebook and maybe then refining it. Like, oh, okay, so they found if you approach from this side instead of that side, there's less scarring and less damage. Great. Well, I've been trying to approach from this side, but what do you do with a patient who has blank, you know, is overweight or underweight or has this preexisting condition. All those little kind of messy nuances of how do you actually do the thing. I think those get sorted out quite nicely on Facebook in the same way we might expect them to get sorted out face to face in a, a more traditional office environment, right? Where there's that classic water cooler vicarious learning that's happening.
Can you Can you talk about vicarious learning?
Yeah. All of this, I think, really boils down to our attempts to learn from other people's experiences, particularly in healthcare, right? Other people's mistakes. Like, trial and error is sort of our least preferred way of people practicing medicine. Note I said least preferred, not never preferred, right? It is— it's the last option. But yeah, again, if you're one of the only robotic surgeons around and you've got nobody to talk to or ask, you're on a learning curve, but you're on your own learning curve. You may end up having more complications or more issues than you otherwise would. In fact, there's some evidence to suggest that there was a great study by Melissa Valentine, Amy Edmondson, Sarah Singer, I believe, and a few others in, in JAMA a while back where they found that physicians who were in group practices scored better on their recertification exams, you know, controlling for a bunch of other factors, but essentially having other people around who you could learn from.
In the same kind of specialty, is that what that—
Exactly.
Yeah.
So if you were a solo practitioner, so you're a cardiologist in an office by yourself, or you're part of Parsons and Co. Cardiology where there's 5 of us working together as cardiologists. If you were from a group practice, you tended to score better on these recertification exams. That was sort of the gist, and I, I may be misremembering the specifics, but the general sense was having people around helped you out because you learned lessons from their experience. You could bounce ideas off of them. It, it all came back to this, how do I glean a lesson from an experience that you've had so I don't have to wait to have it myself? Or if I do have it, I'm one degree more prepared. And that's really this idea of vicarious learning, learning from the lessons of another's experiences rather than our own. And I've studied this in a lot of different contexts throughout healthcare organizations. We've seen it in other organizations as well, but this ability to draw some insight or some lesson from an experience that somebody else has had just arms us with a broader repertoire of experiences. So if I'm that robotic surgeon in the middle of Wisconsin, nobody else is around, but I'm on this Facebook group. And so, yeah, yesterday I happened to watch your video where you showed a thing that you had done to deal with a, a certain kind of issue. Great. I kind of log that away. It's not that I'm learning any real declarative knowledge or gaining anything, but in 2 weeks when I see that same challenge, I kind of go, oh right, I remember that guy went around to the right instead of to the left. You know, maybe I'll try that. And it just kind of gives you these little reference points or frames that you can refer back to rather than figuring it out on your own where you go left and you realize like, oh, okay, that wasn't correct. So next time I'll know. Well, it's hard in an era where the next times are fewer and farther between. And I think we're definitely seeing that across all industries where things are getting so specialized and so complex in the world of knowledge work that it's not like we're getting another chance right away to try it again or to make 10,000 of the same products. You don't build 10,000 of the same building to be able to learn from your experience on that kind of rapid cycle. This may be the one cantilever design you're doing for the next 10 years, and it's going to be a decade before you come back around to another project like that.
Right.
So, yeah.
Or even once you do the design, it's 4 years before it's built. So the feedback cycle is super, super long.
Yeah. Thanks for bailing me out there. I said cantilevered and that was as much as— that was good architecture as I could work in there.
I think that I'm impressed with when it comes to the vicarious learning. I'm curious, like how much of this tends to be serendipitous, like if X person happens to do this thing versus kind of more orchestrated where, you know, we want new people to see a bunch of different stories from a bunch of different people. What's your experience there?
I think again, the unsatisfying answer of yes, both, right? That both of these have a big role to play. There's so many case examples that we could cite of the serendipitous learning examples that pop up where somebody happens to share a failure of a project in making glue. That then solved somebody's problem of needing to put bookmarks on very thin, brittle Bible pages. That net result is Post-it notes, right? You know, we, we've seen all those kind of serendipitous ones, but I do think we're seeing organizations also try to orchestrate this a little bit more. I mean, Google's classic phrase of engineering serendipity, right? You know, wanting people to bump into each other at work, right? Rotational programs that intentionally take people all around the organization and expose them to different experiences. So I think intuitively we kind of know that if we can create the opportunities for these different interactions, then that creates opportunities for us to learn from a broad set of others.
What, what's the role of AI in vicarious learning?
It's an interesting question because in a pure sense, it, it is sort of the ultimate instance of vicarious learning. right? When you look at a large language model, it's essentially saying, okay, when everybody else has had token A, what token has followed it? What word most often follows this? I'm gonna learn, okay, well you said only this, you said only that, you said only this. All right, I'm gonna guess it's probably this is the next word that I should predict here, right? They are the, the ultimate tool in that sense and scale kind of beyond what than any human would even imagine being able to do. Again, what gets lost though is sort of the context and the ability to understand how you're drawing that knowledge out and then where you're placing it in, right? So one of the places I studied vicarious learning was with these air medical transport crews that fly patients by helicopter. I studied how they gleaned a lot from one another's stories of their experience.
Like on the flight back or like in the break room? Is it the same?
Yeah, in the break room, just hanging out, you know, talking like, hey, you won't believe what happened on, Yesterday's shift, right? I listen to all those stories. I am not prepared in any way to treat people in the back of a helicopter. There's a context that, that is necessary there. And that's where I think AI can be both a blessing and a curse in that you could ask AI now, how do I transport this kind of patient in the back of a helicopter? And it will find as much as is out there to be able to help you with that. But it doesn't solve the problem of I still wouldn't know how to do it. It can pull it for us. So it solves the vicarious part, but not necessarily the learning part. It brings the information to me, but it can't help me learn from it in a way that a sort of human-to-human interaction often can.
Do you think it would help though, like one of these, the transport nurses who had 5 years of experience, if they could ask AI about edge conditions or things like that? They have a certain baseline of context to then know how to use that information, right?
Absolutely. And I think that's true across the board with vicarious learning, that the more sort of baseline knowledge, or you could think of it as absorptive capacity that you have as a, as an individual, the more you're going to be able to glean from that, right? You know, I would sit with these flight nurses and we would both listen to the same story and they would pull out 9 gems or takeaways that they're going to do. And I was like, okay, so you were flying. That's as much as I got out of this, right? I didn't know the terminology. I didn't know what you were thinking about, what things you would've been worried about or wouldn't have been worried about. So I think, yes, there will still be room even in an AI-enabled world for people's own expertise to help shape what they're gonna do with that. The other problem is AI can capture what's out there, but there's still just a lot that isn't out there. Yeah. One of the things that I studied with these air medical crews was how often they use just old school storytelling to learn some of these lessons because what they did write down, what got documented, back to our earlier example, is a medical chart, right? So they were defensively documenting, or they were documenting the cleaned-up version of the story that was good for billing and liability and all these other purposes where learning is not necessarily the primary purpose, right? So they leave out, well, I thought of A, then I thought of B, then I thought of C, then I realized B was better than C, but actually that led me to D. Right? The chart just says we did D.
And by storytelling, then you're meaning oral, like that's it. Like not digital.
Yeah, it is just shooting the breeze. Yeah, exactly. And so I think it, again, where I think the world will be different, not necessarily worse or necessarily better, is now the question will become how do we make some of that stuff exposable to AI? So if you do audio recorded stories, if you record all your meetings, if you record everything and feed that to AI, presumably then, yeah, you could get some of these things in there. Now, whether you would be able to do that, whether just logistically the reams and reams of data that that would create can be stored anywhere, whether that would make everyone stop telling stories if they knew the camera was on and they were being recorded. Those are all, I think, fascinating questions that we need to sort out. But at just a logical level, yeah, we could get all this stuff into AI. Theoretically. It isn't there now, though.
I mean, we don't have to stick with the transport nurses, but like, how does it— like, if you're a first year, the way that you got brought up, maybe when you did this research versus the way a first year will learn how to become a transport nurse in 2030. I guess this is getting pretty speculative, but I just— I wonder if I'm combining this idea with a standard of care idea you raised before. Where if it will be acceptable to just have you learn one flight at a time, and if you happen to get across the certain stories that maybe help you build.
Yeah.
Because that apprenticeship model, like, this is certainly a conversation in our community, is like that apprenticeship model of one project at a time and what you happen to hear over the shoulder. That's not scaling and it's not working. We get impacted maybe more by hybrid work than people in a hospital would. But I'm curious your thought about what does it mean to develop? Someone who's gonna be an expert.
Yeah. So the hybrid work thing is really interesting. Let's not lose that. But I think the broader answer is absolutely. I think one of the use cases of AI that I find most compelling is the ability to create simulated environments and experiences, right? I think we have long known the value of learning through simulation and getting to do trial and error in a low-stakes context, all those sorts of things. But one of the biggest barriers was always just the mechanics of creating the simulation. It's a lot of work to even just write a, a tabletop written scenario exercise, right? Let alone actually create realistic client specs for a building and then have somebody create a design and then evaluate it and do all that. AI is simplifying all of those pieces of it. And so I think, yes, training will take advantage of AI almost undoubtedly. There will be opportunities to supplement real-world experience with simulated experience, and it helps solve some of the challenges that we see. I mean, in all knowledge-intensive work, right?
I have a crossover story between our two worlds to share with you.
Yeah.
So one of our clients, Todd Anderson from Boulder Associates, spoke at our conference this year, and he built a simulator to help architects who had never worked in a healthcare setting before understand what it's like to go into a user group meeting full of a couple busy surgeons and hospital administrators. He actually built different roles and backstories for these people and would go in and you're trying to understand what they need or pitch a design and you're just getting grilled. How does this apply to me? And so it was like getting reps just even in these human-to-human, you know, kind of interactions.
Yeah, I like that. And as someone married to a surgeon, you do have to practice sometimes presenting to them if you want, if you want to get an idea actualized. But yeah, I mean, 'cause again, this is a longstanding problem, right? We saw when I was doing this research with the air medical crews, they would practice on simulator mannequins, but one of the things they would try to do is they would have one of their colleagues run around the corner and be the distraught parent of the patient. 'Cause they're doing this out sometimes on a car accident scene, right? And it's like, yeah. The treating of the patient is one thing, but the thing you might not be prepared for coming from a role as a bedside nurse is that somebody's gonna run up and grab your arm while you're in the middle of trying to administer this medication, or it's gonna be super windy and do all these things. And yeah, you could wait till the first time it happens for real, or they could put you in a basement room with a mannequin, crank the fan on, and then have somebody run around the corner and try to grab your arm.
Mm-hmm.
That you could find ways to simulate these things in order to prepare folks better for the first time they encounter it in the real world. Now, until AI powers the robots, there'll still be some sort of human bit that we need to do. But I think a lot of these scenarios and things you can do in a, a fairly convincing manner with AI, and they're more realistic than some of the old school prerecorded video cases and things like that. So I, I see huge upside there.
No, I agree with you. I, and I think that one of the challenges also is that Oftentimes the people that need to write these scenarios are the busiest people in your organization. And so if they can move to be more of an editor versus the writer and they can interact with a thought partner to help them just explore and get, you know, it's the blank page thing. They can even just talk and say like, look, I need to write a scenario. This is roughly what I want to have happen. Like, help me refine this thing. I've done a lot of that. And it is— I think it is quite helpful in that way. Yeah.
I think it also, you know, will allow these things again to be scalable and adjustable in ways that stories might have been too limited for. Because the story is a recounting of one event, right? And the inevitable question anytime you tell a story is like, oh, well, what if you had done this instead? And if we can start building out some real high fidelity simulations based on underlying characters with models and motives and things like that, you could get to the point where you could actually try something different and see how it would've gone. And again, none of this is new, right? You, I mean, Apprenticeship has included simulation from day one. They're not an either or, right? You do practice runs, you do mockups, all that sort of thing. But now we might be able to do it just at a scale that that we wouldn't have thought of before and go back and switch out actions and things that before you might have been able to reflect or debrief or talk about it afterwards, but you couldn't really go back and just get a second rep in because who's going to sit there and have the exact same pretend client interaction with you 8 times to help you prepare for your first pitch meeting? An AI bot will. It'll have it 8 times and you can say, you know, be in a bad mood today, right?
Be in a better mood. And on Saturday and whenever you want it to be. Right, right.
So I think that will be a, a real scaler. I don't wanna lose though, the point you brought up about hybrid.
Oh, sure.
Yeah. Because I, I do think that's hybrid is gonna be a fascinating, um, not going to be hybrid is a fascinating sort of moderator of all of this. I'm doing some work with colleagues led by Kevin Rockman at George Mason University that is looking at the impact of hybrid work arrangements on our relationships at work. And so learning vicariously is essentially a core part of our workplace relationships. But what we're finding is that hybrid has this really unique effect on the way that people intend to and then engage in interactions in the workplace that is different. It's both different from fully in-person, but also different from fully online. So it's not just Oh, the more online you go, the worst. Hybrid is actually the worst, full stop. And in-person and online are both the worst, significantly, from what measure? Yeah, so good question. The worst in terms of disrupting the pattern of relationships and interactions that we see happening. We're calling this, this idea of sort of social compression, that essentially what everybody in a hybrid work schedule does is, oh, I need to ask Chris something. We'll both be in the office tomorrow. I'll wait and ask him then. Scaled to the entire organization so that when everybody's in on Tuesday and Thursday, it is this chaotic free-for-all of, hey, can I bug you for a second? That people all know they're going to do when they get there, but then hate doing while they're there because it's a big disruption. And meanwhile, their email inbox is piling up and they're not able to, quote, get any real work done. And so then they go back on Wednesday and they're totally depleted from the Tuesday interactions and they feel like they're just catching up and they get in this spiral where they feel like interacting with other people becomes this big burden in their role rather than part and parcel. And what was interesting is we did not see that reaction from either the people who were in the office full-time or from the ones who were fully remote because they know they're fully remote. And if I have a question for Chris, I'll ask him. Right now because there's nothing to then wait for. But this idea that we compress all our interactions into these in-office days and then try to not interact with people on our at-home days because we need to get stuff done, one creates this weird, like, now interacting with people is not getting work done.
Yeah, that's what I was going to ask.
It used to be kind of part and part, like, that was part of the work was interacting with people. We've created this totally different orientation to how we interact with others. And I think a big risk is that we lose vicarious learning in the process because we don't want to bother other people or we're already too overloaded with interactions on our in-office days. So when we're home, we think, uh, I could message Chris and ask him how he had done this, but it's probably just easier for me to Google it or figure it out myself. I'll just learn it on my own.
Or your mental model is Chris's head's down trying to get work done. I don't want to bug him now. I'll just see him on Thursday or whatever it is. Right.
And by the time Thursday comes around, you've got a 20-person queue outside your office waiting to ask you for all these things. You're giving curt answers, trying to kick them all out of the door and address your inbox. So we, we've been looking at this. It's very early stage kind of qualitative work, just understanding the change here. But what we find is that ironically, hybrid policies are often marketed as restoring some of the relationships and interactions that were missing when everybody went fully remote. But we find that in some ways they do the exact opposite. They actually deteriorate them more quickly because of this compression effect.
Have you found outliers like positive deviance in this firms that are doing hybrid really well that you could learn from yet?
We haven't yet. So this is small scale. We've looked in a couple organizations, but it was surprisingly consistent. And then we did a broad panel look just from a representative sample of people working fully remote, hybrid, and fully in person. And we found similar sorts of patterns. I have no doubt there are places that are doing it very well. Again, back to your question of serendipitous versus intentional. Interactions. What we are observing is that the compression is worse when everybody shares the same in-office days.
Mm-hmm.
Right? So we did not study this. I'm speculating, but one could imagine then that staggering the overlap where everybody does overlap with one another, but not everybody in on Tuesdays and everybody in on Thursdays where it is more balanced throughout the week. One might imagine then could attenuate some some of these problems where it's not that everybody's trying to have their meetings on Tuesday because that's the day we're all in the office. Half of us are in on Tuesday, then half on Wednesday, then the first, you know, or whatever, a different half are in on Thursday and a different half in on Friday would sort of spread the load a little bit perhaps. So again, we haven't studied that directly, but I would imagine there are organizations who are out there who have thought more intentionally about how they're doing their hybrid in ways that might be addressing some of these concerns.
So we've touched like— I'm trying to identify a few themes that are kind of running through our conversation. I feel like one, one is it kind of— I keep asking the question, what does it mean to be an expert going forward? And we kind of hit on some different angles of that. And also, how does one become an expert? And those two things, I don't know if this is, I don't know if you've thought about it this way, but those feel like those are gonna change a lot more in the next 5 years than they've changed, the way I say it, is in the last 25. And I guess the third thing, so what does it mean to become an expert? What does it mean to be an expert? And then what does it mean to be a learning organization? And that's responsible for development of people and that has expertise.
I think the becoming an expert one is probably the one that I lose the most. Sleepover.
Yeah.
Because I think the deal in, certainly in most knowledge-intensive organizations was as a new person, you traded menial labor for knowledge and experience and sort of wisdom. You got to be in the room for the client presentations because you made the slide deck or you were running the slide deck or you were taking notes or whatever it was. But that was in essence, how you also gained experience and the sort of wisdom of having seen lots of these things without needing to be the one in charge of the project or anything like that. That deal is gonna be broken now because, or, or not needed anymore because I don't need the intern or the junior associate to make the slides or take the notes or do all of those things. What I worry about is that organizations go, great, so I don't need junior associates anymore, but you need senior associates. And so it reminds me, if you know the comedian, he passed away a while ago, but Mitch Hedberg, who did these observational notes, right? He has this joke. He says, hey Mitch, do you want a frozen banana? No, but I want a regular banana later. So yeah. Right. The junior associates are the frozen bananas here, right? You don't want a frozen banana right now cuz you have AI, right? You don't need a junior associate when you have AI, but you still need a well-seasoned, experienced senior associate. And I think organizations are missing sometimes that the price they pay for those well-seasoned senior staff is keeping junior staff on for a while. They probably weren't producing worth their value the value we were paying them already, certainly in kind of high-profile, knowledge-intensive work, right? Juniors at a law firm or at a consulting firm, or I would imagine in architecture as well, right? We weren't really ever paying them for the work they were producing, 'cause the work they were producing was not of huge strategic value to the organization. Slides, notes, things like that.
Bathroom details is ours, you know, picking up red lights. Yeah. Yeah.
Right. Yeah, exactly. So now we outsource that stuff to AI. We're thinking, great, we get rid of the expense line of the junior associate, but that never was what they were there for. It was something to keep them busy while we were essentially just growing them into senior staff. We were waiting, we were nurturing, we were mentoring. There's a world in which we really embrace and recognize that and say, okay, great. So now you don't have to make slides. But still come to the meeting and sit and listen. And I don't know what you're gonna do, but I need you to be well experienced in 5 years. So I gotta keep you around for 5 years and give you that experience. I don't know what I get in return for that right now. We gotta come up with a new deal. But just to throw all that out by saying, oh good, well AI can replace the things that they were doing, I think misses the point of what they were doing there in the first place to develop that expertise.
I think I'll take it even a step further. I'm writing a newsletter issue I'm calling the AI and Expertise Paradox. And so at least in my experience with the kind of tools, like there might be a tool that can help you run a set of plans through and do quality assurance automatically, or run a set of plans through and understand code compliance. And at least in the current state of these tools, it still requires an expert to review. The feedback because there's going to be false positives, there's going to be misses, there's going to be, well, that's kind of true, but there's a new version of the code that's coming next year. There's all this expertise that's required. You can't put a junior person, have them run the plans through and just sign off and say, that's okay.
Like, right.
It's health, safety and welfare of the building is important. Like, we can't have these kind of errors and firms need to stamp them. So AI still needs experts, but those experts are retiring. Yeah, in a massive way in the baby boomer. What we see, and I don't know, I don't know how this one translates to your world, but maybe this is just for us. Fewer of those emerging professionals that are in the business want to replace the kind of deep technical expertise. They want to do sexy stuff like design, you know, and design technology and write software. And then that apprenticeship model that we've just been talking about is kind of breaking down. And it kind of feels like at the, at the very time that that AI stuff is emerging, The experts we need are going away and we don't have a clear pipeline for developing that next generation of experts. And I'm not sure how we break through other than becoming much better at being learning organizations and maybe making some of the investments that you're talking about. It's like, sure, from a pure P&L perspective for this year, this may not make sense to have those junior people, but an investment in our business for 5 years, 10 years down the line, Maybe we need to figure a different way of developing those people out.
Yeah, I mean, I think it circles back to this idea around simulations, around training, around how do we create more opportunities for folks to do that. And then on the back end, how do we create opportunities? How do we leverage technology in ways that allow some of these experts to maybe not all depart at the same time? You know, is there a way to stick around in a 20% capacity doing these spot checks and this mentoring. Way back when I worked with Dorothy Leonard on this case of NASA's Jet Propulsion Lab, where we were watching them go through essentially the same thing just 15 years ago.
One generation earlier, right?
Yeah, yeah, exactly. Because they had all the folks who had worked on the Apollo program and it's sort of grown up and they had an extreme version of the problem most organizations face, which is they only got to launch something to Mars once every couple years. So you put the A team on it every time, every— because there's enough time for the cycles and the consequences of screwing up are massive. The problem was they had an amazing A team and the whole team retired at the same time because they were all of the same vintage and there was no B team.
No fresh—
there was barely like a C, D, E, F, G team, right? We had A team and like some interns.
Yeah.
Because Essentially, there was just not this progression, and I'm exaggerating, of course.
Sure.
They had great folks moving up the pipeline, but they had this massive gap. And we were writing about the struggles they had. They tried projects where they gave earlier career staff responsibility for fake projects, essentially. Mm-hmm. Things that would never get launched. And that was good. It taught them how to do some of the strategic decision-making. But then what they found is it was very hard to integrate those folks back in cuz they went from being the project lead on this fake project to now being back to a subsystem engineer on the real show, or they got poached by Google to be the senior vice president of whatever made-up domain, but it was a cool project and a sexy title, right?
Yeah.
And so they were really struggling with it. And then on the back end, struggling, how do we keep the senior engineers around and engaged to be able to give that feedback, give that advice?
What did they learn?
Last I checked, still struggling. Um, you know, that, that these are ongoing challenges. What they have found is, well, one, the shift to private enterprise being involved in space, uh, actually sort of in perhaps an unsatisfying way for some of the government agencies solved for this problem that you had a lot of folks mid-career essentially go move to become the senior people at all these small startups. But I think finding tools and technologies to be able to give folks on the front end more experience, but on the back end more ways to share knowledge and input to try to help preserve some of their knowledge and some of that experience that they had gained. Searchable video interviews. I mean, all the way back to what we were talking about with YouTube, right? That if you're gonna have somebody do an exit interview of all the lessons and things that they learned, get 'em on camera for 3 hours and you've got a 3-hour YouTube video that nobody will watch, but an AI will scrape it and 100% it's searchable and it can be fed into there. And then yeah, do you need that person back half-time? No. Maybe now you need 'em in one day a week to give some advice and you can create a, a work arrangement that's maybe better for folks where you supplement some of them with AI and some with that personal interaction.
So I think this idea that retaining employees is a knowledge management strategy, I think it's lost on a lot of people because I think a lot of times technology comes to the forefront, but really just still having access, like you said, on maybe even a reduced schedule and shifting their burden to building the— I mean, you know, I think for firms to really take advantage of AI, they're going to need a really solid knowledge foundation in a way. And by that I mean digital knowledge foundation that they haven't had before. But it feels to me like those emerging professionals, what we would call them, are well positioned to interview those experts, develop the knowledge. Maybe they do watch the 3-hour video and try and make a tool out of it. Maybe the 3-hour video is a lot of stories and it's like, cool, like now we need to build standards and processes and things like that out of those stories. Yeah.
And again, I think it also helps solve for this problem of people are too busy to document and build the digital knowledge base, right? If you've got the middle, that is your active producers and the ones who are driving performance, you've got new folks and maybe more senior folks who could be focused on building that, that base underneath. And I think, you know, what's interesting is when you see organizations that have stood the test of time, I think you see some of those same kinds of practices in place. I mean, if you look at academia as the quintessential knowledge industry, right? You know, from thousands and thousands of years ago, and we have doctoral training, we have postdoctoral fellows, like they're not doing the, the sort of primary work of teach the classes, get the grants, publish the papers. We also have emeritus professors, right? Stick around and just kind of hang out and are available.
And you have TAs, right? Also.
Yeah. TAs. Yeah, exactly. When you, you look at, you know, I, I would argue the, the military is another one where they've invested quite a bit in some of these structures and practices and, and the idea that you're a retired general that's still a career path in the way that like, oh, I'm a retired CFO. Like that, that isn't really a thing that people continue carrying around as their job title the way that a retired admiral or retired general would. And I think it's, it's a recognition that there is this lifecycle of knowledge and there might be some ways to plug into it at other stages. And again, back to that, not missing the forest for the trees of getting rid of junior staff because we have AI to perform their sort of nominal functions now. I think recognizing that, yeah, we're gonna spend a bunch of money developing somebody on the front end and there is not much return to us for doing that in the short run. They are not producing right now, and yet we are gonna grow them into somebody who would be important to our organization. It's just a tougher and tougher sell in a competitive landscape these days, but I don't think it's any less true than it was ever before. And if anything, is probably more true now with AI accelerating all these things.
One of the things I've heard recently in a few different places is most emerging professionals today will manage an AI before they manage a human. And by AI, meaning agents, right?
Yeah.
And then I read about a computer science program. I want to say it's Stanford, but I'm not 100% sure that is building a course into the curriculum for, for computer scientists where learning how to manage agentically built software is now part— this kind of goes all the way back to the beginning and we're like, you're not hearing anything change in the medical education around drug conflicts or whatever the different— those could probably be done better by a machine. Let's just start moving the profession that way. And I wonder how much of this— I think you're right. Redesigning how we take a 0 to 10-year, what that journey looks like for somebody. Realizing that that old apprenticeship model is gone in your industries, in our industry, and what does the next thing look like?
Yeah. I mean, apprenticeship-heavy models also suffer the burden of being slow to change because you get a lot of that, well, this is the way I was taught. Totally. And so you, you sort of move one generation at a time, but some of these technologies are paradigm shifts, right? It's a little older now, but Matt Bean at UCSB has this fantastic qualitative study of robotic surgery and how it upended the whole idea of a surgical community of practice. So you used to literally move from the periphery to the core. You might hold the leg, then you might be helping retract, then you're assisting, then you're doing the surgery, right? So you had degrees to move into expertise and, and participation there. With the robot, you're either driving or you're watching on TV. There, there is no in-between. You either are controlling the instruments yourself or you're watching on a screen with everybody else. And so it created this fundamentally different training need. And yet the training model stayed the same. There was still this apprenticeship model of you move from a first year to a second year to a third year resident to a fourth year. And what he describes are these processes of shadow learning where essentially people had to go around the existing training paradigm.
I thought you meant shadow, like shadowing somebody. You mean like literally like off book?
Oh no, I mean off book in the dark. And it was things like going in in off hours to use the robot, right? To get some practice. When things were available to them. So overnight cases where they probably wouldn't call in one of the attending surgeons, some of the senior residents might say, yeah, I'm able to do that. I'm confident doing that case when they might not have done it. Before, right? Having to sort of essentially circumvent the training process because it hadn't changed. And it was partly, at least in my opinion, because the folks who were at the top of that training model had no mental representation for how to become a surgeon other than what they went through. And so it's just been really hard for everybody to say, okay, well, what does residency look like now if the tools are totally different?
Yeah.
And I think it'll take a generation who figured it out through backdoor shadow means and got there, and then they'll be able to say, okay, here's how we're gonna do things differently. And we're starting to see that now. You have a lot more simulator labs. You have a lot more, right?
So if you, the only way to do it is watch on TV or do the thing, then simulating the thing.
Exactly. Do the thing, but on not a real person.
Not on real person.
Right. That, you know, so again, and they're not, it's not rocket surgery, right? There are obvious answers like that, but it's just that requires reorganizing the surgical training schedule, right? Because right now they're supposed to be in the OR watching and it would take somebody going, okay, this is not a good use of my time to be the third person standing here watching this screen. I should be in the simulator lab doing it. It requires building a simulator lab, right? Staffing it, doing all these things. So it's a, simple answer, but a non-trivial change to adjust the training program.
And it almost has to be designed, I would think, in concert with the novices.
Yeah.
Because it's been so long since an expert has been a novice that they don't even really remember what it is to not know what they know in the first place. Yeah.
And so, and to try to figure out live as you go. I mean, the military provides a great example. They were having such high failure rates with some of their remote operated vehicles until they switched the controller from this custom DOD-built controller to an Xbox controller, a literal Xbox controller, because they found it has all the same— it has, it can do everything it needs to do. And suddenly it was second nature to all these young kids coming through and the success of their UAV training program goes through the roof because they just, they made it a tool that made sense to these novice users rather than something designed by someone who had never really used a, hadn't used a remote control since the era of like remote control toy airplanes and things like that. Like it was one of these crazy old things with knobs and sticks and stuff like that. So yeah, you have to kind of go where the learners are in that sense.
This has been, I don't know, I'm, I'm very curious to stay in touch with you as you, because I actually, there's, there's, I don't, I should be clear. I don't think apprenticeship and that kind of like idea that you learn progressively is completely gone. I just feel like it's not enough. I want to square that with the idea you said in the beginning of like, well, the standard of care is going to raise. So not only is that process not good enough because the tools have changed, but then also the expectations have changed. So we need to get somebody higher up the mountain than we used to get them faster. So did that apprenticeship model ever work at all really well is another kind of question that one could ask. Yeah, but I think it feels like people— and I'm inferring from you that the organizations you're spending time with know that it's broken generally. Do you think that's— is that fair assessment?
There's at least some awareness, right? There's concern that it isn't working as well anymore.
Can they see far enough ahead that it's going to get worse before it gets better in some ways, given the AI piece and the junior people not being there, like all that kind of stuff?
Yeah. What I will say is there are so many— this is speaking more kind of my observations of healthcare is there are more other changes happening that I don't know that they're fully grasping the, the consequences of some of these changes around AI. I think we're seeing just broader challenges to healthcare from a funding standpoint, from a demographic standpoint, from a staffing standpoint that are certainly related with all of these things. I think if I had to capture the general sentiment that I see looking at decisions from senior leaders, hospital CEOs, folks like that, it's a hope that AI might help solve some of these other problems and maybe a willingness to overlook the problems that AI will cause because of the problems that they will solve, right? That we can't recruit primary care doctors and ask them to see 40 patients a day if they need to do all their notes. So They stop going into primary care. Okay, well, we need to get them into primary care. What if we paid for scribes to do all their notes? Well, that's really expensive. What if AI could do it at a fraction of the cost? Great, let's do it. Consequences? Yep, we'll sort those out, but it's less bad than what I was facing.
Can a human handle 40 patient visits in a day from an emotional focus, attention? Like if you take out all the busywork, which is like a, from a sports perspective, maybe that's a little bit of recovery time too, as you're kind of going through your notes. Like if you strip all that away and you're purely just on all the time, is that a realistic ask?
It's realistic in the sense that I know that it happens, right? That fair. There are people who have seen 40 people a day, you know, on a sustained basis. I haven't seen that personally, but a couple busy days a week.
Yeah.
You know, you can certainly see, I mean, 10-minute visit slots. It is doable. Again, back to the sort of burnout and wellbeing question, right? If AI gets cast as a tool that makes you faster, better, smarter, you, you should be able to get the diagnosis in 30 seconds cuz all you're really doing is just typing their symptoms into the machine and it's telling you the diagnosis. You could probably see 50 a day then. Yeah. Right. You know, if you do that.
So, but don't make any mistakes cuz you had AI tell you. Yeah.
Cuz you had AI. Right. Exactly. So it's, again, that's what fascinates me. And maybe it's a morbid fascination that, you know, of how bad could it get?
Right.
But I think we will have to wrestle with that. But what I'm seeing right now is, and again, this is my own two cents, but my interpretation would be awareness that it isn't perfect, but in the landscape of the broader challenges to the healthcare industry right now around costs, around staffing, around potential payment changes, that there's a hope that these are magic bullets that will come in and be able to ameliorate some of those problems.
I've got two final questions for you. One, just kind of thinking back, let's just say even go back 5 years from now. And as you're kind of sitting where you sit at the end of 2025, what's something you've changed your mind on with regards to— could be vicarious learning or the way that knowledge gets moved around, how organizations learn. What was a pretty strong conviction you had that you feel differently about now?
I would say if you had asked me pre-pandemic, right, pre-Zoom migration, can tacit knowledge survive in a non-in-person work environment? I would've been extraordinarily pessimistic. I would say I still think it can't all survive.
And by tacit knowledge, do you mean the transfer of it or you mean the, what, what, like what angle of it?
Yeah, sorry. The, the sort of sharing and dissemination of tacit knowledge. This is how we do things around here. Here's how you navigate the, the kind of thing. I, I loved your example. Here's how you talk to a busy surgeon when you're doing a, a client pitch for them. The stuff that you just can't quite write down. I have been pleasantly surprised to see how much of that has persisted and adapted to the Zoom environment. And I think some of it is changing attitudes towards how we consume information and gather insight from other people. By the same token, if you had told me that there were 3-hour-long podcast episodes on YouTube that would have millions of views, I would've told you, you were high and that people's attention spans are dwindling and blah, blah, blah. So I think there was some seismic shift in our willingness to do that. And it, again, it's this double positive whammy in the sense of knowledge management that a 3-hour casual conversation podcast is exactly what we would've done in the office, right? The kind of shooting the breeze, this whole conversation, we would've had this just in the office as a—
or over a long lunch or whatever it is. Yeah.
Yeah.
Yeah.
You know, the waiter would be very annoyed with us by now for wanting to turn over the table, but we'd have this intellectual conversation. But the beauty now is because it's recorded, it is now scalable and searchable and AIable for lack of a better catch-all phrase in ways that then I think propels us that one step closer to a world where tacit knowledge doesn't have to be disseminated in person in interaction. I think there's still limits to it. I think you still need discussion and back and forth, but I'd say that would've been the thing that I would not have predicted migrating as well.
What do you think was the fundamental assumption that you held that prevented you from being able to see the way that it's transpired?
I think some of it was, I don't know, I was gonna say a misunderstanding of what people would be willing to do online, right? And I don't actually know that it was a misunderstanding. I think the world just changed so much that we didn't have a choice. I don't think anybody preemptively would have chosen hybrid work or remotely remote work, right? If you gave people the choice, I think many more than do now would still say, no, I prefer to work in the office. It's essential for connecting and blah, blah, blah. All the things that we're hearing now with companies who have expensive leases trying to drag people back into the office. I think all of us would've been saying those same things before, but there was just such a seismic shift. I mean, it's a, this will be a case study for 10,000 dissertations, but this was the kind of disruptive event that just fully resets expectations. And sorry, by this, I mean the pandemic. Sure. The pandemic was a, a resetter of expectations in ways that unlocked stuff that I, I just thought, yeah, would have never happened. I would never imagine a world where I would prefer a Zoom meeting over an in-person meeting. And I live that world every single day now because on On Zoom, I can mute myself before I say something bad. And in person, you know, I, I can't, right? Like there, it's weird. I would have not predicted that in a million years. I used to actually teach a class on teaming prior to the pandemic. And we did a fun little day on virtual teaming.
Right.
And as a novelty, I made them do it remotely on Adobe Connect.
Oh my gosh.
Right. This is how I'm dating myself now, but right. Everybody had to log onto Adobe Connect. And they hated it. It was like, it was a teaching point to show how different it was and how much you needed to change your approach as a leader if you were gonna work in a virtual environment. And I just laugh, that just feels so quaint and comical now because if you're teaching a teaming course now, the default is a virtual team.
You're gonna have to teach people how to be around each other in person, right?
Right, exactly. That would be the novelty day. It's genuinely true. We teach online courses now. Now that have these in-person residencies where it's like, hey, we're gonna do this weird thing for 3 days. You're gonna come and be around each other.
Oh my God.
Instead of fully online.
Yeah. How do I go off camera? I don't understand. Like, right.
It, it, it's, it's fascinating and not something I would have ever predicted. I mean, I, you can go back, my writing from that era was all, we need to be in the office, you know? Yeah. Sometimes you might have to do this remotely, but really the only way to do it well is get people together.
Yeah.
And now my writing has changed to, okay, well, how do we get it done given that we won't be all together? Right.
Yeah. I think that that parallels the big shift that we've seen. We, we're building a learning management system into our technology and the big seismic shift we've seen is these in-person 60-minute things is no longer going to be the default for most of our clients. It's like they're going to be on-demand native first. I probably flipped classroom stuff. Go watch some of this, then we'll have a discussion. Just because no one wants to— I shouldn't say nobody. Many people do not want to sit through a 60-minute lecture, let alone a recorded 60-minute lecture. Yeah, it's just really shifting. And then the hybrid thing was interesting. I've seen a bunch of our clients move towards like, even if they're in the office, they do Zoom or Teams, right? Because you can't have second-class citizens, like a different experience for people in the room than the people You know, offline.
Yeah. Yeah. And people are right when they ask the question of like, so then what's the point, right? To come into the office to be on Zoom. Yeah. What are we doing? What we've found ironically is even having meetings for our center, if we had an hour-long meeting and we did all the right things, we provided lunch, we provided all that, 75% defaulted to the Zoom option. Interestingly, when we made it a 4-hour meeting. We called it a retreat to make sure that we didn't say the evil phrase of 4-hour meeting, right? What was funny is more people opted for in-person than, than on Zoom. Now I have no solid theory or, or even evidence beyond that kind of N of 1, but it just got me thinking, will it totally change how we start to structure things? Because I get it, right? The logic, like, I'm gonna do a 4-hour retreat on Zoom, that is miserable. I'd rather be here, be able to pop out, you know, or at least have a sidebar conversation with somebody. Like, if I'm going to be doing it for 4 hours, I want to be there, you know, that'll be more fun. We'll interact or whatever. But yeah, again, not what I would have predicted 5 years ago.
The CEO of one of our clients, I was talking to her actually for an episode of the podcast that's going to come out after yours. Hey, fun fact, these things aren't always recorded in order. And she said she says that she is going to do on-demand learning for things like here's how you use the CRM or here's how you do XYZ technology. It's like that doesn't need to be in person. It can be short. Yeah.
And by on-demand, you mean online?
I mean online, prerecorded, like all the things.
Gotcha. Yeah.
But if it's something like we're changing the way we do performance reviews, that'll be in-person. It'll be multiple hours, kind of along the lines of your thing and like This is something we all need to experience as people, cuz it's a fundamental culture change moment that we're going through versus here's some technical information we're trying to give you.
Yeah. Yeah. I mean, it's an extension of this meeting could have been an email of now we have this whole range of you have to really decide what kind of thing is this, right? Is this an information only sort of thing? Is this discussion? Cuz I think for us, part of the move to a for our retreat that I think did draw people in is it was obvious, more obvious to them. There's no way this is just announcements for 4 hours. 4 hours, right?
Right.
If they blocked 4 hours for it, there must be some kind of input or breakout group or something, right?
Right.
Versus when it's a 1-hour meeting over lunch, we fully intended it to be a discussion, but they would always turn into just kind of one-way things because most of the people were online and it was just sort of like, okay, any questions from anybody? Nope, nobody's camera's even on. Okay, we're just gonna go along with it. And so, yeah, really thinking about the mode of meeting and the length of meeting now as a, a communicator of the impact or the intent in a way that maybe wasn't true before where a meeting was a meeting was a meeting.
So it's, it's organizational redesign really is what I think maybe is the thread running through what we talked about as well.
Yeah.
Whether it's expertise or it's these meetings or learning or all of the things.
Yeah.
Awesome. Well, Chris, thank you so much for sharing. I, I think my job was to hopefully help people vicariously learn through you in this new medium that people will sit around and listen to and be able to take asynchronously. So, you know, this would've been a webinar 3 or 4 years ago, and now we're doing it this way. And so I appreciate you making the time and for the great conversation.
I appreciate the invitation. And if anybody learns anything from it, that'll be a, a good and surprising day.
That's awesome. Well, we'll stay in touch and keep, keep clear on your work. So, all right, thanks everybody. Hey, thanks for listening. If you like what you heard, you can subscribe for new episodes on YouTube, Spotify, Apple Podcasts, really anywhere you get your podcasts. And if you've got comments, feedback, or ideas for future shows or guests, you can email us at smarter@knowledge-architecture.com. Thanks for listening. And we'll catch you next time.
[FOREIGN] Cuz I don't need to drive a car, I get too money out of Classic car, anywhere you want to cross the bay, just make sure you don't stay too late. From Coma or Dublin to Millbrae, Orinda, San Bruno, or Ashby to Concord, our Daily City, my favorite transportation, no matter what you Station, wherever I need to be, I take the VRT. Where does the bus, where does the bus, where does the bus go? Wherever I need to be, I take the VRT bus. Where does the bus We take the fight!
More from Christopher
Human in the Loop? Testing Our Assumptions About Where People Add Value in Learning | Todd Henderson
Redesigning Work for the AI Era: How Greenprint Turns Implicit Knowledge into Shared Intelligence
Designing Learning Like a Product: Branding, Culture, and the Unseen Work of Building an LMS
Scaling Design Excellence: How Diamond Schmitt Amplifies Knowledge Through Learning and AI
From Tame to Wicked Problems: Rethinking Learning and Development in a VUCA World | Ellen Bensky
Learning at the Speed of Change: Adapting L&D to Your New AEC Workforce | Kristina Williams
Designing Modern Learning Organizations: Rethinking AEC Learning Development in the Age of AI
The Link Between Learning, Motivation, and Career Fulfillment for AEC Professionals
Why Communication Skills Are Foundational to Teaching, Leading, and Adapting in AEC F
A Framework for Collaboration and Quality: SMMA’s ISO 9001 Journey | Ryan Farias & Jessica Smith