LH134 AI's Hidden Tax with Josh Cavalier

You can create tons of outputs, but still somebody has to go in and put a stamp on it and approve it. And if we continue to crank out so much more content, well, who's responsible for that at the end of the day? I certainly know that agents are not going to be put on a pip. That's for sure.

Welcome to the Learning Hack, a podcast about the people and technologies that are creating the future of learning. I'm John Helmer.

Now guess what?

Learning is cool. Learning is cool. Learning is cool. Learning is cool. I'm learning! Learning is fun. Knowledge

is power. Knowledge. Education. Learning.

What happens to the people who make learning content when a machine can make it faster and cheaper? Will instructional designers become performance designers? What's the hidden time tax that AI imposes? And how do you smoke a potato salad? Over the last couple of seasons, my guests have answered a lot of questions for me about AI and learning, but none of them has raised as many questions as this guest.

Hack Facts.

Josh Cavalier runs joshcavalier.ai, which trains learning teams to work with artificial intelligence. He's also a co-founder of Talent Rewire, and last year had a book published by ATD Applying Learning in Learning and Development: From Platforms to Performance. Important word there. He trained as a medical illustrator at the Rochester Institute of Technology, and for most of the 30-odd years since, he's been teaching L&D teams how to use each new technology tool as it came along— authoring software, video, learning platforms, and the rest. In December 2022, a few days after ChatGPT launched, the new tool became Josh's book proposes a new role for our profession, the human-machine performance analyst, but is frank about where most learning teams are right now.

So many of the practitioners that, that I facilitate for week over week are still just creating content. They're still order takers. And that makes me extremely nervous because we have a limited amount of time here to elevate The type of work that we do.

We talked about what happens to the instructional designer now. We talked about what he calls the supervision tax, the cost of checking everything the machine produces. I asked whether AI is going to create a lot of redundancies, and he didn't duck that one. And we asked why agents, for all the demos, are still rare behind the corporate firewall. Nevertheless, he's upbeat about the future that he can see is happening already. Right here and now.

I believe that we are in a golden age of creating learning, training, and performance experiences to support people where they're at on the journey.

Here's Josh Cavalier. So Josh, welcome to The Learning Hack. Great to have you along.

It's great to be here.

So you published a book at the end of last year called Applying AI in Learning and Development: From Platforms to Performance, and it introduced us to a new role for L&D, the human-machine performance analyst, the HMPA. In plain terms, what does that person do that a training manager or an instructional designer doesn't?

You know, John, since writing the book, I've had some clarity on this role. I knew when AI really started taking off that there was going to be a seismic shift in the way that instructional designers are going to show up and support the business. And my thought was that the HMPA was going to have additional performance consulting capabilities and work with the business to understand exactly the relationship of human and AI. And how does that show up in regards to upskilling, reskilling, or just maintaining work? But since then, there has definitely been a bifurcation, and I have more clarity on that role, especially with my work with Talent Rewire. And so now, well, first let me go ahead and just address the instructional designer and where that's headed, because so many of your listeners are attuned to that. And then I'll get to the HMPA.

Right.

So, I believe that the instructional designer is going to, there's gonna be a little bit more nuance. There's gonna be a shift to becoming a performance designer or a performance orchestrator.

Okay.

And, and that, that change of words is intentional because of working higher up with the business to gather data about current state problems. It, is it truly a training or performance issue? And then orchestrating experiences and then measuring the impact of those experiences. At scale. All right. And that's where, you know, I spent my whole career either creating content or teaching people or facilitating how to create that content. And especially in the last 6 months, there's so much more orchestration that's going to happen. And I see it month over month. There are new capabilities that these models have from a generative AI standpoint that are eliminating things like a timeline, right? I spent my whole career in timelines.

Yeah.

And now I'm just talking to a model doing video edits. Yeah.

It's pretty elegant. Mm-hmm.

And, and so that is the direction I see the instructional designer. Now, back to the human-machine performance analyst.

Yeah.

That is truly becoming a business analyst role, but with performance consulting skills. Okay. And the reason for that is that with each iteration of this technology, we get a change in the surface of work. we get a change in the relationship of humans working with AI. What remains uniquely human, what is collaborative, and then what can be, you know, the low-risk type tasks that can be given completely to AI. And that division of labor has a cost associated with it. Mm-hmm. And the human-machine performance analyst gathers that data, creates a plan or report for leaders who then ultimately make a decision on that division of labor, whether to invest more in AI, how do we invest in our humans to upskill and reskill them, especially when we have freed capacity?

Really interesting what you're saying there. And just so many questions about how you're joining the dots between where we are now and how we get to that world. But I'd like to park that for now and come back to it. For now, could we just look at this change through a slightly wider lens? In this podcast season, I'm talking to quite a lot of the people who are helping learning professionals get to grips with AI as part of their day-to-day work, as you are. And new job roles keep coming up.

Yeah.

For instance, Ross Stevenson from here in the UK talks about the AI enablement architect. And it does make me wonder about the structural changes we might see in our patterns of work with technology, which might be more than just a change of job title, because it's not just within the L&D department, I don't think. From the vendor side, firms across tech are sending in FDEs, I hear now, forward deployed engineers.

Right.

Whose job, and that's coming from the vendor end, as far as I can see, whose job is to make the stuff work in enterprises and map business use cases. I came across this pre-AI with someone called a a business success manager or something like that. Yeah. I've worked for startup software companies and they would have someone who would do this job of going in, sitting with the client, working out what they're trying to do and help them do it. Maybe there's an overlap there with the HMPA, but it does make me wonder if the current structure of practice areas, HR, L&D, IT, will survive the change as this really gets going. Does the sort of change you're talking about create a kind of pressure on the way that organizations organize themselves around the people area?

It does completely because with each iteration of a new model, the jagged edge of AI disrupts work. And the only way that that can be surfaced is by a subject matter expert or somebody with deep domain knowledge, in addition to AI fluency, being able to go ahead and surface that reality. And you see, that is the, that is the issue at hand currently, is that many, many professionals are not fluent in AI so that when a new capability comes within the firewall, they can immediately Understand the impact on process, on workflows, or at the organizational level as a whole. And when that happens, you can have massive shifts in the allocation of work between humans and AI. And we need professionals, you know, whether you wanna call it a for— forward deployed engineer or human machine performance analyst or business process management or whatever you wanna call it.

Mm-hmm.

Somebody has to make sense to leaders about that reality and the shape of it. And because if there's a big shift and, um, you know, a leader is like, let's go ahead and take those tier 1 call center people and make it all AI. Well, that is, that's quite a leap. And within that, there's risk and there's reward. And at what point do we, you know, does it, Flip to where it makes sense. And what do we do with that freed human capacity? And for organizations that are proactive and can transform that freed human capacity into net new work, into innovation, into top-line growth, those are gonna be the winners moving forward. I, I have a, I have a problem with organizations that are costing out humans. Mm-hmm. Because those are the ones who can make sense of things that are external to these models that from a business standpoint allows the velocity to happen. And, and these are the individuals that are going to take on like the V-shape of work, meaning that they're deep in one area and as they gain fluency with AI, they begin to acquire new skills naturally of working with the model. And I think about learning and development professionals who may not be as fluent around marketing or building awareness around their training programs, and they tap into AI and all of a sudden they can write an email sequence. They can go ahead and, you know, build that explainer video or, you know, something that gains attention as far as a big program or learning journey. And they didn't have those skills before, but then they start gaining nuance in that area. And they start going, you know, that wider V-type pattern. And those are gonna be extremely, extremely beneficial associates to have on staff. And I, I think that we're going to see mistakes being made in regards to costing out humans. And we see it already with organizations bringing people back because they went too far too fast. And not realizing exactly how beneficial it is to having a human on board that can make sense of what's happening with AI and how does it impact what we do within the organization to support our customers.

All this does make me wonder if it becomes a practice area, or does it kind of get subsumed into IT? That, in a way, doesn't seem all that likely. I don't know. Veering off left-field science fiction, I've really enjoyed Silo. the science fiction series on Apple TV. And there the head of IT is effectively in charge of the government. IT becomes such a big role— That's right. That it kind of swallows everything. But I mean, there's also an alternative scenario perhaps where it becomes the use of AI for this kind of business process re-engineering becomes a practice area just like finance or marketing with its own kind of clarity and set of skills because not everybody necessarily has the kind of foundational knowledge and aptitude and just the interest in this stuff. You can see some people really take to it, really like it, and others are like, no, I'm not going to go there. I'm not going to.

Yeah. And John, it's been that way with every technology shift. And I've been through desktop computer, internet, mobile phone, There has always been, and I think, you know, if you want to call it growth mindset or adaptability, whatever you want to call it, I think inherently some people, you know, they thrive off of it, and others, they are concerned because their world is completely rocked because everything has shifted, and they don't know quite how to navigate that because they're extremely comfortable where they're at. And we are definitely in an environment where it is not comfortable. Because again, the, the doubling of capability every 3 to 4 months is really forcing everyone to maintain some type of momentum in regards to their own upskilling and reskilling so that they feel like they have agency throughout this whole entire process as opposed to just blowing in the wind. Which unfortunately, a lot of people feel that way, like they don't have control. over their career for maybe the first time in their career. And that is a really horrible position to be in. So they're just trying to find some grounding and some direction on which way to go, especially as we move into 2027.

Yeah. I think I'm less talking about motivation than about proclivities. I mean, you'll see people in organization who are people people and people who are mathsy, and it's kind of, they're both equally motivated, but they have a very different skillset and they go to different areas of the organization for that reason. I mean, some people just like to make stuff. Some people like to deal with people. And some people like the strategic side. Do you know what I mean? Yes, completely.

There

are multiple intelligences and people— that's why I wonder about it becoming a practice area. Or is it that this is just something that everybody has to get their heads around and do to a kind of equal level in as much as they can.

Well, I do think it's going to be a skill that everyone's going to be required to have. I mean, when you have an interface into collaborating with this technology that accelerates the way that you do work, and, you know, that big skill that everyone, in addition to taking on all this AI fluency, is also, it's something I call the supervision tax, which is you can go in and you can be, you know, on the surface efficient, but are you being effective? And that's where the supervision tax comes in because you can create tons of outputs, but still somebody has to go in and put a stamp on it and approve it.

Yeah.

And if we continue to crank out so much more content, Well, who's responsible for that at the end of the day? I certainly know that agents are not going to be put on a PIP, that's for sure. Like, some human is going to have to go ahead and take, you know, um, the bumps if some output does not align to the work output and the quality floor that needs to be there. Um, so we still have— Yeah. Especially on the, on the HR side of the house, we still have a lot of work to do On shoring up leaders and how leaders support their individual contributors in the age of AI. And what does that look like? How do they create safe spaces that experimentation, that work reconfiguration can be done in real time, you know, in a, as a natural flow of work week over week or month over month throughout the year? And I, I go to leaders, I'm like, listen, you're gonna have to, you're gonna have to think about the support level of your teams and making sure that when you acquire new capabilities of AI, that those spaces are created to where you can get together as a team. You can quickly, within an hour or 2, attack a business problem, this new capability, and then disseminate out, you know, to other teams the success or failure. And what does that look like and how does it change the way that we do work? And we've never really operated like that before. I know when I, like, I owned a training business for years teaching Adobe software and that software wouldn't update, like, it would be like every 2 years, you know? So you had, you had some time to like adjust and, and, and, you know, work through that. Now it's every 2 weeks. Yes, it's every week now.

Every week.

And so we've never been in this situation where knowledge workers at scale have before them a complete change in the way that they do work. I know for myself, especially the last week, you know, Anthropic releases Opus 5.5.

I was just going to say, what a big change that's been.

Yes. And on the surface, if you don't have your head in the game, it'd be like, oh, well, isn't that novel? Maybe it can write better emails. But then it solves for video editing. It solves for design. And I'm making explainer videos that would take a day and a half in 45 minutes. And my jaw is on the floor. I've Yeah. I've seen it all. You know, I've experienced for over 30 years and still every single week there is something that just amazes me because I know, you know, on the one hand we have this pressure of people who as professionals, they have done this hard work. They spent 20 or 30 years knowing how to do video editing and they see this before them. They're like, oh my gosh, I, I, all that effort on a timeline, now it's all gone. But then there's the other half where I think about all the individuals that have come through my courses or classes or anything like that, and they were very apprehensive about doing video because there were so many things to understand— the lighting, the camera, goes on and on. It's just like, it's very, very intimidating. And now they're empowered to, to create.

Hmm.

You know, those barriers are removed, and that's what I see. I mean, I'm a very, like, you know, glass half full kind of person, and I see a world where— here's another one, accessibility. I know throughout my whole entire career, accessibility has been secondary. Like, you create the thing, the PDF, the website, the mobile, whatever, and then you make it accessible. And, but now we can leverage this technology to be accessibility first. Like, all the design elements and all the things within your course or, you know, whatever media that you put out there, it could be accessibility first. And my gosh, isn't that wonderful? Yeah. Like, to think that we can go ahead and really support individuals where they're at. And I believe that we are in a golden age of creating learning, training, and performance experiences to support people where they're at on the journey.

If you're looking to connect with the top minds in learning and talent, there's no better place than an iVentiv event. Since 2009, iVentiv has hosted over 300 events for global heads of L&D and talent. I've attended multiple iVentiv events myself and found them to be invaluable. Less about sitting and listening and more about real, impactful conversations. It's an unmatched opportunity to collaborate with leading Chief Learning Officers on the challenges they're tackling today. Join the conversation. Visit iVentiv.com/learninghack. That's iVentiv.com/learninghack. So to ground this in the present moment a bit, how much of all this do you actually see happening on the ground right now in your work? In other words, where are they? You know, what place are they in their development? I think this is something that everybody is interested in because it's all changing so quickly. You describe L&D's journey as creation, curation, generation, automation. Where are most of the teams you work with on that path right now? And maybe you can give us a kind of brief description of what those steps mean.

Yeah. So where I see most practitioners is just in the creation sliding into generation. Most are still order takers and until we can gain skills around performance consulting. And, and the reason why we need that in data and analytics is because it's, it's about acquiring a source of truth from the business that allows you then to, one, identify, is it truly a performance issue? And if it is, in real time, to go in and create experiences to fill those gaps. Right. To fill those knowledge and fill those skill gaps in real time. So you need to understand acquiring data, correct data from the business. And again, is it a performance issue? So many of the practitioners that I facilitate for week over week are still just creating content. They're still order takers. And that makes me extremely nervous because we have a limited amount of time here. To elevate the type of work that we do. And it's going to take some, you know, gaining of skills because the generation, you know, slotting into generation is moving more individuals and not at scale. But I do see practitioners who are AI fluent, who are now leveraging generation at scale sliding into automation, meaning they are moving from turn-based chats to leveraging harnesses, AI harnesses that use models at scale to do deep work.

Mm-hmm.

So as opposed to doing a single prompt to do turn-based work, they're creating a goal for the model. That model creates a plan. And then drives towards goal completion. Now, to orchestrate all that, you really have to understand what's happening in the background.

Yeah.

With context, uh, being acquired, with tools that are being called.

Hmm.

And what is your relationship as a human throughout that whole entire process in telling the model, yes, continue, yes, continue, oh no, hold on, uh, make an adjustment. Okay, now continue. Um, And that is the shift that I see happening, but not at scale because of the cost associated with working like that.

Hmm.

This is more than just your $30 US a month Copilot license. Now we're talking about token burn and what is the budget of the individual when they leverage AI at that level. And so it's very, I see it in small, medium-sized businesses. I see it in organizations that are very pro- proactive when it comes to AI operations, but at scale, we are not there. So I think we're going to see some barriers to accessing generation-to-automation capabilities until the price comes down. Now, that is happening today. We do see increasing capability exponentially. The cost is starting to drop. So I could see as we get into the latter half of 2027, the potential for more individual contributors getting access to Claude Cowork or ChatGPT Work or Gemini Spark, these types of capabilities that are coming online that are token or credit-based charges.

Yeah. And of course there's also, I'm beginning to see people talking about, well, it's all very well people as individuals accessing their own AI and building up the context for it to do stuff for them. But we work in teams. Right. We also have to perhaps have a type of team AI that you will do certain things for as well as the individual ones you'll complete your own work with. But I'd like to go back now and just have a look at your background a little bit. In a way, perhaps we should have started with this, but we like to start in media res. You trained You trained as a medical illustrator. You've got a strong background in visual design and digital media, which I found very interesting reading your bio. So how does a medical illustrator end up in R&D? And what from that training do you still use?

You know, back in 1983, I got my hands on a TRS-80. And one of the first things that I wanted to do was create an animation Using BASIC.

Explain what a TRS-80 is.

Right. Yeah. So it's one of the first desktop computers with, um, you know, we use tape drives, uh, to actually go ahead and read and write and save our programs.

I know what you mean. Some people won't.

Yeah. You may affectionately know it as the Trash 80. Um, so it, you know, I've always been intrigued with multimedia.

Yeah.

At a very young age. And this mix of art and science has been a part of my world for the longest time. And medical illustration was a fantastic practice. But, you know, I did my school at the Rochester Institute of Technology and we got exposed to Mac very early with the first versions of Illustrator and Photoshop and then Director. Um, um, you know, very early on. And something about movement on the screen and controlling it really intrigued me. So even though I was on this path of becoming a medical illustrator, and I've done medical legal and demonstrative evidence for court cases, and I've, I've had my illustrations published in, in magazines, um, I had an opportunity in 1994 to work as an art director for this gentleman, Dick Hanshaw. Mm-hmm. And here in Charlotte, North Carolina. And Dick has written up— he wrote a book with Jim and Dana Robinson, Jack and Patty Phillips on performance consulting.

Uh-huh. Yes.

And I got exposed to performance consulting really early on. Now, as an art director, full transparency, I found extremely boring because of all of these interviews that you have to do with leaders to find source of truth, like what's happening with the business.

Yeah.

But as a, as I matured in the industry, it became apparent that I was very lucky to get exposed to that early on and understand what happens at that level and why is it so important at the, at the front end to identify what's happening with the business. And so when I transitioned from that business, I started my own training company and I was the lead consultant in that business to go into L&D departments and spin them up on tech, which typically meant facilitating authoring-based tools and implementing LMS/LXPs. So I've been in many, many Fortune 1000 organizations around the globe doing that type of work. So that's the, that's the path that I made, that I went from medical illustration into learning and development, and I, I just fell in love with it. And it's about helping people. You know, that's, that's the, that's my motivation. That's what gets me up in the morning is to assist people in becoming better versions of themselves. And that's what, that's the work that we're doing over at Talent Rewire. My other business is that when it comes to the division of human and AI labor, we don't believe in the cost out. We don't believe in getting rid of humans. We believe in double down, um, you know, on the human and upskilling and reskilling so they can accelerate within the organization in the age of AI. And so I, it's, it feels like I'm coming full circle in that, you know, I started with this, you know, very kind of limited view of how to help people out. And it kind of, it started to widen as I had my training business. And then now the work that I'm doing, it goes extremely wide because we are at that inflection point where individuals have to make a choice about their career and how are they going to address and like, what does that personal learning plan for them look like in 2027 so that they can remain relevant and not just remain relevant, but lean into this reality so that they are extremely valuable and they can, they can show up so much more stronger for themselves. individually as a professional and also for the people or in the businesses that they support.

Yeah. And your road to Damascus moment with AI was finding ChatGPT in December 2022, where a lot of people did, I think then during that month, 5 days after it launched, that was. So what did you see at that point that made you rebuild your business around it? Was it I mean, was that instant? You thought, you know, the world must change now.

Yeah. I'm fairly pragmatic because I've been through, you know, when you go through so many iterations of technology, like, is this gonna be another Google Glass? Like, where, where is this headed? Right.

Another Second Life.

So, you know, I always, I always, I, I, I have reservations. Same thing with AR and VR. Like I went through it, you know, back in like the early 2000s with 3D and Director. And I was like, wait a second, there's, there's a huge cost. to this and maintenance and hardware and like only few people are going to use this. So when I got my hands on ChatGPT, you know, I did the typical, write me a limerick about me and my friends drinking beer. You know, I was like, wow, this is novel. This is fun. But then, you know, I'm a workflow junkie. I was like, okay, well, can it do a learning objective? Okay, can kind of, can it write a video script? Oh, well, this is interesting. Can it write a storyboard? Oh my. Huh. Can it write multiple choice questions? It can. Uh, and I was like, well, if this really does take off, then this is going to radically change the way that we do our, our job. And, but, you know, I had to dig in deep for like 2 months and watch tons of YouTube videos about how a transformer works and the calculus and what happens to a word when it makes it, you know, it's created into a token and it's a new number and that number runs through the transformer. Like, I need to understand how this works. And constant prompting for 2 months and I'm like, oh, okay, I think I now have some confidence around this to actually go ahead and post some stuff on LinkedIn about what's happening. And it caught a lot of people's attention. And I think I was like one of the first out of the gate to at least expose how our industry is going to shift and change. And I've been trying to lead the way. I'm not right all the time, but I try to get out there and allow people to think about this reality and give them some insights, at least through my lens. And my hope is that through that, they can make some correct choices about the way that they spend their time in learning about the technology and how they apply it in their careers.

Let's get into the worries that people have, the fears that people have about being replaced. Yeah. You have a line, one analyst can impact thousands, which is a great statement about leverage and about the power of this technology to help us with productivity. But a financial director could read that line as, so we need fewer people. Does that mean we're headed for mass layoffs of people who can't make the transition to HMPAs? Do you think, I mean, some people, they're just not going to get there or they're going to get there very slowly?

It's going to happen. I hate to say it, but there is so much more evidence out there that work is shifting and that we have certain roles that You know, will support the business that once AI hits a certain level of quality and risk is mitigated, if those individuals don't upskill or reskill to maintain their value in the business, that job will go away or that role will shift so dramatically that they're not a good fit for that role anymore. And I think that's where we're going to see more of it. I don't see AI truly Like in one fell swoop, eliminating thousands of jobs. Now, on the surface, what you read in the trades and what you read in the paper, you see that, but it's not AI eliminating jobs because of the, the role has changed so dramatically. It's because organizations are trying to find capital to pay for AI.

Yeah.

And the only way they do it is through operational expenditures and getting rid of 4,000 people in one shot, which, you know, the stockholders love, Uh, but from an op— you know, from a business standpoint, don't know if it's the best move. You know, all that tacit knowledge walking out the door that could be used to actually go ahead and, uh, be used to, you know, figure out AI, uh, within a function or a line of business, I think is essential. And I think that's the reason why I am all for double down on people. Mm-hmm. And to get them to a point where They understand how the technology works and they know how it's going to impact the business, and they can surface that information to leaders. And that's the human-machine performance analyst. That is their role, is to actually go in and acquire that data, to get that information, and to make sense of it in such a way that a chief financial officer, a chief operating officer can go ahead and take a look at the bigger picture. And then if they want to go ahead and make a big move as far as acquiring additional AI capability, paying for tokens as opposed to, you know, a fixed fee and letting a group go in and take it and take off with it. You know, that takes some calculated risk and somebody's got to make sense of that. And so I believe that's where the profession is headed. Some individuals, again, who love that front-end work, that love that performance consulting work, are going to shift off in that direction. And as I mentioned earlier, we're going to have Instructional designers become more performance designers, orchestrating experiences at scale. And my gosh, that's gonna be so important because we are gonna be in a, and we're in it right now, this constant state of upskilling and reskilling. But individuals only have so much throughput when it comes to learning. Yeah. And so, for it to be laser sharp and focused and in the moment of need, that takes a lot of of professionalism to where you can go ahead and not just create those experiences, but, but it's also partnering with other parts of the business so that if you do have an internal system, that those, those experiences are built into the system.

Yeah.

And my, you know, uh, it, it's not gonna take much to go ahead and get a spec from the IT department to then go ahead and Vibe program, and they would take that code and integrate it in with the system so that a video or a walkthrough or anything like that can be integrated directly into the system that could then self-heal with AI. That capability exists today. It's just that those bonds, those partnerships across functions have yet to be established or are just now being established.

Okay. It's great to hear you talk so much about the importance of humans in the process and human values and so on. But Talent Rewire, with Talent Rewire, you're breaking jobs down task by task. You're analyzing workflow to see what AI will automate. In a sense, you are part of that process. And I do appreciate the value of that work because I feel like I've been doing that all year. I've been looking at which bits of my workflow can I kind of hand off. Right. And the progress is dramatic. It only seems to me that It's in this quarter that AI has really become useful and I've really begun to save days of time with it. And so I appreciate the importance of what you're doing. People could say that you are complicit in the job of letting machines take stuff up that maybe humans used to do. But if you ran that analysis on an L&D department, I'm sure you must do that. What would it show? Which parts of the L&D work are most exposed? Which are the safest? I mean, I suppose, just to answer a bit of that question, the obvious answer, content creation, kind of out the door.

Yeah, content creation is exposed now. Again, with each iteration of these models, the jagged edge of work changes, and content creation is front and center. And there's some recent vendor announcements that really shine a light on that. But the reality is, is that with every workflow, with every process within L&D, there are tasks that are extremely complex. And if that task fails, it could be detrimental to a whole entire effort. Yeah. And I think about all the front-end work that we do when it comes to performance consulting, learning needs analysis, action mapping, creating a learning journey, um, evaluation, understanding how that's done to, to validate, you know, did we actually change behavior? Did we go ahead and change or impact that KPI? And to say that AI is actually going to go ahead and do all of that, I don't believe it for a moment. There is too much risk. Now, within each of those items I mentioned, we're going to be partnering with a model to get certain parts of that work done. Mm-hmm. But then there's also just the human-to-human work. That's, that's going to a leader, going to a subject matter expert, and having a conversation. Or hitting the record button and getting the transcript and having AI go ahead and, you know, take 5 SME interviews and then surface patterns that will indicate a performance issue, right? That's that collaboration. Mm-hmm. So, you know, at the end, again, at the end of the day, who's gonna take the blame for a work output not being of quality? You can't blame a machine. It, it's gonna come back to a human. And so I, I, I think we're going to be in this period of mistakes being made to where people go too far, too fast with a machine, come back to the human. But, you know, there's a balance to be struck within every task, and we're gonna find that rhythm. We're, we're going to find the shape of that over time, and it's just gonna be integrated into work so that when we do have a new capability that comes on board, it's gonna be extremely evident. How work is gonna change, and then what does that mean for me in the way that I'm, you know, work in my role? And to go ahead and have a conversation with a leader about that, you know, now I have freed capacity, what are we gonna do with that freed capacity? And for organizations that are proactive in setting up systems and setting up, you know, work to where that freed capacity can be used for mentoring, Mm-hmm. And I especially see new hires. I, I, I just had a conversation this weekend with a gentleman who's a lawyer, and he is like, I'll never have to have a paralegal before. Now, he has a small firm, and I can see that, but I, I have concerns about organizations that feel they're going to cost out or not have junior associates. Like, how do you have talent succession when your experts are knocking things out with AI? Or, oh, and this is really horrible, that you have new hires that are getting work done, but they don't understand the work. Like, AI is, is being used to get work outputs complete, but they don't understand the how and the why of the work that they're doing. And so to have coaches and have mentors guide them through and explain, this is the reason why we do this, even though AI is doing all the work for you. To understand the intricacies, the nuances, because I would, I would, you know, take pause in situations to where learning or performance experiences were just automated.

Mm-hmm. Right?

And, and then at that moment, you'll look like a hero because you're, you're meeting the needs of the business. But if you don't have the instructional designer with those learning sciences, with those adult learning principles, Those are the ones that are going to make sense of these new capabilities coming on board and how experiences change. And I'll give you an example. We just did a Copilot enablement for an organization, 300 people. And of course, you got to baseline their knowledge. Where are they at? Well, I can go ahead and interview them, and that's going to take some time. But we use this latest interview technology. It uses audio-based agents. Sounds amazing. And it asked both behavioral and cognitive questions. And we interviewed hundreds of people in 2 days, 8 to 10 minutes each. You were never able to do that before, like economically at scale. And to have all of this qualitative data, uh, you know, that we assign rubrics to, to baseline numbers, you know, to actually go ahead and give us an idea of where people were at and then run through those same interviews at the end of the experience to see that lift. But even, even within that capability, I had to know as a professional about behavioral and cognitive questioning.

Hmm.

Right? Like, there's a nuance there that you have to understand. And that's why we're so important, is that all these nuances that we understand about our profession needs to be applied to new capability. Because if you, if you current state it and you just like, okay, I have AI that's going to go ahead and make content. Yeah, it gives you value in the moment, but that value is going to decrease over time because you don't have the right professional in place to take advantage of new capabilities, or you're going to rely upon a vendor that has a 6-month lag in applying that new capability within their tool or platform.

Hmm. Interesting. While we're talking about examples, You're presenting at DevLearn, or probably will have done so by the time this comes out, on Rob Roy Industries, a family business founded in 1905 that makes electrical conduit. So not a Silicon Valley startup, but tell us a bit about that story that you tell and the part that AI plays in it.

So Rob Roy is fascinating in that they are this blue-collar, manufacturing, electrical manufacturing company. But as part of their DNA, and certain organizations have this, is that when a new technology shows up, it's just a Monday for them. You know, and we, and we see organizations that are like that. You know, another one that reminds me of, of Rob Roy and their attitude towards technology is John Deere. You know, you think they're just like a tractor company, but they have cameras that will monitor the exact amount of herbicide that's sprayed onto the ground, saving, you know, a farmer millions of dollars a year in less waste and things like that. So, you know, for Rob Roy, they saw AI and decided internally, well, how, how is this gonna shape our business? How is this gonna reconstitute the way that we do work? And that's from leadership, that's from top down. And so immediately they went ahead and started their AI operations initiative. And figuring out exactly what does that look like. Now, my part in that process was enablement, was, you know, taking the access that they gave to all associates and spinning up champions and making sure that they could build a bridge to their expertise and how AI fluency changes the way that they do work. Right. And so then those champions would would disseminate down into a function or a line of business these new opportunities of doing work. And one thing I absolutely love about Rob Roy, and this is something we're going to bring up at DevLearn, is how they were very intentional about giving within throughout the whole entire business access to data, specifically sales data or data that people can Uh, you know, lean into a workflow with. And that was phenomenal. Not only that, but giving a platform where you could pick from 7 or 8 AI models, not just one, not just, oh, you're on ChatGPT Enterprise or Copilot or anything like that. No, they can go ahead and pick any and all models. And those models are constantly being updated. Mm-hmm. They could create agents. They can go in and, you know, create automations. And that's all accessible there within their own internal platform. So it was really fascinating, this, this kind of juxtaposition of this manufacturing company that at first glance you'd be like, oh, they're not technology savvy, but behind the scenes they're the opposite. They're actually, you know, allowing business operations to flourish because of their support of people in this transition to AI, which is— that's an— that's incredibly important. Yeah. is being able to go ahead and make the investment in the people as the technology's being implemented. And I see it lopsided to more on the tech side than the people side.

You mentioned agents there. What sort of traction is agentic AI getting with L&D? It's getting a bit of a bad press at the moment, at the time of recording. You've got a chapter on autonomous agents, and you wrote From Prompts to Agents over a year ago. But what I agents actually doing in L&T teams today? How much of it is there about?

Yeah, it's not showing up as much as you would think, right? We do have, if you go up to professionals like LinkedIn and you do see individuals up there on the bleeding edge kind of showing everybody like what's possible. But I know that when I go and I speak to learning professionals at conferences or, you know, within when I'm facilitating, Uh, we don't get evidence of agentic, let alone automation, uh, work being done. Now that is changing. Again, the access to AI harnesses that leverage models as agents that exist today, just not at scale behind a corporate firewall. So I do see when the cost goes down and access increases, to Claude CoWork, Copilot CoWork, ChatGPT Work, Gemini Spark. All those environments allow an individual to work with a model as an agent. And it's all because of memory, like, and tool calling. So those 2 things right there allow deep work to happen to where the model can plan towards goal completion. And you're orchestrating that agent to where It spins up subagents to do work at scale. And, you know, each and every week I'm shocked at how elegant that's becoming to where when you have a huge ask of the model, like, I need you to do deep research or I need you to go in and change these values at scale, and it fires up subagents to run in parallel to do the work. And it is impressive to see the panel come up and all the subagents are working in parallel. But the unfortunate thing is that we, we just don't have access to that at scale.

Hmm.

Now, people that are working independently and they have their own businesses, they're doing that today. Small and medium-sized businesses that have a little bit more flexibility in regards to access to AI, that work is being done. And for those companies that are extremely proactive with access to AI and putting the budget in place to have variable-based token burn, that is also happening. But for a majority of L&D professionals, they do not have access to that capability behind the firewall.

Yeah. I think you've given us a great idea overall during this interview of exactly where people are, which I think is important knowledge at the moment. People really want to know it. So that's great. I'm going to do that very annoying thing now that people do at conferences of saying, this isn't a question, it's a reflection. But you've mentioned performance consulting and how energy should move towards a more performance-orientated sort of approach. And I haven't kind of taken the bait and asked follow-up questions about that for the simple reason that every single one of the people I've interviewed in this season so far has brought it up spontaneously. In fact, want to base their answers around it in a lot of case. And I think that's very interesting. And I wonder if it's the case that AI is one of those things where technology improves and suddenly you can do all the stuff that you wanted to do all along, but just couldn't because it was too expensive or too time-consuming. You see a lot of little examples of that, people doing extra stuff rather than just taking tasks away from people. So I think that's really interesting. But we're short of time, So move on to the final question. You smoke barbecue, I read.

Yes.

And to be honest, you're kind of aware that Brits have a different sort of sense of vocabulary. That sentence falls very strangely on the ears of a Brit, and it's even got some rather kind of spicy connotations in certain areas of slang. But we have barbecues over here, obviously, and personally, I love a lot of smoked food, smoked mackerel, Smoked bacon, always smoked salmon, of course, a staple if you've ever been to a corporate function in England, but you smoke potato salad. I mean, to me, that just sounds fucking weird. Can you tell us about how that works?

Oh my goodness. Well, first of all, I can probably talk about smoking food more than AI because it's my hobby.

You've got 5 minutes.

Yeah. And so the way that that works is that you have to go ahead and you boil the potatoes first, typically cube them up, boil them, and then you will put them in the smoker. And it's only a kiss. It's only just a little bit of smoke. And you'll, you'll go in and you'll add all of the accoutrements. And I like, I like them loaded. You know, I like them with, you know, bacon and cheese. And of course the dressing that goes along with it. But it's just a hint, you know, it's just, it's a hint of smoke. It's nothing even overpowering. Just a little bit.

When you say smoker, that's like a kind of desktop thing, is it? Or what, you know, cause it makes me think of these little wooden huts where people smoke.

Yeah.

So herrings and stuff like that in—

Smoker actually use is what's called a Weber kettle. You know, it's this spherical metal container that holds charcoal and wood.

Right.

And during COVID my wife acquired one of these for free on the side of the road. And because of all the additional time I had, I'm— I wonder if I could smoke food with this. And I figured it out through YouTube. And then I went down the rabbit hole and I actually will— I smoke my— I'm excited about winter because that's when I actually smoke salmon. So I'll cold smoke salmon. And eggs and cheese, uh, during the winter. And, uh, you know, so it's a year-round, uh, hobby for me. And it, and, you know, it, it comes full circle, John, because this is art and science, right? You know, my mind, I just, I, I gravitate towards anything with art and science and that's, um, smoking food scratches that itch.

Okay. So if it moves, train it. If it doesn't, smoke it. Now this has been completely irrelevant, this segment to L&D, but I've learned something. So thank you for that.

Oh, you're welcome.

And thank you for taking the time to do this interview. I've really enjoyed it and I'm sure people will find a lot in there. There's an awful lot of clippable stuff in there, Josh. So thanks a lot for doing the interview.

John, thanks for all your work. It's been a real pleasure talking with you.

Thanks to our guest and to our sponsors. We'll be back in your feed in 2 weeks' time. Meanwhile, stay classy, learning people.

Now I finally get it.

Let me ask you, have you learned anything from this podcast? And are you wondering maybe whether a podcast could help your people learn stuff? Well, when I'm not making my own shows, I walk organisations through making podcasts for learning of their own in-house. And now I've written a paper setting out the evidence for where it works and where it doesn't, together with 10 questions that will help you decide. It's free to read or download, no form, no follow-up, at johnhelmerconsulting.com. The Learning Hack Podcast is supported by Learning News, the learning sector's newswire. Rob and his team are good friends of the podcast, and we really value the help and advice we've had from them, and they do a great job. For the very latest news from around the learning sector, for interviews with learning leaders, the latest from learning sector events. For more on LENDERS and features on workplace learning, go to learningnews.com.