How Science Suggests You Change Your Organization - with Prosci’s Tim Creasey and Paul Gonzalez

0:00You know, change management is something that comes up again and again and again in these conversations. And today we are so excited to have Tim Creasy and Paul Gonzalez from ProSci. These folks have been studying change management in organizations for the last 30 years. And today we're going to talk about how the timeless techniques of change management are being applied at the cutting edge of AI adoption today. Hi, Tim Creasy, Chief Innovation Officer of ProSci. Excited to talk about the conditions of successful AI adoption and here with Paul Gonzalez. Thanks, Tim. I'm Paul Gonzalez, VP of product at ProSci.

0:37Been working in technology and product for about 10 years now. And I'm excited to share what we're seeing with organizations and teams that are using some of the key change management tools and frameworks to drive successful AI transformation. So maybe we just start there. So for folks who are normal listeners to our show and who don't know ProSci, give us a little teaser for why should they turn it up this week? Very good. Thanks. and excited to be here. ProSci is at its core an organization focused on understanding why there's a pile of successful change and what pile of unsuccessful change.

1:14So founded by an engineer 25 years ago, just to start to explore what are the differentiators when we over deliver on change and outcomes and when we end up struggling. And sure enough, the people side of change was right there at the very top. So for as long as we've done change, we've designed, developed, and delivered great technical solutions. And then often wished and hoped that the other side of that change coin, where people engage, adopt, and use the solution, actually comes to life. And so, yeah, we put a bunch of science behind how do you put people in a position to succeed when you ask them to make change inside an organization.

1:50So maybe we start with why is change so difficult? I mean, at its core, I guess I'll take this one, Tim, but inertia, right? I think fundamentally, when it comes to why change is hard, is that people are hard, right? People have a lot of inertia. Organizations have inertia. And at the end of the day, there tends to be large parts of organizations that, I mean, as you think about, you know, 10 ,000, 100 ,000 person organizations, a lot becomes like calcified and ossified about the way things operate. To make real change happen, a lot of times you have to sort of thaw that out and move people in a new direction.

2:30And that's just a tricky endeavor, right? but from the technical side there's like requirements you can sort of set them in you can build your solution there's no one sort of like resisting that the technology you might fight with a little bit but ultimately you can sort of control the outcome when you think about changing behavior there just tends to be a lot of baked in in terms of systems incentives culture all those things have like a massive amount of inertia and at the core that's really why change tends to be hard. And when it's done right, it can feel so magical, right? Because this is like a huge mountain to climb.

3:03That's really what ProSci has been working on for the last 30 years is how do we help organizations have the right tools, frameworks, methods to make change a little bit easier. It's not going to be easy. It's still hard. But that's really at its core why it's so challenging. I don't know if you have anything else to add on that, Tim. Well, I think this AI change is a fascinating, even a different type of change for organizations to navigate. And so Ethan Mollick talks a lot about the secret cyborgs, right? The folks that are sneaking AI into work in their pockets, while at the same time, we're watching enterprise AI deployments really struggle to gain traction.

3:40And so not only has change hard, you know, human inertia and all the making sure we answer the questions people have when they have them, this particular change is really unique in terms of the individual and enterprise size of the coin and how that's going. I think a lot of folks are experiencing that for sure. A lot of the organizational leaders feel like this is a particularly gnarly change management effort. Why do you think that is? What's unique about the generative AI transformation? Well, I think the first thing I had to start with is that nobody ever smuggled an ERP to work in their pocket before.

4:16Like the commercial accessibility or the consumer accessibility to the very top models is like nothing we've ever seen in comparison to a lot of the other technology deployments that were rolled out. And so I think that's one of the big, big changes. And then Paul talks a lot about just the pace of this change and what it truly means to the human side of organizational change. Yeah, I think a lot of leaders that may not be taking change or change management all that seriously, a lot of times approach change as like a comms and training endeavor. Let's tell people what we're doing and then let's give them training.

4:56If you just believe that's true, which by the way, we do not. At its very core, from a comms perspective, it is a tough message to deliver, right? If you're leaning in from a leadership perspective and just saying this is about cost, that's like not a really inspiring way for people to get on board. let's go why am i gonna like work myself out of this right and on the training side i mean when you look at the the pace here you look at things like you know performance is doubling every six months we're just not used to that type of pace of change of like transformation technology so what that means is our old playbooks around hey we're gonna like you know tell people what they need to do and then we're gonna give them training on how to do the thing within three months that might be obsolete, right?

5:37Everything's changed. I would have taught all this stuff about prompt engineering and chain of thought. And then all of a sudden reasoning comes around or like, hey, you can't do this with image models. And now all of a sudden, that a banana's around. Like, there's so much change that's happening that if you just take our old tactics around, you know, comms and training, again, that's always been limiting, but certainly the pace is different here. So we need a different approach for how we like get people. Do you think there's some reason why some people seem to be excited about this new stuff and kind of like see this an opportunity, see the opportunity for learning, see the opportunity for getting more autonomy.

6:10And there are some people that seem almost kind of angry at it where it's like, hey, this thing is coming from my job. I want to resist it. Is there a profile or how do you think about that? I think not necessarily. I don't know. I'm not sure about the profile of person. I think your first exposure matters. Like when you watch it truly do something that's value add above and beyond, you know give you a recipe for tomato soup and snoop dog's voice um i think when you see people go into that moment of this is a partner for me to to work with to achieve whatever it is i've set out to achieve i think that's the moment when you see people kind of tip over i love that you go when they see something of a true value tomato soups and snoop dog's voice say oh well for the record for the record truly valuable yeah i wonder by the way like contextual hyper personal right because the thing about ai is it can do stuff hyper personally it gives us the ability to actually get done what's right in front of us and so when i ask it to teach me about mcp servers i ask it to teach me in the terminology of barbecue grilling because i grill a lot and so like the ability to mash up different bodies of knowledge is i think one of those things that i don't know.

7:26I leaned into it. And I think when you watch people mash up its capabilities and truly something they're trying to unlock that, that's, it's that hyper personal experience. So you're kind of bringing in, we've heard what the bad is, right? Bad approach to change management is comms and training. If you could wave a magic wand and kind of prescribe, what are the ingredients to bend the odds in the direction of success? What are the kind of call it top three or four key leverage points. Assume you have way too much demand. You've got a thousand less spots than customers are demanding. And you have the ability to sort through a customer list and say, we're going to take on one client this quarter that is primed for transformation.

8:12What are you looking for in terms of being primed for transformation? So rather than wave a magic wand, we actually did research on this. So we did research with 1 ,107 participants, 500 frontline, fund 400 team leaders, 200 execs, and asked them all about how they were using AI, perceiving it being used in their organization, bringing it into their particular part of the organization. Then we looked at all of that data through the lens of what are the patterns that are emerging that demonstrate the conditions where you get successful AI deployment? Because there's a big pile of unsuccessful ones, but there was a pile of successful ones.

8:49And what are the conditions that really exist there. So we turned it into a bit of a diagnostic, but if I were to pick a couple of the ones right at the very top, leadership and bold AI vision. But what also came out of the research was a balanced vision. So a bold and balanced vision that really articulates how this technology will transform us, but also what we are doing in the near term to try to bring it to fruition. And I know even in some of the conversations with other guests, right, Paul, there's been like some real specific examples where we're like, we're hearing bold and balanced vision right in the conversation.

9:25Yeah. Give us an example. What's an example of bold and balanced vision? Well, I think there's a couple of first stuff when you talk about like the code reds, right? That happened. Like that's like from the top, right? Leaders at the top are really saying, okay, this is an urgent moment for us. And it's an urgent moment for the future of our organization. And we are going to attack it holistically. That means finding ways to use AI internally to, you know, manage our organization more effectively, more efficiently. That means finding ways to use AI to solve customer problems in ways that we never could have historically, right?

9:57That means AI out for our customers, new products, new services, new offerings. So it really needs to be sort of balanced in that approach. It can't be, hey, we're just going to use it to like cut out a whole bunch of call center costs, right? Again, not inspiring. So it really needs to be across that wheel of both internal, external use cases. and that allows people to sort of internalize, you know, what's in it for me? How might it help my team? How might it help my department and my organization? And again, I think if you put the customer at the center, assuming you have a culture where people are bought into the fact that we are here to serve our customers in a meaningful way, that's part of what Bold AI Vision looks like.

10:34And I think, you know, one of the things Sims brings up is that this is also not, while it is like very specific to AI adoption, in this case, this is not new. You know, strong sponsorship at the top tends to be a leading indicator of transformation success. You know that, we all sort of know that. But sort of that key role that sponsors and leaders play to get their team on board and sort of paint the vision of the future has always been an important part of change. But specifically when it relates to AI adoption, there's just a couple of different nuances like this balanced approach, like this bolder vision, other than just, hey, we need to cut costs.

11:10Can we dive into the word balance for a second? Because I think bold everybody gets, but if I, if I, I kind of like to think in terms of Goldilocks spectrums, right? So balance is kind of in the Goldilocks spot. I would say a little bit obvious on one end of the spectrum, which is this is all about efficiency and productivity. And we're going to do, you know, it's only going to take 10 % of what it used to. Right. So we get that what's on the other end, what should be equally avoided as just leading with productivity. The other balance I would, in addition to kind of the content or the topic, I think there's a timeframe dimension around balance.

11:49And we're talking about the near term work we're doing, the wins we're having, the experiments that we're promoting, the pilots that we are shutting down intentionally on purpose so that we can put our focus where we need to. And we're also talking about what we're going to look like in three years as an AI-infused organization, where all of the work that we do in zeros and ones ultimately is up for grabs to be touched or disrupted by this new technology. And so leaders are, from a balanced perspective, are talking both about opportunity and efficiency AI, and also near-term impact and long-term vision.

12:31On a very practical level, as you are advising companies that want to do this, and we talk about getting the leadership involved, I get the sense that there's a lot of CEOs and board members that are very big on getting kind of the AI, AI coefficient up in the organization. Then sometimes that could be, it could be more challenging, even the senior management, let alone the middle management. As you kind of see successful kind of implementation, how important is it that you have top and then all the way through the organization? Or can you have like sporadic kind of like bright lights that you can build on?

13:08I think, and it was interesting, Jeremy asked about like what metric we might use to decide if this organization is set up to succeed. And the one that came to my head, and I'll have Paul talk a little bit about it, is exposure hours of the executive team. if you could give me literally a measure of how many hours executive team has spent with their fingers on the keyboards not talking about ai not just listening to ai but actually pushing the boundaries pushing the buttons i would put my money that might be the metric i would use to decide if this is going to be a and so paul exposed me to the notion of exposure hours but i think when we talk about what's different for ai change and readers i think this is a key piece yeah i think it matters a lot and to your point henrik you know we think about how that cascades down the organization um you know i think there could be a few bright lights in terms of champions and how that plays out i do think at some point you end up with a bit of like a swiss cheese state at the end of this transformation which is that if there are certain department leaders that are resisting it certainly if they're openly resisting it in some capacity that's going to be a problem for how that then cascades down within their departments that is like a known challenge um So I do think it starts at the top.

14:17It starts at the executive team level. I do think you need to build a strong sponsor coalition across that sort of middle layer of the organization. And, you know, champions can carry things, but it's also not their job necessarily to champion this across the whole organization. and I think to Tim's point the amount that those leaders are engaged with using AI themselves tends to be a good insight just because they start to see patterns they start to see what use cases might be tailored for their teens to me that's like some of the leading indicators that you see so I think you really need to think about holistically enabling your your leadership level executives and that sort of key sponsor coalition level with at least some AI literacy AI foundations so they can start to pattern match, right?

15:03How it might apply to their team specifically. Okay. So we've covered the first one, bold AI vision. Going back to this question of what are the three or four kind of leverage points you're looking for? What's another, or a key indicator of readiness for transformation? What's the next one? Next one, I think is really around that kind of people side and the experimentation culture. So the organization encourages employees at all levels to experiment with AI tools, share insights and learn from small failures. That was kind of the statement that came out of the research. And again, we started to see the success pile was not just decreeing experimentation, but it was creating a system, the conditions, the structure to encourage that rapid learning, sharing kind of within and across the folks that are really doing the work.

15:49So how do you build that system? I mean, that feels like on the one hand you go, it's either present or it's not, you know, pre-AI or not. Now, say an organization goes, OK, we need to build an environment where experimentation can be encouraged and unleashed. Are there are there ways to establish or to cultivate a culture of experimentation if it's not present? I think there's a few things that come to mind for me there, Jeremy. First, I think I was talking to one leader and they were bringing up Toby's like Shopify memo about like how we're going to move in this direction. and they're like, we need something like that.

16:27Like that would not work here. And like the way you've hired, the way you've built the organization is really around a builder's mentality. It's a lot of experimentation baked into the culture. That's not gonna be the right like option. You can't just jump to that in an organization. So to me, there's like a few things really, German that I think about. First off, it's really disingenuous in my mind to say like, hey, just start experimenting, right? If you haven't built that culture within the company, it's not what's rewarded. it. It's not what's incentivized. Hey, it takes some time to fail or try things.

16:58It's just not baked in. I do think you have to carve out some time specifically within teams on very low stakes things, right? If you say, hey, let's do a hackathon to solve this one problem. That's an actual problem we have. People are ready. Like, am I going to get this wrong? Am I going to get this right? I think it's almost better to find some time that's almost completely unrelated just so that they can start to use and try out the tools. I also think you need to, in these moments, take an opportunity to broaden out the tool set that people are trying within AI to sort of see the art of the possible.

17:29If there are just sort of enterprise AI tools that they're using, maybe they're not getting exposed to some of the prototyping tool to try things out, right? Hey, did you even know that you could build your own landing page or build your own app or do deep research on a topic? So I think as you start to lower the stakes on how they're experimenting with it, give them some space to try things that are maybe not directly even related to like a problem they have. It starts to create a little bit more psychological safety around trying something crazy, failing a little bit to sort of build that culture in.

18:03Have you seen other examples? We've talked to a lot of folks that have done hackathons. We've done a lot of CEOs who have like a session where they do show and tell or they kind of like share the other team for some of the organization involved you know like their loom is being used a lot to kind of like show a different example in an effort to try to create like i would say like excitement and in like entertainment out of like this have you seen other tools that people are using just to kind of like get the excitement level out so it becomes as much about like here's cool thing you can do with it that makes you or your customer kind of having a better experience rather than just yeah here's the thing that will make your make your job for you?

18:44Yeah. I mean, I think, I mean, just one small example, even internally at ProSci, I don't think everyone had known about what you could do in Claude with like artifacts, as an example. Someone on our sales team turned in like, or they were sort of sharing out the pipeline, how that was progressing. They built like a bit of like Claude artifact, like small app to showcase that. Like what a great way to actually experiment, try something and then show it to the rest of the team. And they're like, Hey, how'd you do that? How does this play out? So I think there's some of that Henrik that is. And that was through Slack or was that that through like in all hands or what's it actually kind of like the mechanism yeah in that regard we have like a weekly email where they're sending stuff out for what's going on with um the pipeline and then that email they all of a sudden i see this cloud artifact i'm like oh what is that um so again that's something that as an email i would typically read anyway but when i see this this new way to expose um the data the content bring it to life in an interesting way And now I'm going out to that person and saying, hey, that was interesting.

19:42Like, are other people in your team aware of this possibility? How might we sort of bring that that other places? I think there's like, again, to me, that's like a bit of a novel, at least internally, a pro-sci was a way to share that out. Now, one of the things I wonder is this, where does experimentation live? One thought is we're all experimenters, you know, and everybody's experimenting, right? Which is you call it building culture, whatever. The other thought is more of there is a team that's dedicated to AI experimentation. I think we had Adam and Andy who wrote AI first on recently, and they kind of reference Ethan's interview with them is in the kind of, you know, the, the last chapter of the book, I read the book after our interview with them, which I enjoyed.

20:27And, um, one of the things that Ethan talks about is establishing an AI R and D lab that every organization needs an AI R and D lab. And to me, that actually, that shows the way towards a little bit different structure, which is it's not everybody's job to experiment. It's this labs. I mean, I don't know how you defined lab, but can you talk about whether experimentation has to be kind of broadly distributed or centralized? Yeah, I think, I don't think there's a reason to veer off of Ethan's framework too much because I think it makes sense. And the important part to add to that is that in his framework, it's not just the lab, right?

21:00There's leaders, the lab and the crowd that are all working together to create this, this AI transformation. I think the lab is going to be that area where those are the people that as GPT-5 is released, as NanoBanam is released, are on the frontier seeing what might be possible. But you still need access to the crowd of like employees broadly to see how that might actually apply to drive outcomes for the business, right? Like you need to still have access to people who are dealing with problems on a day-to-day basis with customers that are trying to get through a process. They're trying to build a new, like a new solution or something that can then like push the learnings from the lab into the crowd to actually see if it's making an impact on the business.

21:38So I definitely like the idea of like, I don't expect every employee everywhere to be up to date on the latest and greatest from AI and what's possible. I mean, we're us four probably in the weeds way more than others. And I don't even know if we can keep up with all the change. So it's disingenuous to assume that employees can. So I like, in that regard, keeping sort of a centralized lab to experiment with the latest and greatest. But you still need to get that to the edges of the organization to see the impact of how it's going to actually impact the organization's day to day or customer outcomes in some new way.

22:12So that. Can I ask a little bit on the impact? I mean, like if you look at just, let's say Barbrox that obviously they're involved. We seem to have gone through these like two phases and going into the third phase. The first one is kind of how do you get the coefficient up? how do you get the crowd involved, how do you get the lab up and running, all the, I would say, the fundamentals. And then I think we've been through a phase where we're like, okay, let's figure out how do we basically create an agentic kind of framework for a lot of the different organizations. So how do we take everything from supply chain to HR and we create different agents that can either enhance what that organization is doing or make it for them, right?

22:52The third phase seemed to be this kind of like emerging kind of like new architecture that almost need to be of the organizational design. Because what I don't see is that we're just replacing kind of FTEs with agents. What I'm seeing is that suddenly the designer can do a little bit of the marketeer's job because, you know, they can do something, but it can also write a little bit more basically like the narrative around like a toy character, for example, right? Or the supply chain could do a little bit of like the planner's people. And so the FTE reduction is probably not going to be like, and suddenly we'll just remove people that.

23:32It is that there's like this new patchwork kind of like emerging where people can, if they're allowed to do other people's job. And that would suggest to me that kind of like the third and final maybe phase, at least the third phase, would be that actually like the little bit static orc chart that I guess in many ways was designed for when we did railroads in the U.S. in the late 18th century, right, like is kind of like being in many ways displaced. A, by this thesis, and two, how do you feel people start to do the leaps from kind of like phase one to at least phase two, or at least phase two to phase three, as we obviously are also seeing like a lot of kind of writing about people who think they are doing a lot of like the phase one stuff, but they're not really seeing the returns and therefore getting increasingly frustrated.

24:24I'm pretty aligned with kind of your premise. I think very early on, I got exposed to a website. It was called There's an AI for That. And it was essentially just a big library of AI tools, but it was organized around different jobs. And when you clicked into a job, it gave the tasks that that job executed where AI could augment. And that was kind of a moment for me to realize the task. I think the task is the denominator here. The task is what we pile together to make jobs. We string tasks together to make workflows. We string tasks together in project plans to make projects. I think if we start to actually break down the work an organization does into the tasks, we can start to reassemble and redesign the org in really neat new ways.

25:11And I think you're right that AI expertise and capabilities has enabled folks to start to expand into some different tasks that they might not have necessarily extended into before. It does take a huge role in redesigning what work actually means in the organization. And there's even this, you know, the notion of the work chart that you're starting to hear talked about above and beyond an org chart, but really looking at the assembly of tasks to achieve the outcomes that we as an organization are setting out to. And then which of those tasks do we want to keep human? Which of those tasks can we agentify and automate?

25:48And then which of those tasks do we bring together in sort of an augmented fashion? And do you think, how far do you think we are from like these new type of organization designs that are more agentic? I seem for the startup side, We seem to be seeing it more and more because you start to see companies that are kind of built from their AI kind of like from inception. So when do you think we'll start to see this not just happening in the startup side of things, but like happening in organizations that you advise, for example? Yeah, I think you'll in the next couple of years, I mean, you'll start to see things where let's say like a new product line is spun up.

26:25You might take this approach right out of the gate on like a new product. So as an example, like even in ProSci, we're launching some new products. The team stack feels very different. as we're starting to invest and build there, just because there's much more of like a IC, super IC sort of focus where people can do more with their current skill set and sort of bleed into other things. I think the bigger change, Henrik, honestly, around that as we move forward will be how do organizations align incentives, align career ladders, align compensation, because there's a lot changing. In my view, this is actually the bigger change for organizations than the technical side over the next two to three years, because it is incredibly disruptive.