How to Create Decision Clarity During AI Transformation

Galen Low: A RACI chart is one of those things that people either love or hate, mostly hate. Not because it isn't useful, but because it's a bit too easy to get carried away documenting who is responsible, authorized, consulted, or informed for every facet of your business instead of just getting work done. But as organizations get deeper into their AI transformation journey and start reimagining the very structures that the work exists within, clarity on roles, responsibilities, and decision-making is becoming a wee bit more important.

So today, I've brought in the person that I know to be the expert on the subject of RACI-led decision clarity, and who also happens to be working very closely with enterprises and government agencies on their AI transformation. Together, we're gonna explore whether RACI can be used not just as a tool to capture how decisions are made today, but also as a language and a mindset to design the organizations of the future. Hope you enjoy the episode.

Welcome to the Digital Project Manager Podcast—the show that helps delivery leaders work smarter, deliver smoother, and lead their teams with confidence in the age of AI. I'm Galen, and every week we dive into real-world strategies, emerging trends, proven frameworks, and the occasional war story from the project front lines.

Whether you're steering massive transformation projects, wrangling AI workflows, or just trying to keep the chaos under control, you're in the right place. And if you've been liking what you've been hearing from us lately, please consider following us wherever you're finding us and maybe even leaving us a review. All right, let's get into it.

So today we are talking about AI, decision-making, and why clarity around roles and responsibilities is more important than ever. With me today is friend of the podcast, Cassie Solomon, president and CEO of The New Group. Cassie is an award-winning author and enterprise consultant who is an expert at leading successful change.

She has helped senior leaders transform their operations at organizations like Twitter, Pandora, Bayer Pharmaceuticals, Penn Medicine, Geisinger Health Systems, and ChristianaCare Health Systems. In addition, she teaches executives at the Wharton School of Business, and her book, Leading Successful Change: Eight Keys to Making Change Work, is still the manual helping many leaders navigate today's pace of disruption.

But for folks who have followed this podcast over the years, Cassie is the person to talk to when it comes to using RACI as a tool and as a language to clarify roles, responsibilities, and decision-making so that organizations can accelerate their transformation velocity.

Cassie, thanks so much for being with me today.

Cassie Solomon: Thank you so much, Galen, for having me. It's always really fun to be talking to you.

Galen Low: First of all, I just wanted to hit you with a spicy question that my listeners would love your take on, and then I'm hoping we can just unpack that into some practical tips for avoiding AI role and responsibility chaos on projects, and just for organizational decision-making overall.

And then maybe to round out, I thought maybe we could just look into the future of how decision-making, teamwork, and org structures will look with AI embedded, and what we need to get right today in order to make it a positive thing, n-not a negative thing. How does that sound?

Cassie Solomon: That sounds wonderful.

Galen Low: Let me then start out with a bit of a spicy question, because at this very moment, almost every organization that I talk to is trying to create AI agents and incorporate them into their operating model, and these agents are meant to operate with pretty much relative autonomy, which specifically includes making decisions for themselves.

And you are, in my books, the expert when it comes to using RACI as a language to create a shared understanding of who is responsible, authorized, consulted, and informed around decisions made throughout the collaboration process. So my question is this: Is RACI still even relevant in the age of AI?

Cassie Solomon: I think yes.

I wanna thank you for plugging the book, Leading Successful Change. It's a way of describing a system. One of the eight levers happens to be decision-making, and one of the things I love to point out when I'm teaching that model is this is an aspect of our organizational life that is largely invisible.

I can show you my compensation plan, I can show you my org chart, I can show you my technology, but it's very hard to show you my decision-making, and yet it's a very important part of the system. And so it's a very interesting lens to look at AI transformation through because we're about to get into a kind of change that I think really transforms our decision-making practices, and we don't even half the time know how to talk about that.

Galen Low: I like that you brought up the transformation for two reasons, especially AI transformation. A, I agree with you. These aren't things that we always r- you know, write down. They are, in a lot of ways, unofficial, invisible, you know, documented to a very limited extent, which is But bad for two things.

A, it's bad for AI because if we haven't trained it on it and it doesn't know and it can't find the information, it won't know until we tell it. And the other one is yeah, transformation requires this clarity. Everyone's moving at pace. They're trying to do things quickly and stay competitive, and this is the moment where decision-making and roles and responsibilities and, you know, the things that are in flux need to be talked about and then written down so that we can keep moving quickly and not getting, you know, stalled out in decision purgatory or in massive ambiguity.

It is the way we talk about it.

Cassie Solomon: Yeah. The way AI gets talked about in the popular press often, you think, "Oh, is this about replacing someone's work? Am I gonna replace this worker because I no longer need them?" And then this other word comes along which is, but wait a minute, what about judgment? What about human judgment?

The judgment word actually refers to expertise and authority, so judgment often comes in at the point where someone is saying, "That's a good decision," or, "That's a bad decision," or, "Wait a minute, I better make that decision. I have 25 years of experience in..." So, e- embedded in that word judgment, I think there is already we're...

It's pointing us in the right direction. It's saying, "Look over here. This is decision-making territory."

But the other thing I wanna talk about is Ajay Agrawal, Josh Gans, and Avi Goldfarb's book, Power and Prediction. It came out in 2022, so it actually came out before ChatGPT was launched, but it's pointing us towards this understanding of AI as a prediction engine and a way of making better decisions.

We're kind of like awash right now in all these AI projects and AI conversations and... There was a really good study in 2025 that showed something like 80% of AI projects are not achieving productivity. That was an MIT study. And there's a 2026 study from McKinsey that basically said the same thing.

These guys have a model that I think starts to really be helpful in understanding what kind of AI project are we talking about, and their framework is really simple. So the first thing is a point solution, where you just take a single thing that humans used to do and now we do that with AI. I kinda think of that in an individual frame as when I go and ask the AI to make my meal plan for the week, or if I go and ask the AI to do the research for my paper.

That's a really good thing, but it's limited. The second level up for them is called application solution, and that's where you're changing multiple things at the same time, and that might be where we're getting closer to now with AI agents. The ultimate, the highest level, is systems applications, which is where you're stepping back and saying, "How can I transform the entire system using this technology?"

And those are much harder to achieve, and there are fewer examples of them. But their theory, and I agree, is that that's where we're gonna see real impact on the bottom line, if we can get to the total transformation piece.

Galen Low: I really like this framework because it gets us into the head space of not all AI transformation projects or projects, or not all AI applications are, are the same.

And, you know, while it would be easy to be like, "Okay, well now I can replace this one tool with ChatGPT," or, "Now I can add an agent into this workflow," there's this other tier that we might need to think differently about because the implications are bigger, the design challenges are bigger, and the problem is bigger.

And it kinda puts us in this territory, as you were saying it, I'm like, oh, that puts us in territory where we're not clinging onto the side of the pool anymore. We don't really have a comfort zone to go back to. We're sort of out in the wild, in sort of net new territory, and we need to sort of figure it out, not based on what we have before, but based on something new altogether.

Cassie Solomon: I think if we do point solutions first, which makes sense to me 'cause they're the simplest to imagine, that lends us to thinking, "Oh, I don't need that 8,000-person workforce to do that task anymore." You know, that's replaceable. But I just was reviewing some research this week which was really interesting.

Last year, so this is pretty recent, out of the UK, about 30% of the organizations that let people go because they're going to implement AI end up hiring them back. And then there was a, a second study, which wasn't done by the same people, that basically said, "And oh, by the way, when you hire them back, they're more expensive the second time."

I think by trying to be broader in our thinking and going towards the solutions and the transformation piece, we can start to ask better questions. If these employees are no longer spending time doing this, what else could they do? I mean, the classic example from back in the day, which most of us remember, is when ATMs came out and people were like, "Oh my God, there won't be bank tellers anymore because the machine can now give us money, and that's the death of that entire job."

But it ended up not being true. It wasn't the death of the entire job. They just decided, "Oh, tellers are still important to us 'cause they can have a conversation with our customers. Let's do more with them." They became more business development kinds of people. And I think that that's the kind of thinking that will help us kind of tamp down some of the, the night terrors that we're having about the workforce and will AI replace all the jobs.

Manufacturing, different case study, but it's, it's still a good- aspiration, I think.

Galen Low: I really like the idea that almost the point solution is what's scariest because it just looks like this swap, and you're like, "Okay, well now a robot does my job." I'm glad you brought it back to what we've seen in history, recent history really, of disruption and how it can go.

And I, I like that tilt towards, okay, well, when we start zooming out at the system, right? When we're looking even at the sort of application level, you're like, "Okay, well I guess we have to lay off all the tellers." And then you're like, "Wait, wait, wait. We need them back," because actually when we zoomed out to the system, we're like, "Wait, there's still value that can be created in the system.

We just have to think about it in a net new way, not using our old model." We almost react based on what we know today, even though the mission is to, to figure something new out for tomorrow.

Cassie Solomon: It's 100% mindset. One of the things I'm obsessed with, which my friends tease me about, is the transition that the world went through from the steam engine to the electric motor, and it took such a long time, Gil.

And it was like a 20, 25-year transition. And one of the things that really held people back is they had a very hard time imagining what the new technology was capable of doing. And the firms that figured it out just succeeded and thrived, and the firms that didn't figure it out ended up falling by the wayside.

And the, the thing that intrigued me the most about that example is they, they would build new factories after the electric e- motor was Already adopted, but they would put them in the same central location that they used to put the giant steam engine, 'cause they couldn't imagine a different layout for the factory.

So a lot of it is mindset, kind of what can we imagine to be different?

Galen Low: I love that, and I think the-- it's relevant today. It's very relevant today. It's like, okay, well, where are we gonna put the new engine? Well, the same place that we always put it. And I think there's a lot of parallels there in terms of well, where are decisions made?

Okay, well, where are we gonna put it in the new model? Exactly where it was before. That's probably our best bet. We don't know what we don't know, and it's very you know, it's, it's such a human thing to be like, "Well, we'll just do it the way we've always done it," and that should be fine. We've, you know, gotten this far doing this.

And then there's that moment sometimes where you're like, "Actually, no, we have to think bigger. We don't need it to be here anymore."

Cassie Solomon: Well, thank you for bringing it back to decision-making, 'cause I have a really nice example that we're gonna cite in an article that we're working on right now from McKinsey about a huge company called Blackstone.

And they, they went in to transform their legal and compliance work because they really anticipated just a incredible growth in volume, and they, they had been processing everything manually. They couldn't afford to add 25% more head count. And they came to McKinsey and said, "Well, we, we should be able to do this with AI."

And McKinsey said, "Whoa, slow your roll. We're not just gonna slap this technology down on top of your old process, because that's not gonna get you very far." That's gonna be like the cases that we see where people say, "Oh, well, I used AI, but nothing really changed. Didn't impact my bottom line." What McKinsey said instead was, "You know what?

You've got all these layers of approval. You've got this way of doing this process that's really clunky, and you have to clean up that process first and streamline the decision-making and automate as much of the rules-based decision-making as possible, and then create an escalation pathway if somebody doesn't like what the, you know, the new automation accomplishes, and then you can put the AI down on top of it."

And after they f- cleaned up the clunky process and adopted the AI, they saw unbelievable productivity gains, like 30% more productive. Senior people weren't doing these kind of multiple reviews. They were freed up to do other things. There were just a lot of benefits that came out of it. Now we're back to my friend RACI.

You know, can you use RACI to diagram the process that you're currently using? Can you notice how clunky it is in terms of How many approvals you're tormenting yourself through and can you streamline that stuff before you adopt?

Galen Low: Torment is such a, a strong and good and appropriate word for some of the things around decision-making.

I think it's it's interesting because, I mean, so at this very moment, you are actually launching a RACI certification program, first one I'm aware of. And the idea is to sort of help enterprise organizations train people at, like, all different levels to harmonize how decisions are made throughout their entire operation.

But I thought I'd ask you what's something that is fundamentally different about using RACI in the AI era that business leaders and project professionals need to understand to kind of avoid the, you know, steam engine in the same place or the, the electrical, you know, like the, the, the, the mindsets of the previous age getting through to limit progress in, in the new age?

Cassie Solomon: Oh, what a great question. So let's roll back to what you started with, the RACI certification course is in beta right now. The, the first real running of it will be in October, and we're, we're searching for the right title for it because a lot of people say, "Well, these RACI's this very simple tool.

Why would I need a certification? I can go online, watch a couple of YouTube videos, I'm done." And I completely agree with that. The tool as a tool is pretty simple, but we're trying to teach it as a skill or a capability, and it has so many applications to creating engagement and high-performance teams. So I often am contacted by companies who say, "Oh my gosh, my employee engagement survey came back and people are so frustrated with our decision-making processes.

They can't get anything done. Can you help?" I'm like, "Yes, of course we can help." But one-- So one of the things we're baking into this course as a skill is, how do you step back and do some kind of decision audit? How do you look at your decision-making practices and map them, you know, just as if you were mapping a workflow?

Except that we almost never map our decision-making processes. We have a lot of emphasis on what we're doing or on how we're doing it, very little emphasis on where is the power, where is the authority, how are, how are we using that in the organization? So first you have to see what your current state is, and that's a RACI capability.

And then you have to start saying: Why is this so slow? And, you know, the PMI told us a long time ago, you're only supposed to have one A for every project. Mm, you know me, I tend to disagree with them and think that that's not terribly realistic. And in fact, we actually wanna push decisions down in the organization if we wanna achieve speed.

We're all coming from a legacy, legacy systems that are pretty hierarchical and pretty command and control. If you look around, the world has been migrating for now a couple of decades to flatter and less hierarchical and less command and control and those organizations are more agile and speedier.

So step one is diagnose what you've got right now in terms of decision-making. Step number two is step back and look at that, and then step number three is redesign that with an eye towards empowering people so that you're pushing decisions closer to the edge, I think is what they would say in the military.

Galen Low: I really like that in terms of the nimbleness of it. It's funny because, you know, you mentioned PMI, and I'm sure at some point in the model, like original RACI, they were like, "Okay, one, one A, one accountable person, that's gonna be the most efficient." And then today we're like, we're realizing that that's a bottleneck actually.

And a lot of what, you know, the people we talk to are running into that, is that having a single decision maker for, like, all the things is not always the best approach. That person might not be equipped or informed enough to make that decision. And then when it's not clear who that person is or when there's any kind of waiting game, it slows everything down.

Meanwhile, there's a lot of examples in the world of, yeah, teams that are more empowered to make decisions, you know, at, at, at the, at the fringes or closer to where value is being created, and that's, you know, supporting this more nimble thing. And the other thing I f- find fascinating is that y- I think you're right, right?

People are like, "RACI certification, yeah," you know. "What are you gonna do? Teach me an acronym in five minutes, and then I have homework, and then I'm done?" But it's not that. It's actually RACI as a, as a tool, as a framework, as a mindset for high performance teams to understand where decisions are being made.

And then the other thing that I, I, I think you and I were chatting in the green room or, or as we were prepping for this, but you introduc- introduced this notion of RACI as a design tool as well. It's like now we've got ... We understand what we've got now, but this isn't just about documenting current state.

This is about understanding in the new world, in this new episteme, in this new framework, in this new era, where does authority lie? W- who can make decisions? And then sort of creating the system, right? That system level change, the sort of tier three transformation in your model.

Cassie Solomon: Although I actually wanna circle back because the one decision maker works really well for startups And they usually don't need RACI.

They don't need that structure. They're like, "I'm the founder. I make the decisions. I'm not a bottleneck. And then we have a very small team. We like to get together the way an agile team gets together, and I get some consultation, everyone's clear, and we move very quickly." Now you get into success, your startup starts to grow, that's when the one decision maker starts to be really painful, 'cause you've got a founding team, they're really used to touching every decision, and they have all of this history and they're familiar with each other, and then you just add 350 new people, many of whom come in with tremendous experience of their own.

They're great. And then it's chaos. That's usually the point where the decision rights issues emerge and the pain points are. The other point that I think we talked about a little bit in the green room is the analogy to the electric motor being distributed. Th- those motors getting smaller and smaller and then distributed throughout the factory is kind of a nice metaphor for the way we're flattening organizations now.

But the reason that we're flattening them is because of access to information. That's a really interesting signal to pay attention to. So the one that I was telling you about earlier, there was an article in yesterday's New York Times about the transformation of the Ukraine-Russia war as they've moved more and more into drone warfare, and we'll talk about when does the machine start making the decision, 'cause that's what everybody's gonna wonder.

But they are now using satellite imagery, which we're, we're getting to the front lines in Ukraine, to look over the, the landscape and identify targets. But now the technology has improved, the access has improved. They're now pushing that technology out to the field, like into somebody's phone or iPad. If they were still making the decisions back at central command, they wouldn't be taking advantage of the fact that the intelligence is now right out there in the field on the edges, and so that's where the decision-making has gone.

But they haven't gotten to the point where they want the drones making the decision. They still want human in the loop, and I think that's really important, at least at this point in our evolution with AI. A lot of the most important decisions in clinical care, for example, you know, I saw something today, which I haven't even had a chance to read, it's only 80% of doctors are using AI.

But they're not using them without that element of human judgment and human in the loop.

Galen Low: What you're saying is really interesting because it's not just the technology that's bringing intelligence closer to the people who are doing the work or creating the value or, you know, executing on the mission.

But also there's this sort of like trust and I guess education that comes along with it. Just because somebody has the right information or the right intelligence to take action doesn't mean that, A, they have the training or that they have the view of the bigger picture to make really good decisions.

How do you solve for that? I, you know, by all means, we can, you know, have a direct pipeline of information going to, you know, our project teams or, you know, people on frontline customer support. But what's the thing that also helps them make good decisions once they have that intel? And how can organizations even trust that giving people the right information will actually drive better decisions?

Cassie Solomon: We built an empowerment model based on just years of RACI consulting- because I feel like that, that word and that concept is just too broad, and it gets into all this normative stuff like, oh, I, I don't want to be a micromanager. I want to empower my team, so I have to back away from these decisions.

Oops, that's not going well. No, so I better step back in. Well, now I've just discouraged everyone and demoralized the entire system, but-- But one of the ingredients in our model is this idea of skill which comes from the Dreyfus research that happened in the m- mid-'80s, which now seems like 100 years ago.

Basically Dreyfus and Dreyfus studied how do human beings build skill. And it starts out if you're a novice with absolute step-by-step checklists. I'm gonna tell you exactly what to do because you've never done it before, and you're gonna have to grow step by step. I always say I'm a really experienced cook and I'm a terrible baker, so if I'm gonna bake something, I have to go back to that recipe like 15 times.

My friend Laurie is a genius baker who's been doing it for decades. I was like, "What are you doing?" This is all so intuitive to her. And as you gain experience, you march up this Dreyfus model to the point where your experience starts to become judgment. And by the time you're at the top of their ladder and you're an expert, you can't even remember the checklist anymore.

You've c- so internalized all of those steps and what you learned initially that among other things it makes you a miserable teacher of new people coming in, 'cause most of the time what you say is, "Well, that's just common sense. Why don't you know that?" So at the time, which was the mid-'80s, they said this proves that computers will never achieve intelligence because all they can do is follow rules.

That was the state of the art in the mid-'80s. Then machine learning came along and lo and behold, computers started to teach themselves and gain experience just the way humans do. And now they can put two and two together, and they can break out of their sandbox and They can go hack Hugging Face, which no one anticipated, you know, when Dreyfus was writing.

But the model is still really useful because if I have somebody on the frontline, and that's what actually what the New York Times article said, they are keeping the human in the loop 'cause they don't want inexperienced soldiers making those decisions by themselves. And so the question becomes, when does the drone make the decision by itself?

And that's a really interesting question for medicine as well, because we, we always test our AIs, our clinical AIs, against expert clinicians, but they, they way outperform the, the junior people. So when you're asking trust, do would I trust an AI to make a medical decision for me, or would I trust this doctor, my first question would be, how much experience does this doctor have?

Galen Low: And it's interesting that you mentioned machine learning because, I mean, in a way we're talking about experience because you kind of get exposed to the work and decision-making, and are supported in your decision-making, and often have an opportunity to make the wrong decision a whole bunch of times so you figure it out, and, you know, you have to iterate through to get that experience.

So the experience is not just years of just sitting and looking at stuff. It's actually by making decisions and learning from them. Machine learning obviously happens a lot faster, right? You know, like the, the, the way AI has been taught to, you know, identify a cat versus a dog is just by getting it wrong a whole bunch of times and being like, and figuring it out.

So arguably they're gonna get more experience in decision-making through that process. But I, I agree with you that it becomes this like almost more of like the human comfort level. Not to say that AI's gonna make the right decision every time, but guess what? Neither will humans, and like what do we feel comfortable with?

A- and actually maybe it's time to go there. Maybe we should go there because, you know, I'm thinking about RACI. We're talking about it in our community in terms of roles and responsibilities, and then like I started thinking. I'm like, "Okay, well, like up until now, you know, we, we've had tools, right? We have software.

We have platforms. They don't show up in the RACI because they don't make decisions. I don't put Photoshop in the RACI. I don't put Git in the RACI. I don't put, you know, Microsoft Word in the RACI because they're not making decisions. We don't have to account for that. And now that's a bit different.

Actually, I wanted your take. When it comes to team roles and responsibilities, should AI be treated as a team member or as a tool?

Cassie Solomon: Well, well, you know I'm gonna answer it in this really annoying way and say both, because if you're at the point solution, then that's a tool, right? That's I need an answer.

I need, I need one thing maxim- you know, optimized, so let me use AI to do that. Think about the way the big banks are using AI to do fraud detection, right? Which kind of used to be human. And now the machines just run through a billion transactions and say, "Oh, wow, that's an unusual pattern. Cassie hasn't ever bought $800 worth of pizza in Brooklyn 'cause we happen to know most of her transactions are in Philadelphia, so I'm gonna text her all of this."

There's no human in this loop, right? I'm gonna text her and say, "Did you just buy $800 worth of pizza in Brooklyn?" And I'm gonna say, "Huh. No, actually, that, that is, that is fraudulent." To take this back to RACI, how much of the R, how much of the work itself Can I subtract and automate? The A, which is the judgment and the decisions, is where we're kind of in conversation about I don't know if I wanna put the decisions inside the machine as if it was a team member.

But in that fraud example, it actually is in the machine. It's communicating with the customer and saying, "I'm checking here." Obviously, when we put these algorithms into our bank fraud systems, we sometimes catch the wrong people. And then they're really hard to unwind. So I'm not saying it's perfect.

I'm just saying we have examples already where decisions are getting made by the AI system and we don't need a human in the loop. But if you're getting to more important stuff like, "Should I transplant a lung into you?" Or, "Should I fire this drone?" I think we still really do need human in the loop.

In that case, it's more a tool and less a teammate.

Galen Low: Honestly, I like that because I think it's tr- it's been true of RACI all along. And, you know, going back to sort of like rudimentary RACI as a sort of project professional, you know, we're like, "Okay, this team doesn't normally work together. We just need some clarity around who's doing what."

And then the question I always get from folks that I'm, like, teaching this to is like, "Well, how deep do you go?" Right? Just like a risk register, like anything. I could put anything there. There's trillions of decisions. Maybe I'm over, over-exaggerating. There's a lot of decisions that happen within a project.

We can't write them all down. And I'm like, "Yeah, don't." Write down the ones that are consequential and a bit blurry or a bit ambiguous. And so, you know, y- y- your example with the, the sort of credit card fraud detection you know, most teams will be like, "Yeah, okay, we built the system that way."

I don't have to put, you know, who notifies the customer when we detect fraud and then put, you know, like the automation or the AI workflow in R. It's just, it's just part of the system. It's you know, it's not, it's not that different. But what I also like is that, you know, when we're talking about transformation, and it used to be, you know, Sugandha was the R, and now it's an agent, might be worth writing down to sort of like communicate the change in responsibility of okay, and well, now we can have a conversation about what Sugandha is doing, you know, and where she's an R or an A in the overall system, not even just like a single slice, but in the overall system.

I like it as that sort of like seeing where the chess pieces are moving, even though, you know, whatever, 10 months from now, maybe that won't even matter. But documenting that change so that we can have a conversation about what's shifting around and making sure that everyone's on the same page, and maybe having that conversation about "Ooh, isn't that kind of risky?"

Like we've just put, you know, our agent that we built. We've vibe coded this thing overnight, and now it's responsible for 80% of our revenue operations. Like it's like, okay, well, maybe this is worth talking about so that we can figure out, you know, is that okay? And if that's okay, you know, who's getting consulted or informed?

And importantly, who is authorized to make decisions around that work? Inquiring minds want to know, can AI ever be in the A column of a RACI, in your opinion? Can they be authorized or accountable at this stage in the technology's progression? It is making decisions. Is that something that we should be writing in?

And I guess why or why not?

Cassie Solomon: I think yes. So if you think about the, the top of the RACI chart is all your stakeholders that are gonna be involved, right? Including how many stakeholders do I have to consult so that I'm not getting into trouble, and how many people need to be informed, all that, that long tail of the RACI.

Absolutely, AI agent number one should be a part of that, that group so that you can have the conversation about, do we give it the first A and then we escalate if the following conditions apply? I mean, that's, that's a pretty low-risk way of defining its authority. If it's an, if it's an easy call.

For example, in I think it's Utah right now, they are piloting an AI system that can give you prescription renewals without a doctor, and this has been hugely controversial, and they've been really fighting over which drugs are so benign that if this AI gets it wrong and renews this drug, that it's gonna be terrible for the person.

And at the same time, kids who are in their 20s, and they're like, "What do you mean I have to go to the doctor to get a prescription? That's the stupidest thing I ever heard." And I'm like, "Wait for the Utah experiment," you know, which is very controversial. It's getting a lot of pushback from the medical society.

Is that a decision that we want the machine to make? And then what, what they've done to refine it is they've made the decision space narrower and narrower, right? Under these conditions for this kind of person, if they've had this drug three times in the past, then the AI is allowed to decide without review, right?

I love making it visible that way in a RACI so that you can have those conversations. The other thing you said that I loved, Galen, is when I meet people that hate RACI, it is always because they have sat through a session where they've had to RACI, like, 1,000 lines. And they have literally said to me, "I would rather watch paint dry than do another RACI session like that," and I don't blame them.

I'm, I'm gonna go in a slightly different direction than you did, though, because I don't think we have to RACI the entire process. Some of that process is pretty clear to us or intuitive. We just should RACI the pain points. Where is the process not clear? And sometimes I tell people, "Just RACI the decisions."

If there are trillions of them, that's another problem which we'd have to attack, but what if we just did a RACI that looked at a variety of decisions? So yes, AI should be on that chart, and that allows you to have the conversation about what are the parameters that we wanna give this tool so that it makes decisions that we think it's capable of making and reserves other decisions that we wanna reserve for the humans.

Galen Low: It's actually really interesting because, you know, in my head I'm, I'm, you know, going through risk. What is the risk of, you know, prescribing medication without a physician involved? And the answer is it depends, and it depends on what we're comfortable with, and what we're comfortable with is about the impact of the risk.

Right? So actually this is all like, it becomes y- r- the RACI becomes your document of your risk tolerance in a way, which to your point, is shifting. And that's why we need to have the mindset and have the dialogue because sometimes they change beneath our feet. You said it at the beginning, a lot of it is still invisible A, to each other, like humans to humans might not be documented, we just know how it works.

Definitely invisible to AI, who is like, "Tell me how you make decisions," and we're all like.

Cassie Solomon: And the most painful thing about cross-functional teamwork is that everybody comes to that team from their own department, and most of those departments wanna preserve some of their decision-making authority. How can the team make the decision if I have to take everything back to my boss?

How can I empower this team if I'm not getting the appropriate amount of legal review or IT resource review or risk tolerance review? And then those teams just swirl. They can't make decisions. They can't move forward. And generally someone will say, "Ah, we need RACI." But that's the breakdown that I see, because if you throw just the tool on top of that mess, it, it'll help, but often it'll help in the way that people will acknowledge how messed up it is.

But it's what, everything that comes after that, that is the skill of RACI. How do I deal with that? How do I negotiate a more streamlined system that can move?

Galen Low: I wondered if we can like return to the i- idea of RACI as a design tool for system level change. I was wondering if maybe we could just like step through it you know, where does someone start?

How is RACI a tool that helps drive the dialogue to sort of, you know, capture what's going on? How has it been used to decide what the new new is? And also, yeah, how do we get around the, the, the idea that what we uncover is usually like power politics, right? Like that's what I like about your model.

The A is not accountable, it's authorized. Accountable, I think of oh, whose, who, whose fault will it be when it goes wrong? Authorized is actually who can make the call? Which also is power. And like you said, groups within an organization are gonna wanna retain power. Can we like step through like maybe just like a thin slice example of okay, yeah, we can design using RACI.

It's more than just an acronym. It's more than just a matrix on a piece of paper. It's it's a way to dialogue and design something new so that we don't end up with, you know, the electric motor i- where the steam engine was.

Cassie Solomon: Yes, we can. Fun. I borrowed a really cool technique from a client, so we're using it now where there's like the workflow is on top and the RACI chart that corresponds to the workflow is underneath that.

And one of the things that fills in is that there's so many workflows diagrams in the world, and they're, they're lonely. They're not populated with people and roles and who's doing what. They're just kind of describing the, the what. But in, in my certification beta class last week, one of my participants said, "You know, I don't like the way things are working in my business right now.

I wanna use RACI to redesign the roles. How can I do that?" And we said "Oh, we have a template for that And if you think about it, you can just start and say "Here's all the work I want you to do. Those... Here are your Rs. And do I want you to make any decisions?" That's a great question. So if I know that, I can say, "Where, which kinds of decisions c- are, are gonna be baked into your role?

What kind of expertise do I want you to bring to your job and this project? That's where we put you down in the C column." And they went off to redesign the roles in their company through that lens of there's certain work that has to be done, and that's usually captured in a workflow somewhere. People are not asking that decision question nearly enough, and they need to 'cause that's where all the trouble lies.

When I really think about your question from a how do we design transformation, I kinda have to go back to the model in the book, because I think of RACI as combining a couple of those levers, decision-making and people, which stands in for what are you capable of, how much do you know, what's your expertise, all of those pieces combined with decision-making.

But there's also task, which is your workflow. There's also what's your technology. There's also how do you organize the work. It's an intriguing time, because we do not know what this future looks like, and we have to kind of acknowledge that and then really look around and watch for these weird signals.

Okay, you know, I can push the intelligence to the front line and allow that decision to be made at the level of an individual soldier or an individual doctor. What else is it gonna be like? How else is it gonna change? We didn't anticipate remote work. That sort of took a global pandemic, but that's been a very profound shift.

How can we imagine those shifts kind of in advance?

Galen Low: And I like that it went to skills and capabilities and experience, because I think that is how a lot of organizations at least aspire to redesign their org. You know, there's dialogue around, you know, job titles being probably one of the worst descriptors of what people do in an organization, right?

Sure, shove them into a box and we'll be like, "Okay, well, you know, every physician with this many years of experience is, they're all the same." So they are accountable for the decision versus, okay, well, we need to figure out what needs to be done, who is qualified to do that and has a skill set of it, instead of boxing them into a title.

Maybe they are, you know, like an R or an A across all these different things, and that's maybe a new role. I can see it working really well on a project team where we don't have to necessarily think about, you know, permanent org structure, but I could also see it as a worthwhile exercise to arrive at clarity in answering the question, to your point, when point solutions look like it's just gonna replace a bunch of humans, like, how can we have that dialogue of, like, where people's responsibilities shift in this new model?

But I can see... I, I just, I, I'm picturing it, and I like that workflow with the RACI underneath. I think that's so interesting because You know, on a workflow diagram, decisions are just diamonds.

Cassie Solomon: There's a really lovely case, so I just have to say it. Maybe this will be how we close. IKEA replaced a whole bunch of its humans with chatbots, and they were doing very simple, like answering customer questions.

But when they stepped back and they analyzed the questions that were coming in, what they realized is that a whole lot of people were actually asking for design advice, and the chatbot wasn't gonna be able to help them with that. And they had to retrain 8,500 people so that they could start giving some basic design advice, and they made a boatload of money.

So I think that we have to remain open to this possibility that if we're really willing to look at the system and do the redesign work, like the AI's gonna expose where the gaps are. We have more data and more intelligence at our fingertips than we ever had in the entire history of the human race.

What are we gonna do with that? Fire people? That's dumb. You know, we're gonna take all of that intelligence and that data and see where are the gaps. I always tell people, "Calm down because this technology will eat your job like a little Mario Brothers nibbler from the bottom up." Let's eat the scut work with the AI technology, and let's use the data that that generates to teach us, like IKEA learned, where's the gap that we can fill?

Galen Low: I love that. And that's, you know, we can see the gaps you know, through RACI, a RACI mindset, a mindset of understanding how decisions are made and where there are gaps as we create something absolutely net new that we have never envisaged before.

Cassie Solomon: I love net new. That's really good.

Galen Low: Speaking of net new, I mean, what, what does the future look like for you?

I mean, just in terms of decision-making, we've been talking about, you know, organizations that, yeah, can have, you know, a single decision-maker at a certain scale. We've been talking about, you know, central command making decisions, but that can be slow. We've been talking about decisions being pushed to the fringes so that w- you know, teams and organizations can be more nimble, and we've been talking about AI making decisions.

What does decision-making and collaboration look like in two, three years' time for you?

Cassie Solomon: You know, the point of the certification is to say basic level of RACI work you, you knew how to do. The demands have just become exponential on us to talk about and understand role and redesign role. So we just all need to become real ninjas with this tool and, and allow it to help us kind of illuminate these invisible parts of the organization that we know are either holding us back or that present opportunity for us to do something really novel and innovative.

If we can't speak that language of role like a master, we're, we're just gonna be stuck. I can't predict the future but I can tell you that the signals of what is coming are presenting themselves today. And if we can sort of tune our ear to "Oh, that looks like an example of where decision-making has shifted.

Oh, that looks like an example where decision-making has followed information to a new place, I think we can start constructing our understanding of what the new new looks like.

Galen Low: I really like that. Yeah, an abundance of information now, but a language that helps us to have the dialogue, you know, the ninjas of understanding decision-making and how stuff works so that we can actually have that conversation.

Cassie Solomon: It's cool. It's a terrifying and cool time to be trying to design the future with this incredible new technology.

Galen Low: Awesome. Cassie, this has been amazing. I love chatting with you. I've learned a whole ton. Thank you for coming on the show again and sharing your insights. For folks who wanna learn more about you and what you're doing with RACI and the RACI certification, where can they go?

Cassie Solomon: RACI Solutions is the website that we built around our RACI consulting practice. And eventually there will be something on it about the certification course. We're limiting the cohorts to 20 people 'cause we really want the dialogue. We want people learning from each other at this stage of the certification.

So if somebody's interested, I'd love for them to reach out.

Galen Low: Perfect. I will include your profile in the show notes as well. And Cassie, thanks again.

Cassie Solomon: Thanks, Galen.

Galen Low: All right, folks. That's it for this episode of Digital Project Manager Podcast. If you enjoyed this conversation, make sure to subscribe wherever you're listening.

And if you want even more tactical insights, case studies, and playbooks, create a free account with us at thedigitalprojectmanager.com. Until next time, thanks for listening