In this episode of Executive Connect, Melissa Aarskaug sits down with Kathleen Walch, leader at the Project Management Institute and co-creator of the CPMAI methodology, to talk about what actually makes AI succeed. Kathleen explains why AI failure is rarely just a tech problem and why leadership, change management, data quality, workflow design, and trust matter far more than most teams realize. She also shares what strong organizations do differently before they invest, how leaders can spot failure early, and why AI literacy has to become part of the culture.

If your company is experimenting with AI, rolling out copilots, or trying to move from hype to business value, this episode will help you approach it with more clarity and fewer expensive mistakes.

Chapters
(0:00) Why leaders must stop waiting
(0:29) Kathleen Walch on AI execution
(1:23) Why AI is change management
(2:31) What leaders underestimate most
(6:58) Why AI is not just IT
(9:27) The top reasons AI projects fail
(11:57) Early warning signs before failure
(14:11) Who owns AI outcomes
(15:44) How leaders quietly block progress
(18:17) Where hierarchy slows AI learning
(21:51) What successful organizations do first
(25:47) Why so many AI projects fail
(29:02) How to build trustworthy AI
(32:25) What leaders should start now
(37:42) The AI myth leaders must drop
(43:16) Final advice and where to connect

Kathleen

(0:00) I mean, this is hard because I’m like, there’s so many myths when it comes to AI, but I would say really let go of feeling that you need to have everything perfect, right? (0:09) I say perfect is the enemy of good. (0:10) Just get started.

(0:12) Don’t put unnecessary roadblocks in your way. (0:16) So don’t wait for that perfection. (0:18) Just get started.

(0:19) Just experiment. (0:20) Learn fast. (0:21) Iterate.

(0:22) It’s okay. (0:23) Come from a world of experimentation rather than a world of debating and never do.

Melissa

(0:29) Kathleen Walsh is one of the clearest voices in AI execution today. (0:35) As a leader at the Project Management Institute and co-creator of the CPMAI methodology, she’s helped organizations turn AI from pilots and hype into real business results. (0:51) Kathleen works directly with executives to fix what actually causes AI to fail, not the technology, but leadership, decision-making, and operating models.

(1:04) Welcome, Kathleen.

Kathleen

(1:06) Thank you so much. (1:08) I’m so excited to be here.

Melissa

(1:10) Excited to have you. (1:12) Now, most organizations rush into AI without changing the decision-making. (1:19) Where’s their break in that choice?

Kathleen

(1:23) Yeah. (1:23) It’s interesting. (1:24) So I have been in this space, covering this space for about a decade now.

(1:30) And so I’ve really gotten to see a lot of different trends. (1:33) And this was, I always say, I’ve been involved with AI before Gen AI made it hot, right? (1:38) Before large language models, before open AI really kind of hit the scene and Claude and Gemini and all of these things.

(1:45) So what I really see is that leaders underestimate some of that human side, far more than the technical side, and that AI really is about change management. (1:56) And so when we talk about AI from that kind of leadership perspective, AI and change management really go hand in hand.

Melissa

(2:04) I love it. (2:04) Now you sit at a very rare intersection of AI strategy, execution, and leadership capability. (2:13) Through your work at PMI and CPM AI, now that’s a tongue twister, you see to hundreds of initiatives across the industry.

(2:23) From your viewpoint, what do leaders consistently underestimate when they decide to quote unquote, do AI?

Kathleen

(2:31) Yeah. (2:31) That’s really interesting too. (2:33) So CPM AI for listeners that maybe don’t know about it, it is a methodology for running and managing AI projects.

(2:40) And I co-developed it back in 2018. (2:44) So a little bit of history about kind of how it got to PMI as well, before I get into the question that you asked. (2:51) So CPM AI was developed from Cognolitica, which was an AI focused research advisory and education firm that me and my business partner founded back in 2017.

(3:02) So we were a boutique analyst firm and had some clients, right? (3:07) I’m from the DC area. (3:10) And so we had some government agencies, we had some large banks that we were working with and they said, okay, this is great.

(3:17) You’re covering this space, but we want to run an AI project. (3:21) So can you help us get started? (3:23) And we go, there clearly must be a methodology out there, right?

(3:26) And we looked and there was no methodology specifically designed for running and managing AI projects. (3:31) So that’s how CPM AI was developed. (3:33) And so what we did is we looked to see what was out there and we saw that there was CrispDM, which was really around data mining.

(3:42) It was built in 1999. (3:44) The Agile Manifesto was published in 2001. (3:47) So it was not agile and iterative, right?

(3:50) How software development was really starting to get done. (3:53) And then what we also realized is that AI projects are about data. (3:57) They’re really data centric.

(3:58) So you need a data centric methodology. (4:00) That’s why we took the data mining, you know, as a foundation, but then brought some of those agile best practices into there to make this iterative. (4:09) You run this in short iterative sprints.

(4:12) So that’s the kind of, you know, backstory of CPM AI. (4:17) PMI approached us back in 2024. (4:21) They are focused on eight content and innovation pillars.

(4:24) AI is one of them. (4:26) And they said, you know, we can do two things. (4:29) We can acquire what’s already out in the market or we can build it.

(4:32) We had momentum and traction. (4:34) It was a really great synergy because it was a methodology for running and managing projects. (4:38) Project Management Institute is about project managers.

(4:41) So we would get a lot of project managers who would sign up for this. (4:45) And then we also get project adjacent folks. (4:47) So, you know, PMI, we call them project professionals.

(4:50) So they may not have that title of project manager, but they’re tasked with running and managing projects. (4:54) When it comes to AI projects, what does that look like? (4:56) It could be a data engineer, it could be a data scientist, it could be someone in the IT department.

(5:02) So the, you know, the marriage was really great. (5:05) And I’m excited to share, now we’ve been there for about a year and a half and we are the fastest growing certification at PMI. (5:11) Third behind only our two core PMP and CAPM certifications.

(5:16) So that’s exciting. (5:17) And I think that kind of helps give a framing for the discussion. (5:21) And then you think about, okay, well, what did the world look like back in 2018?

(5:25) It was much different than what it looks like today. (5:29) So when we think about, you know, why leaders are maybe underestimating what they want to do with AI, there’s a few things that I see. (5:39) So one, they underestimate the human side far more than the technical side, right?

(5:43) Because I talked about how, you know, it’s really data centric. (5:46) There’s a lot of change management that comes around. (5:49) So I think about people, process and technology, it’s usually never a technology issue, but that’s the easiest thing to quote unquote fix or throw money at, but it’s the people in process side that you really need to think about.

(6:02) And that’s what CPMI helps with. (6:04) And then they also underestimate some of that process debt that you have at an organization. (6:09) So back in 2018, the world looked different.

(6:12) RPA, robotic process automation was what everybody was talking about. (6:16) And how do you take different processes, right? (6:18) And automate it so that you’re not manually doing it anymore.

(6:22) But if you have a bad process, automating it doesn’t make it a better process. (6:27) The same thing happens with AI. (6:29) It’s just amplified, right?

(6:31) Your bad processes are amplified. (6:32) Your good processes are amplified. (6:34) And I have found that if you don’t get into the core of those processes and actually re-imagining the workflows, you’re still going to run into issues as well.

Melissa

(6:43) Ready to lead smarter and invest wiser? (6:47) On the Executive Connect podcast, we unpack executive strategies for wealth and influence. (6:54) Hit the subscribe button now.

(6:56) Don’t just watch, act. (6:58) Yeah. (6:59) And it’s so true because I think there’s a difference between what leaders say they want from AI and what their organizations are actually set up to deliver.

(7:12) And I find that there’s a very large disconnect because it used to be that this fell under IT and now it’s quietly kind of becoming a leadership mandate. (7:23) So talk to me a little bit about that from your lens.

Kathleen

(7:30) Yeah. (7:30) So I think, you know, sometimes people think, yes, this is an IT issue, but we say, no, this is an everyone issue, right? (7:37) So with CPM AI, there’s six phases.

(7:40) And so it’s supposed to be iterative. (7:41) But the first phase is business understanding, figuring out what problem you’re trying to solve and make sure that it’s a real business problem. (7:49) Far too often people just jump right in and they go, we want to do AI, which I don’t even know what doing AI means.

(7:56) Right. (7:57) And then it’s like, well, why do you want to do it? (7:59) Oh, because we have FOMO, fear of missing out because our competitors are doing it because we think we’re lagging in the market.

(8:04) Those are reasons that you want to do AI, but those are not why like what business problem you’re trying to solve. (8:11) Because I always say AI is an amazing technology, right? (8:15) There’s so many different applications and ways that you can use it, but it does not solve every single problem.

(8:21) Sometimes you can have automation. (8:23) You know, I talked earlier about RPA. (8:25) You don’t have to have intelligent automation, or you can just have humans do it, or you can code your way to a solution.

(8:31) AI doesn’t always have to be the answer to every single problem. (8:35) So if you don’t actually understand what problem you’re trying to solve, then you’re going to run into issues. (8:42) And what I say is that this costs money, time and resources.

(8:47) And so you have to think about what that ROI, that return on investment is going to be for your AI investment, because if it’s a negative return, then people are going to say, that was great. (8:57) We’re never doing that again. (8:58) And that doesn’t actually help the organization.

Melissa

(9:04) Yeah, I agree. (9:05) Now, are there specific patterns that repeat across various industries, whether they’re a mature company or a startup? (9:17) What patterns are you seeing repeat in these cases?

Kathleen

(9:21) So do you mean what patterns I’m seeing why AI projects fail? (9:25) Exactly. (9:26) Yeah.

(9:27) So actually I give an entire talk on this, right? (9:29) And I present common reasons why AI projects fail, and I’ve identified 10 common reasons. (9:34) And it doesn’t matter what industry you’re in.

(9:36) It doesn’t matter what you’re trying to do. (9:38) These are common reasons. (9:40) So the first one is treating your AI project like a software development project, right?

(9:46) I said earlier, it’s all about the data. (9:49) So when you think about software development, the most important part of that is the code, right? (9:54) You would never give away your code.

(9:56) That’s the most important part. (9:57) But when it comes to an AI project, your most important part is the data. (10:01) You would never give away that data, right?

(10:03) That’s what you use to train these systems and get better over time. (10:07) And so that’s a shift. (10:08) And you have to say, okay, now the data is most important.

(10:11) So what does that mean? (10:12) Well, now we need to make sure that our data, right? (10:14) That leads into the next common reason why AI projects fail, which is around data quality and data quantity.

(10:19) How much data do you have? (10:21) We say you don’t need Google sized amounts of data, but you do need to have a nice representative sample of data for whatever it is you’re trying to do because we learn over that, right? (10:31) So we don’t want to train it on just a few data points and then find out that that’s not representative of the whole and then data quality, right?

(10:39) Garbage in is garbage out. (10:40) Absolutely true when it comes to AI. (10:43) Other reasons that we’ve seen is the return on investment isn’t justified, that AI is not set it and forget it.

(10:50) So you can’t just think that you can start your project and say, okay, we’re done with one iteration. (10:56) We’re done with our AI initiative. (10:57) It’s like, no, you can’t do that.

(10:59) You have to continually train, retrain these models, monitor them. (11:04) And so that also goes into resource allocation. (11:07) And you have to understand you might have more resources need upfront, but you’re never going to not need resources on these projects.

(11:15) So think about that from a budgeting standpoint. (11:18) And then the other, the biggest, biggest reason that we see is over promising and under delivering on what the technology can do. (11:26) So people get very wide eyed, get very excited about what you can do, think that it can do all these different things.

(11:32) And then only to realize maybe it’s taking a little longer than we realized. (11:35) Maybe we’re not mature enough to do that yet. (11:38) And that’s where a lot of initiatives have failed over the decades that AI has been around.

Melissa

(11:45) Now, are there early signals that leaders can spot when AI initiatives are going to fail or almost failing? (11:53) Are there any kind of symptoms to that, that one would look for?

Kathleen

(11:57) Yeah. (11:58) So what I always say is when we’re figuring out what problem we’re trying to solve, right? (12:02) Start with phase one business understanding.

(12:04) Within phase one, there are a series of questions that we ask. (12:09) We call it the AI go, no go. (12:12) And so it’s around business feasibility, data feasibility, and implementation feasibility.

(12:17) So with business feasibility, it’s, is there a clear problem definition, right? (12:22) What problem are we solving? (12:24) And then the second question that you have to ask is, is the organization willing to invest and change?

(12:30) And that’s really important to answer because if the folks involve stakeholders, and if they are not willing to actually invest in change, you’ve just spent all of this money, time, and resources to build something that people won’t use. (12:43) So these are some of the early signs. (12:45) And then we have, is there sufficient ROI or impact?

(12:48) You have to be measuring that, right? (12:49) You have to measure what kind of impact you’re getting. (12:52) And then when it comes to data feasibility, make sure that you have that data that you need.

(12:56) And then implementation feasibility, do you have the required technology and skill sets internally? (13:02) If you don’t, you have to look externally, make sure you’re understanding that. (13:06) Can you execute the model as required in a timely manner?

(13:09) And then does it make sense to have the model where you live? (13:12) So an example I bring up there is with phones, right? (13:14) You can unlock your phone with your face.

(13:17) If you needed to always be connected to the internet, I wouldn’t be able to use it if I’m out hiking and I have bad cell reception, or if I’m in my kid’s school, which seems to also never have good cell reception, or if I’m on an airplane, right? (13:29) And I’m in airplane mode. (13:31) So you have to say, where is this going to be used?

(13:33) And do we have the ability to use it in that desired way? (13:38) If you’re not addressing these upfront on a project, those are like, you know, we think about this as a traffic light. (13:45) So green means you’re good to go.

(13:47) Yellow is proceed with caution. (13:49) And red is let’s stop for a minute. (13:51) If you have a lot of reds, the chances of your project failing increase significantly.

(13:57) And so these are some of those early indicators that you could have addressed this back in phase one, back maybe before you started the project. (14:04) But if you’re so excited and you just want to move forward and skip all of this, that’s usually how it can snowball.

Melissa

(14:11) Now, since you mentioned ROI, who owns the outcomes when AI decisions affect not just revenue, but risk and a company’s reputation?

Kathleen

(14:23) Yeah, so when you think about, right, who owns AI at the organization, I say it’s everybody’s responsibility, because everybody’s going to be using these tools. (14:32) And so again, think about it, right? (14:34) So I think about at PMI, we have different groups.

(14:37) And so marketing uses it, maybe they use some tools specifically for marketing and finance uses some finance specific tools. (14:45) But overall, if you have, you know, an AI transformation going on at your organization, you’re all responsible. (14:51) And one thing that we do is we have a trustworthy AI framework.

(14:55) And we have a series of questions that we’re supposed to ask for, you know, to make sure that we are addressing them. (15:01) And then we also think about AI governance. (15:04) So if you have an AI governance committee, you should be having representation from every single group.

(15:09) And if you don’t, that’s a problem.

Melissa

(15:12) Yeah, and a lot of leaders believe that they’re progressive with AI, I feel like, but not necessarily, maybe unnecessarily blocking the success of projects, maybe they’re taking too long in certain phases, or, you know, their proof of concept isn’t moving forward. (15:32) So where do you see that most clearly, where leaders are being progressive, but actually also blocking their project from moving forward?

Kathleen

(15:44) Yeah, you know, it’s interesting, because as I talked to so many different leaders, right, or, and then folks at the organization as well. (15:51) So leaders have one perspective, but then the actual doers have a different perspective. (15:55) And so organizations right now are rolling out different AI tools.

(16:01) And they will give either very basic training to their employees, or have them, you know, read an article, watch a small video, do a bit of training, and then they go, okay, great, we’ve trained our employees, now go out and use it. (16:18) And so when I think about how that how you roll out technology, that usually never works. (16:24) And guess what, it’s not working at organizations, because they need ongoing training and ongoing support, and this ongoing like AI literacy program at the organization.

(16:33) So I’ve seen there’s a big disconnect. (16:36) And folks also need to have examples of how others are using it, and identified, we’re (16:42) calling them change champions internally, so that they are really understanding the tools, taking (16:48) it upon themselves, because there’s some people that just love playing around with different (16:52) technology, learning it, right, they’re really early adopters, they love doing this, identify them (16:57) at the organization, and then have them be a resource for others to go to.

(17:01) That’s usually how you’re going to see success, right. (17:03) And I also say, create this safe space to fail, let people experiment. (17:07) And one thing that leaders really need to be doing is using AI out in the open, let people see, this is how I’m using it, here’s some of the struggles that I’ve faced, don’t worry if you’re facing that as well.

(17:18) And also know that you might be slower before you’re faster, right, it’s that learning curve, and that’s okay.

Melissa

(17:24) This episode is brought to you by Summit Ventures. (17:29) If you’re an accredited investor, Summit gives you access to one of the greatest tax advantage opportunities, direct ownership in oil and gas. (17:41) Their projects deliver what they call the triple play, cash flow, equity growth, and powerful tax benefits.

(17:49) And here’s the best part, these investments qualify for 1031 exchanges. (17:56) That means you can roll gains from real estate into energy while deferring capital gains. (18:02) To learn more and get a free white paper, oil and gas demystified, just visit www.summitven.com forward slash executive connect. (18:17) And I love that. (18:18) I think there’s so much truth to that. (18:20) I think talking to your peers and getting insights, you know, working inside an organization, we tend to get kind of rose colored glasses and laser focused on our thing, but getting, like you said, communicating with others and being open, it offers more of an independent perspective into what we’re working on and maybe things we don’t see internally.

(18:45) I’m curious to get your perspective on where does hierarchy slow AI learning the most?

Kathleen

(18:54) You know, that can be interesting too. (18:57) Sometimes there can be a disconnect from the leadership team and then the actual implementers. (19:02) And so what I’ve seen at organizations is when the leaders say, we are going to embrace AI technologies.

(19:09) We want you to use this, but what we are not going to do is bring these tools on to replace you. (19:16) We are going to use these tools to help you. (19:19) And I call this augmented intelligence, where you’re not replacing the human, but helping them do their job better.

(19:24) There’s a lot of fears and concerns when it comes to AI. (19:27) And so if people feel at all threatened, they will back off and they will say, what, we’re not going to use this technology. (19:35) And we don’t trust it.

(19:37) Right. (19:37) And we don’t understand it. (19:39) So leaders really need to say, we’re on this journey together.

(19:44) We are not going to replace you, but we’re going to help you do your job better. (19:47) Yes. (19:47) There might be some tasks that you do that are replaced, but that’s probably a good thing because you shouldn’t be doing some of these tasks.

(19:54) So from the project management lens, we say, well, what are some of the tasks that project managers always have to do? (19:59) Meeting minutes. (20:00) Well, AI is really great at that.

(20:02) So now you can be an engaged participant in the meeting rather than worrying about taking meeting minutes. (20:07) It can summarize it at the end. (20:08) It can create action items and follow-ups.

(20:11) It can do a lot of really great things that now free up your time so that you can be a more engaged participant. (20:17) You can not have to spend the time after the meeting doing those meeting minutes. (20:23) Maybe you just spend five minutes reviewing it rather than 30 minutes creating it and sending it out.

(20:28) So you have to think about it from that perspective.

Melissa

(20:32) And I agree. (20:32) I was just going to say, I think you’re spot on with the human perspective. (20:36) We get in our head like, oh my gosh, this is going to take my job, or I’m not going to have a job, or this is going to be outsourced to somebody else.

(20:44) And I find when we communicate with people on even something simple, like when I got a co-pilot license for the first time from an employer I worked for, I didn’t use the license for six months. (20:57) I didn’t even know it was enabled. (20:59) I had the license.

(21:00) The company paid it. (21:01) Nobody told me I had it. (21:03) Nobody told me to use it, or I was allowed to use it.

(21:06) So it sat for six months unused. (21:09) And I think about all the productivity that could have happened in those six months. (21:13) Somebody told me, hey, here’s a license.

(21:16) Hey, here’s how to use it. (21:18) Here’s how others are using it. (21:19) Or just that simple thing.

(21:21) So in my example, it was costing the company money. (21:25) I was less productive. (21:27) And had somebody just told me, like you had mentioned, or communicated with me, I would have been happy to leverage it for meeting minutes or in any of my admin type tasks at my job.

(21:38) And so you’re so spot on with the human piece. (21:41) And so I want to flip the script a little bit and talk about what successful organizations do differently before they’re deploying AI.

Kathleen

(21:51) So when I think about that, I always say, again, figure out what problem we’re trying to solve. (21:57) And we want to think about this not necessarily as tool first, but problem first. (22:04) So I say, when you’re figuring out what problem you’re trying to solve, then you can figure out if AI is the right solution to that problem or not.

(22:11) So that’s what I’ve seen. (22:12) I’ve also seen executive leadership invest in this and say, this is a priority. (22:20) And then provide that AI.

(22:22) You can call it AI literacy, AI upskilling, AI training. (22:26) People call it different things. (22:27) Some people argue, well, I don’t like the upskilling word, or I don’t like the training word.

(22:31) So I say, call it whatever you want. (22:32) But the idea is what you really should be caring about. (22:35) And giving your employees that knowledge and meeting them where they’re at.

(22:41) I think that that’s also really important. (22:43) Because the example that you gave is, unfortunately, far too often the reality of many organizations. (22:49) And that it just appears one day.

(22:51) And you’re like, what is this? (22:53) And there’s so many different tools, especially even when it comes to Copilot, for example. (22:57) Because we’re a Microsoft shop.

(22:59) There’s Copilot Chat. (23:00) There’s Copilot Studio, where you can build agents. (23:03) There’s Copilot M365.

(23:05) All of them are different. (23:07) And they all have different capabilities. (23:09) But if you don’t even let people know that they’re enabled, then people will reach out to me and they’re like, what is this?

(23:14) What does this mean? (23:15) What data can I put into this? (23:16) Is this safe?

(23:17) Is this secure? (23:18) Can I be putting this data in? (23:20) And I’m like, this is where an AI literacy program really is important.

(23:24) And kind of that lifelong, we talk about lifelong learning. (23:28) You have kids. (23:28) I have kids.

(23:29) Growth mindset is the new term that they use. (23:32) But how do you help people? (23:34) And understanding that you are going to have a wide range.

(23:37) I always say AI is a journey. (23:38) And you’re going to have a wide range of people at different points of their journey. (23:43) And so, what does that look like?

(23:45) And so, one thing that I’m really excited that how we’re rolling out AI is we’re following all of these best practices, right, that I’ve been talking about for years. (23:52) And I’m saying, okay, great. (23:53) We’re figuring out what problem we’re trying to solve.

(23:55) We’re using the right tools for the right problem. (23:58) And we are helping to build an AI training and literacy program so that we’re not just throwing a tool at someone and saying, go figure out how to use it. (24:08) We’re going to help.

(24:09) We’re just starting on that. (24:12) So, I’m really excited about that. (24:13) And so, we’re saying, we’re going to have office hours.

(24:16) We’re going to have training. (24:17) We’re going to have and training will be two flavors, right? (24:19) Training will be online kind of self-paced training.

(24:22) And then also in-person for that hands-on training, we’re identifying change champions so that they can help empower individuals and be that resource that people can go to. (24:31) Because sometimes it’s just a small thing that you’re stuck on. (24:34) And if you have to figure it out, it can take hours.

Melissa

(24:38) And I love that. (24:39) And I think so spot on with the training, I think about like my example. (24:44) So, when I found out that I had Copilot was the first thing I did, I reached out to someone to talk to me about how’s the best way they’re using it.

(24:50) So, not only did I step away from my job, I also distracted another person. (24:56) And it’s a snowball effect if you’re a large organization, if everybody’s calling to figure out how to use it, when you could just either record, deliver, or have these little trainings that people can click on to solve a problem versus having your key people that should be working on these AI initiatives, focused on the AI initiatives versus helping everybody configure their Copilot license. (25:23) So, I think to your point, training is so key and communication when we are moving forward.

(25:32) I want to get your opinion on some of these. (25:37) I heard some number recently and it’s 80 something percent. (25:42) And I don’t know if this is true or the person that said it, where they got this number from, but I’m curious to get your perspective.

(25:47) 80 something percent of AI initiatives fail within side of an organization. (25:55) I think of all the money spent to move these initiatives forward and then just to stop them. (26:03) So, is that a true number?

(26:06) What is the percent of these that are actually successful versus not? (26:10) And I know we’ve talked about why, but I’m curious to get your thoughts on that.

Kathleen

(26:16) Yeah. (26:16) Actually, I quote that number all the time. (26:18) 80 plus percent of AI projects fail.

(26:21) And so, what does fail mean? (26:23) Well, it fails to deliver value. (26:25) It fails to return a positive ROI.

(26:30) And so, then those are the common reasons why AI projects fail, right? (26:34) It fails because you’re treating it like an application development project, or it fails because your data isn’t of the right quality, or you don’t have enough of that data, or it fails because maybe you are using the wrong tool, right? (26:48) We say, don’t fall for vendor hype.

(26:50) Vendors are happy to sell you anything and everything. (26:53) And sometimes it’s not the right solution. (26:55) So, I always say, pause, figure out what problem you’re trying to solve, and then make sure that you’re actually solving that problem.

(27:03) Also, another reason that I didn’t share earlier was, you can fall into this uncanny valley of data. (27:10) So, uncanny valleys usually are around anthropomorphizing objects, right? (27:15) So, we think about robots, for example, and you try and make it more human-like.

(27:19) And then you’re like, wow, this is kind of creepy. (27:21) You fall into the uncanny valley where you feel uncomfortable with it. (27:24) And then you go back to the human and you’re like, okay, I feel comfortable.

(27:26) Well, that can happen with technology too. (27:29) And the more data that we have, the more personalized offerings that we provide, some people think that’s great. (27:35) And then you get to a point where you’re like, you know too much about me and I don’t want to use this anymore.

(27:40) And then you fall into that uncanny valley. (27:42) Once you’re in that uncanny valley, it’s really hard, if not impossible, to get out of it. (27:46) And everybody’s line is unfortunately different.

(27:50) So, you can’t go, okay, well, I’m going to go right up to the line and then not go into it. (27:54) Because maybe they are in my uncanny valley, but for you, you really enjoy how personalized it is when you think about personalized offerings and shopping. (28:04) But then imagine that at an organization, right?

(28:08) And it’s like, well, how personal do I want them to know? (28:10) How much do I want them to know about this? (28:13) I was thinking about taking a day off.

(28:15) And then the system goes, would you like me to book that for you? (28:17) Some people are like, yeah, that’s so great. (28:19) Thank you.

(28:19) And other people are like, I can’t believe it. (28:20) Now I don’t want to touch the technology at all.

Melissa

(28:23) So true. (28:25) Money Ripples is on a mission to help professionals like you get their money working harder. (28:32) Their clients free up an average of $35,000 their first year without having to work extra hours.

(28:40) To show you how, they’ve put together a powerful training called Cashflow Secrets. (28:45) And as a listener of the Executive Connect podcast, you can get it completely free. (28:52) Just visit moneyripples.com forward slash secrets and enter the promo code E X E C. (29:02) Now to your point, responsible AI matters, but we’re seeing it used as a reason to slow or stop their progress. (29:12) Exactly those examples you just explained, some are like, oh my gosh, it knows too much. (29:17) And others like me, I’m like, yes, give me a solution.

(29:20) So how do leaders get it right for all the different types of personalities either stop or are excited about it?

Kathleen

(29:29) Yeah. (29:29) So when I think about responsible AI, a few years ago, it was, can I do AI, right? (29:35) And it was like, what, just, can I do it?

(29:38) Is it going to be providing any sort of benefit to my organization? (29:41) And the answer was yes. (29:42) So then people said, okay, how can I do it in a responsible, trustworthy, ethical way?

(29:48) And it’s important to understand you can do that. (29:51) And that’s what organizations are doing. (29:53) So they’re saying, well, we’re building this with responsible AI at the forefront.

(29:57) And so when I think about our layers of trustworthy AI, we have five and it’s important to address all of them. (30:03) So the first one is ethical, right from wrong. (30:05) Then the second one is responsible just because you can do it, should you do it.

(30:10) And then the third layer, which is what I really focus on, especially for leaders and organizations is transparent AI. (30:17) And this isn’t about how transparent your algorithms are. (30:20) This is about how transparent you are with different things.

(30:24) So I think about, so it’s, are you disclosing that your end user, your customer, your employees are talking to an AI enabled chatbot versus a human? (30:34) Are you disclosing what data you’re using to train the systems and how their data that they’re giving will be used to train these systems? (30:42) And this is so important because this is really what builds that trust.

(30:45) So I think about, it’s really frustrating when I go onto a banking website and there’s a chatbot that comes up and I don’t know if it’s a human or an AI bot and it doesn’t disclose that. (30:55) And it really frustrates me. (30:57) And every time I say this, everybody shakes their head and nods and laughs because that’s exactly the same experience that they get.

(31:04) And it’s frustrating. (31:06) And if they disclosed that it was an AI chatbot, I’d have a much, maybe I wouldn’t have a much different conversation, but I’d have different expectations. (31:13) Not better, not worse, just different expectations.

(31:16) Not disclosing it frustrates me and it makes me not necessarily want to use that tool. (31:21) That’s why it’s so important. (31:23) Then we have our governed AI layer.

(31:25) So this is around those policies, rules, regulations, procedures. (31:28) And then we have our explainable AI layer. (31:31) So whenever possible, try and pick algorithms that are more explainable than others.

(31:35) But it’s really that transparent layer that’s so important. (31:38) And that’s what builds that trust.

Melissa

(31:42) Yeah. (31:42) I love that you said that because I use chatbots a lot and I have a unique last name. (31:49) So I can’t even get past my name a lot of times because the functionality isn’t there yet, especially medically when I have to talk into it versus type into the tool.

(32:03) And so I agree with you. (32:05) I think if we were just clear, my expectations would be much lower with the AI bot. (32:12) So when we look ahead, let’s say 24 to 36 months, which is like a lifetime in where we are right now, what’s about to change for leaders, whether they’re ready or not?

Kathleen

(32:25) Yeah. (32:25) When it comes to AI, I’m always like predicting six months out or looking forward six months is a lot. (32:31) So 12, 24, 36 is even longer.

(32:34) But what I say is for executives that maybe don’t have an AI-first organization, they don’t have these enterprise-wide AI transformations and they’re thinking about this, I say make sure that you get started with building AI literacy across your teams and also making sure that you’re cleaning your data and getting that prepared. (32:57) Even internally, we’ve been doing that. (33:00) And it’s so critically important because as I said earlier, that garbage in is garbage out.

(33:06) So make sure that your data is really foundational. (33:09) And then with CPM AI, our big motto is think big, start small, and iterate often. (33:18) So think about the big picture, the big goals that you want, and then start small, start with maybe a small pilot.

(33:23) Make sure that you’re getting it underway, that folks feel comfortable with this. (33:26) And I say do a pilot and not a proof of concept because a pilot is going to be using real-world data, real-world examples, and people that are not so closely tied to building that project or that AI tool or that AI example that you have. (33:42) And why is that so important?

(33:44) Because if you built it, you know how it should be used and you tested exactly how you want it to be used. (33:49) And in the real world, not everybody uses it that way. (33:52) So that’s why I say make sure you start with those pilots and really get that under your belt.

(33:57) And then because those organizations I see that are investing in that upskilling, that experimenting, and really upgrading and grounding and, you know, spending time cleaning their data are the ones that are going to be succeeding when it comes to AI.

Melissa

(34:17) So many questions from that. (34:18) So a year, six months to a year from now, what will executives wish they had started doing today?

Kathleen

(34:27) I mean, I think it goes into, right, AI literacy and upskilling, really thinking about what you need for your organization, figuring out workflows, right? (34:39) So when we think about AI projects and AI initiatives, we have to also think about what that workflow is, right? (34:47) So everybody talks about agents now, but like what does it really look like to have an agentic workflow?

(34:54) Don’t just take the workflows of today and make them automated. (34:58) Reimagine the workflows. (35:00) That’s hard.

(35:00) That’s process reengineering, right? (35:02) I mean, and you have to really think about that and say, do I need this? (35:06) Are these steps necessary?

(35:08) But it’s so critically important for success.

Melissa

(35:11) Now, which leadership roles will quietly disappear?

Kathleen

(35:18) Wow. (35:19) You know, I don’t know. (35:21) I don’t know if leadership roles, like leadership, maybe when we’re thinking C-suite will quietly disappear, but I know that there’ll probably be ones that are created.

(35:31) For better or worse, a lot of organizations are having chief AI officers. (35:35) I say you shouldn’t have a chief AI officer because if AI is so important to all parts of the organization, why would you have that? (35:43) Like we didn’t have a chief mobile officer.

(35:45) We didn’t have a chief internet officer. (35:47) Why would we have a chief AI officer? (35:49) It should fall into one of the other, the other, you know, large buckets or just across the entire organization.

(35:56) But I do see additional roles, you know, either being elevated or created.

Melissa

(36:03) I like that. (36:04) Now there’s a lot of people that are sitting and waiting and watching to see what happens in our world of AI right now. (36:13) Even some of these companies are large companies.

(36:15) And so tell me a little bit about what is the cost of waiting that most boards are underestimating?

Kathleen

(36:26) Yeah. (36:26) I mean, you know, that’s hard, but I say get started now, or you’re just going to continue to be farther behind. (36:33) And so it’s hard to measure how far behind you’re going to be, but organizations that are AI-enabled will far outshine organizations that aren’t.

(36:42) And what I’ve even been seeing is that large organizations have a lot of debt, right? (36:47) A lot of people debt, technical debt, data debt, and baggage, if you want to call it. (36:53) For better or worse, I get it.

(36:55) That’s how large organizations function. (36:57) But these small, nimble organizations who have small teams, who can be AI-first, AI-enabled, AI-forward, whatever it is the term that you want to use, will far outperform some of these large organizations that get bogged down in their processes and bottlenecks. (37:15) So I would say, reimagine what you can with those different workflows and your ways of working so that you can stay relevant.

Melissa

(37:23) I love it. (37:23) That’s such a great answer. (37:26) All right.

(37:27) We’ve covered so much ground today. (37:30) I want to do just kind of a quick, fast lightning round and get your perspective on some questions I’m wondering today. (37:37) So one AI myth leaders should let go today.

Kathleen

(37:42) You know, I have to say, I mean, this is hard because I’m like, there’s so many myths when it comes to AI, but I would say, really let go of feeling that you need to have everything perfect, right? (37:53) I say perfect is the enemy of good. (37:55) Just get started.

(37:57) Don’t put unnecessary roadblocks in your way. (38:01) So don’t wait for that perfection. (38:04) Just get started.

(38:05) Just experiment, learn fast, iterate. (38:08) It’s okay. (38:10) Come from a world of experimentation rather than a world of debating and never doing.

Melissa

(38:17) One behavior every executive must adopt.

Kathleen

(38:22) I think that organizations that are going to be the most successful are really embracing AI, giving employees that safe space to fail, to experiment, and then also challenge your executive team to start working in the open with AI. (38:38) So don’t delegate it. (38:39) Don’t theorize about this.

(38:41) Don’t just read articles and push that out and do very asynchronous training, but actually use it in your own workflows and start talking about this and learning as a team.

Melissa

(38:53) One question to ask before approving AI spend.

Kathleen

(38:58) So when it comes to AI projects, I always say, figure out what business problem you’re trying to solve. (39:05) So don’t ask about what workflows or what tools or what you should be doing there, but really figuring out what problem it is that you’re trying to solve. (39:16) And that will make sure that you’re saying, do I need to use AI?

(39:21) Is this an AI project or should this be something else? (39:25) And when we think about spend, and then also the other thing that’s so critically important is thinking about that ROI. (39:31) What is the return that we want?

(39:33) A lot of times people aren’t measuring that return. (39:35) So you don’t even know how to, like, if you don’t know what return you want, you don’t know how to measure it.

Melissa

(39:42) One KPI that lies about AI success.

Kathleen

(39:48) That lies?

Melissa

(39:49) Yeah. (39:49) So yeah, one statistic maybe is a better, we set these KPIs when we are doing AI projects, one that lies. (40:02) It tells you you’re successful that you’re not actually successful.

Kathleen

(40:07) Okay. (40:10) That’s interesting. (40:11) So what we say at PMI, right, is we think about project success is greater than project management success.

(40:20) And so when you’re thinking about AI, just because you’ve built it doesn’t mean that it’s successful. (40:26) What is the adoption of it? (40:29) And if your folks are not adopting it, that can be, so maybe you want to meet these deadlines, meet these goals, but if folks are not actually using it, that’s not successful.

Melissa

(40:42) Ethics should live where inside the organization.

Kathleen

(40:46) Wow. (40:46) I mean, I say if everybody’s using AI, everybody needs to be involved in those ethics discussions and that you have to, you know, when I was talking about the AI governance committee, you have to have folks from all parts of the organization. (40:59) And then what I say is the outputs of this committee should not be closed to just the committee.

(41:05) They should be widely known, widely shared, widely adopted. (41:09) Otherwise you build shelfware and that benefits nobody.

Melissa

(41:13) Yep. (41:14) AI maturity is measured by what?

Kathleen

(41:19) I would say AI maturity is measured by adoption and it’s measured by how your organization embraces the technology, right? (41:30) AI is a journey. (41:31) You don’t have to be super far along on that journey, but just get started and understand that it is the journey.

Melissa

(41:40) Yep. (41:41) What scares you the most about AI?

Kathleen

(41:46) So this is a great question because I always say I’m a polyanna. (41:50) I love it. (41:50) I’m like, let’s embrace this.

(41:52) Let’s use it. (41:53) Let’s move forward. (41:54) But what scares me the most about AI is investing heavily in tools without understanding what you’re using it for.

(42:01) And then just, you know, chasing different tools and never actually creating value.

Melissa

(42:09) I like it. (42:10) AI success ultimately requires courage or competency? (42:17) Courage.

(42:19) All right. (42:19) Last question. (42:21) One habit leaders must unlearn.

Kathleen

(42:27) When it comes to AI, I would say fear of being perfect, right? (42:36) Just get started.

Melissa

(42:38) Yeah. (42:38) And I think also kind of people depending on it for everything, right? (42:42) Depending on, I see a lot of people using it for everything and not doing as much critical thinking and the AI tools are not 100% correct 100% of the time, in my opinion yet.

(42:55) And so I think, yeah. (42:58) Well, this has been so fun. (43:00) I really appreciate you being here, Kathleen, and sharing your knowledge and all your insights with us today.

(43:06) Any final thoughts? (43:08) And then kind of secondly, what is the best place for our listeners to connect with you and learn more about what you’re doing?

Kathleen

(43:16) Yeah. (43:16) You know, I mean, final thoughts are just hopefully this was inspiring. (43:20) Get started, right?

(43:22) Don’t fear the technology. (43:23) Just start to understand it, embrace it. (43:25) Like I said, think big, start small and iterate often.

(43:29) And for reaching out, I always encourage folks to connect with me on LinkedIn. (43:33) You can find me Kathleen Walsh, W-A-L-C-H. (43:36) I usually accept every request to grow my network.

(43:41) Follow me on there. (43:42) I post a lot about different things when it comes to AI.

Melissa

(43:48) That’s great. (43:50) That’s the Executive Connect podcast.