With Jennifer Harrity

Most firms measure AI adoption by efficiency gains and dashboards full of green checkmarks. Jennifer Harrity, author of The Human Cost of AI, argues on Episode 286 of The Unique CPA that the real numbers worth tracking are psychological safety and burnout. Associates who once learned judgment through repetition now babysit AI agents instead, quietly losing the muscle memory that built professional skepticism in the first place. Managers absorb the overflow, catching hallucinated outputs while their own workloads climb 40% or more. Jennifer traces how polished corporate messaging about “freeing up time for higher-level work” collides with a workforce doing more, understanding less, and afraid to say so. Randy then connects this to survey data showing rising dissatisfaction even as productivity climbs, and together they map out what leadership grounded in honesty and psychological safety could look like, and why firms that teach fundamentals before automation will be the ones still standing when the dust settles.

Show Transcript Hide Transcript

AI is changing accounting fast, and most of the conversation, at least that I hear, is around productivity, efficiency, and what that technology is going to do for us. But today’s guest, Jennifer Harrity, is asking a different question: what is the change doing to the people inside our firms? Jennifer has spent more than 25 years at the intersection of communications, leadership, and organizational change, including most recently as ESG and sustainability director at Sensiba. Side note, anybody that’s listening to this podcast before knows that’s one of my favorite firms. She is author of The Human Cost of AI, and creator of The Trust Layer where she explores AI leadership, culture, and trust. Jennifer has twice been named one of Accounting Today’s Top 100 Most Influential People. Today we’re going to talk about, well, we’ll see where we go, but I expect we’ll be talking about what better leadership can actually look like in the age of AI. Jennifer, welcome to The Unique CPA.

Thank you, Randy. Thank you for having me on.

Yeah. Before I go, anything I missed in that intro that you’d like to highlight?

I think that pretty well covers it.

Alright! So, before we even get into this, I just want to give people a story. We literally decided to record this two days ago, I think, or was it Wednesday?

Yeah, Wednesday afternoon. It’s now Friday, right? Like, yes, so this was a pretty quick turn.

This was a quick turn. But when I heard the name of the book and what you’re explaining and talk about in there, and your expertise with leadership and culture and AI, it’s a topic that obviously everybody’s kind of leaning into right now and talking about, and you and I were lucky enough to be at a CalCPA event this week where a major thing we were talking about was this topic, and so let’s give CalCPA a little love. That was a really good event, don’t you think so?

It was fantastic. I mean, the room was packed with super intelligent people, all different walks of life, not just CPAs. I mean, we had educators in the room, we had industry folks, and talking about not just what’s going on in the accounting world, like the challenges, but really looking at the future of accounting and how we can collectively overcome those challenges. So it was a really amazing conversation, discussions that were happening.

Yeah, it was a lot of fun and I give them kudos for putting together something and thinking, forward thinking like that, because obviously we’re at a time where things are changing rapidly. Before I even go into questions, I teased the book. Why don’t you give it a little synopsis of the book before we get going.

Sure. So The Human Cost of AI, available on Amazon, is basically looking at the human side of AI implementations and what it’s actually doing to the humans using AI, right? The humans that are part of this implementation, and how leaders are communicating these implementations out to their firms, and they’re probably doing it wrong. They’re probably making mistakes that are eroding the trust in the culture of their organizations. And it’s not honestly that hard to do it right. It’s just scary. So we’re going to talk, I think we’re going to talk about that today.

Yep. I mean, boy, if you think change is scary, I mean, it’s just a natural reaction for most people, there’s a lot happening right now, so people are scared for sure. Let’s just talk about leaders in general: What do you think they’re missing when, you know, like you talk about the human side of AI, it’s easy to just bury our heads in the sand and be like, okay, change is coming and we’re going to let it happen. But leaders have to do things differently, because there’s a disconnect with probably what we are saying AI is going to do and what employees are feeling AI is doing right now. So how do leaders need to lean into that better?

Absolutely. And I think leaders have their messages so well polished, right? The message has gone through marketing, it’s gone through HR, it’s gone through Legal, right? And leaders are scared to say they don’t know something. And honestly, I look at predicting how AI is going to go like you’re trying to predict the stock market. Nobody gets it a hundred percent right all the time. You can’t say you know what’s really going to happen, so why not say you don’t know? That creates more credibility with your employees, with your investors, and it’s a scary thing to say, right? It’s a scary thing to say to your board, to your investors, to your employees that you don’t know something. But if you’re willing to say, “Here’s what I do know, here’s what I promise to you, that I will tell you when I do know something, and if things change, I will communicate it,” and that you actually follow up on that, that builds trust. Employees are fine with the “I don’t know” message, what they’re not fine with is saying that, which I think we’ve all probably heard this, not just in accounting firms, in any organization, that we’re bringing in AI to make people work more efficiently, to knock out all of the admin tasks that we don’t like to do, to free us up for higher level work. But in reality, it’s freeing us up to be able to babysit ten different AI agents that are doing what we have done our whole lives, and the outputs are not as good as if we were to just do it ourselves manually, right? It’s just not there yet.

And people are starting to see that they’re training these AI agents, and then suddenly many folks are finding themselves out of a job because they put all of their experience into these AI agents, and now people are saying, well, wait a minute, this was supposed to help us be better employees, but it helped you cut 80% of the marketing team, or 100% of the customer service team, right. Then there’s all these different levels of anxiety that this creates, right? Of course, what everyone talks about, will I lose my job, that job security fear. There’s also the “I don’t know what this AI is doing, I can’t see the backend,” right? Even engineers that are creating this, they’re at the point now where they can’t see the backend of what the AI is doing. That’s scary. I don’t want to get this wrong, I don’t want to use it wrong, so I’m just not going to use it, right? There’s that anxiety of getting it wrong. And then there’s a fourth one, which is, “I have no idea what this means for me. If I use this, am I making myself obsolete? What does this mean for me? What does this mean for my future?”

And so there’s all these different anxieties and fears, but firms respond to this with just, oh, here’s more training. That’s only going to solve one of those fears. So leaders are communicating this and approaching it with this severe gap between what they understand and what’s actually happening at the worker level, and they’re not seeing what’s happening and the conversations that are happening at the worker level, and that disconnect, that gap, is what is really troubling right now.

Oh, I agree. So one of my favorite things to talk about is vulnerable leadership, and that’s exactly what you just said: Most leaders don’t want to say “I don’t know,” because they think that’s not what they have to be. But there are so many studies, and I’m sure you’ve seen this, that when you have that vulnerable leadership, when you aren’t armored and bulletproof, people are going to follow your lead a lot better when you say, “I don’t understand this, I don’t get how to do this, I need your help, or we don’t know the answer, but what we’re going to do is learn this together and we all need to be part of this,” that just makes such a stronger team. And I’m going to go on a rant here for a second.

Go for it.

I just did a talk on this at a state society event last week, and there was this really cool story that I never looked into before. It was more on culture, which I know, you were at Sensiba, and I love the culture that was created at Sensiba, which I’m sure you were a big part of. But this culture by design, which I think is leadership by design as well, rather than just letting it happen, I can kind of equate it to what’s going on with AI right now, from a standpoint that Microsoft, before 2014, created a culture where everybody was stack ranked. No matter if you were 99.9% just as good at everything as somebody else, you could have been last on the list, because everybody had to be ranked. And then after 2014, they changed to a “let’s share knowledge, let’s figure out what we learned, let’s support each other” approach, and that made a huge impact. They ten timesed revenue in the ten years after that. That “let’s help each other, let’s understand that we’re in this together” approach, where I think what you said, sometimes leaders right now are just not at that mindset. It’s like, well, I don’t really understand it either, but I’m not going to say I don’t understand it. Let’s just see where this goes, but I’m not going to show you my concerns. We can’t have that at this point in time.

Right, right. If you bring your staff along with you during the journey, that gives them a greater buy-in on that journey, that we’re all figuring it out together. It’s going to be a little messy, and mess is okay. There’s that “you need to fail often and fail fast” theory out there, where we need to figure this out. And when doing that, you also need to create a level of psychological safety within your organization so that the leaders actually hear what’s happening in reality, down in the lower echelons, where the employees are actually working and using this AI tool, right? If they can’t speak up and say, hey, this AI tool that you told us to implement, that you’re paying a ton of money for, it doesn’t actually work. If they’re too pressured to have quote unquote “implementation success numbers,” your scorecard is going to read green, you’ve checked all the boxes, but the employees are maybe staying an extra two to three hours every day to try to clean up what the AI actually outputted, because they had to use the AI and it’s just not working good enough.

Well, that’s the other thing that I wanted to get into, and I’m glad you reminded me of that. I do a survey for the profession, it’s on satisfaction levels, and we find out what creates satisfaction and what creates dissatisfaction, trying to lean into the positive and fix the negative, And so this year we actually added some questions on AI. And the answers that are coming in show that AI right now is causing a lot more stress for people. And I think that kind of goes to what you just said: “Hey, I don’t know what it’s doing, I don’t know if it’s doing it right, I do know it’s not doing it right, but I don’t want to tell you because you just spent 5 million on this,” or whatever the thing is. “But then now I’m spending extra hours fixing and learning.” So are leaders seeing that impact? Do they have the empathy to say, “Yes, I know this has taken more time and I know it’s going to hurt?” And before I ask the final question there, what our survey is seeing is, yes, we’re seeing more stress, but people in the survey are pointing to the positives that are coming, which is cool, because normally people see the short-term pain and not the long-term gain. So I think they’re seeing that there. But are our leaders showing that empathy? Are they in that mode, or are they again just ignoring it and hoping everything works out?

Yeah, I think the surveys that leaders are sending out are asking questions that are leading to those answers. So many surveys I’ve seen around implementations that are being done, the questions are leading people to answer yes. They’re geared. I doubt many organizations, there might be a few, I could be surprised to be wrong, I would love to be wrong, that are actually measuring the psychological safety index of their organizations and the burnout index. What is the burnout? Because now your folks have this tool that’s supposed to make them quote unquote “more efficient,” and now they are producing 20% more work, 40% more work in many cases? And now that increased output is going up to the managers and the partners and the directors to review that work, there’s no AI tools for them to be able to review stuff to make sure it’s correct, because then they have to review that AI output, right? So now think about accounting firms, the typical manager or partner, the partners are signing off, legally putting their name and reputation and license on the line for everything they’re doing, now think of how many staff the average partner oversees, and each of those staff now has increased their workload by 40% to pass up to that partner. We’ve got partners and managers burning out because the level of work has just increased far beyond what they ever could have imagined.

And there are managers in particular are hitting burnout because they’re not only trying to produce the work, they’re trying to manage their staff, make sure their staff understand the AI tools, and get the correct implementation numbers for the scorecard, so they’ve got the pressure, but then they also see it’s not working right, and do they feel safe to be able to talk to folks above them and tell them it’s not working? So you’ve got these managers that are between a rock and a hard place here, and everybody has so much more work now, and with these prompts, with these agents, your work pace is faster. There’s no “let me set this up and have this computer do this search while I go get another cup of coffee.” It’s like the search is coming back, and you’ve got no time to get up to the bathroom, to refill your water, to actually stop and breathe for a second. For folks, their brains to have this level of pace of work, we’re not built to do that. And so there are some real problems now that are starting to become visible as we start to see early adopters that have started to work full-time with AI and have, you know, ten different agents they’re babysitting, or 20 different agents they’re babysitting. They’re not doing the work that they went to school for or that they were hired for, they’re just managing and babysitting agents at this point. So that’s a real, we can get into the whole skill hollowing and professional identity part, which is what I like to talk about a lot. This is a big thing. This burnout is a big thing.

Yeah. I want to get into that skill how… I can’t say that word. You say it for me.

Hollowing.

Hollowing, there you go.

Hollowing. Skill hollowing. Yes.

Alright, we’ll get into that in a second, but I want to comment on what you just said. The one thing in the ’26 survey that I ran, it’s still preliminary, we’ve got about 150 responses right now, I think we want to get to at least 200 before we feel good with the data, but this year, year three of the survey, we’ve actually seen an increase in dissatisfaction levels at the partner level, and that probably goes to what you just said: They’re seeing 40% more work from each person that they’re having to see. And from year one to year two, satisfaction levels went up in those areas. This year they’ve gone down, and actually we’ve seen them go down in more entry level areas as well this year, which we didn’t see the first year. We probably need more years of data to see where this is going. Before we get into the skill hollowing, I think I said it right, I feel like I’m saying Halloween—

Yeah! Tis the season!

Yeah, oh, we’re almost there. Before we get into that then, are you seeing anything on how we’re measuring things now? Because if all this is happening and we’re fixing problems and we’re working more, are we actually measuring the right output that we’re delivering now? Because one, I don’t know how pricing’s going to happen if people are tracking hours, that whole thing’s going to change. But two, like realization, I mean, is that getting skewed and are people freaking out about that? Are you seeing anything on how AI is affecting what we’re measuring in firms?

You know, I think the traditional measurements that we used to use, the utilization, the realization, I think there are better metrics to be measuring for that. There are several folks that are actually working on what is the scorecard of now and in the future versus what is the scorecard of the past for the accounting industry as a whole. I think measuring the output, measuring not only how much of the job is AI versus how much of that job was actually humans, you’re looking at metrics that are broken down by partner, director, professional staff, admin, also AI agent in there. You have to start looking at AI agents, you know, they’re not a person, but they are doing a big level of work now, right, and they’re being held accountable for it, right? You’re paying for that usage, you’re paying those tokens to that AI agent, right? So you need to start shifting your metrics and adding in AI. Don’t set a whole separate dashboard, just like AI policies, don’t set separate AI policies. You need to start weaving it into the metrics that you’re already measuring, the policies, your governance documents that you already have. Don’t put it on as a tack on. And just start reimagining some of these metrics that we’re measuring and how AI should be woven into them as opposed to separate.

I also think we really need to start with a human’s first mentality. We need to, because when we think about what is the biggest risk and biggest opportunity at accounting firms, it’s our people. It’s always been our people. And if you’re not measuring the impacts of your people first, then everything else follows after. We know this. If we have people that have high satisfaction and high engagement in an organization, your bottom line looks great, right? So we really need to focus on those people-first metrics: the psychological safety, the burnout, how much “human in the loop” are we having on those projects? As you start to infuse AI more and more into your daily working process, you need to make sure that you have those human touchpoints throughout, and that they’re reasonable. So if somebody is saying, “Wow, 80% of my work now is just AI,” then you really need to look at that.

Right. Well, let’s go into the skill hollowing then, because if 80% of my work’s AI, how is that impacting people in the profession?

Right, yeah! So we’re well aware that there is this kind of mental mindset of “I am what I do for my living,” at least here in the United States. It’s very different elsewhere in the world. Here in the US we’re very connected to our profession, what we do for work, our title, it’s how we exude our identity and our success in life.

Right or wrong, it is true.

Right or wrong, that’s how it is, I know. So if we start to take that away, an individual, maybe they’re in their late 30s, early 40s, or 50s, right, they’re sitting there and they went to school for four years or more. They’ve invested money and time and blood, sweat, and tears into a career. They have decades of experience now under their belt. They’ve been doing this work, it’s their passion, it’s how they identify. When you talk to somebody who is an accountant, they say, “I’m an accountant.” If somebody asks somebody, who are you, describe yourself, they usually say, “I’m a CPA, I’m an accountant.” And like I said, good or bad, that’s what it is.

Exactly.

But suddenly now we have AI that can come in, and I mean, it’s not a hundred percent there yet, but it’ll be able to do an entire tax return, it’ll be able to produce an entire financial statement, do a complete audit. Those folks, we’re already starting to see it happen in customer service, we’re starting to see it happen in marketing positions, with writers’ positions, with engineers, right. Engineers that are coding are saying, well, I’m not coding anymore. Like Anthropic and Claude, no one’s actually coding Claude anymore except for Claude. So that scary realization of, you’re just at some point managing a bunch of AI agents that do all the work that you did. And slowly, if you’re using AI agents, we’re already starting to see people starting to feel like, “Wow, I’ve been using AI so much for this, it’s hard for me to go back and do that now.” It’s similar to if you’ve learned to speak a different language, and maybe you lived in a country for a couple of years, and then you come back to the U.S. and you’re not speaking that language fluently every day, you start to lose it a little.

Right.

You don’t use it, you lose it. And so folks are saying, “Well, wait a minute, I’m no longer doing what I spent years going to school and doing and practicing and have experience in, I have this AI tool now, can I go back?” And you’re not a hundred percent when you go back. And so it really starts to erode your own skills and your muscle memory on these things and your brain power on these things, and it also is taking away our personal identities in our careers. And for a lot of people whose whole identity is basically their career, you are like, “Who am I now? What am I? Am I just an AI babysitter?” So it’s really starting to impact, like I said, we’re starting to see it in certain industries and job positions as this early adoption starts to sweep through different areas.

And you and I heard about IKEA at our event, and how IKEA took all their customer service, which was over 8,000 people, and retrained them to be interior designers, so that they could create this whole other revenue stream with these people, and just upskilled them and completely replaced their customer service with AI. So that’s the innovation we really need to look at, is how do we take and pull things out, but also upskill and empower our people to be better. That’s the messaging that leaders are trying to put out, but they’re not backing it with actions.

Well, yeah, and that was a big part of what we did talk about this week, but one of the biggest topics with that is the entry level job, is that going away? Do people not have, you know, people have learned in the past by doing, and we’re doing the data entry, we’re doing the basic returns. A friend of mine, Alan Whitman, we’ve talked about this before, and he said, “You know what, I did a thousand returns; I really learned everything in the first hundred returns.” So maybe there’s a way to start training people differently, but we still need them to know the basic skills, even if AI is going to be doing something. So this was a big part of what we were talking about this week, so I don’t know if you and I are going to solve it right now, but do you have an idea of how we bring in people and teach them the skills they need to know for this profession when AI is going to be starting to do a lot of what they would’ve done when they came in? And maybe it’s just that whole IKEA story, something like that.

Right, right. And it’s this awkward stage that we’re in, in this kind of implementation phase, where we’re starting to realize that our first and second year associates, the work that they’re doing, the tangible work, is something that they’re passing over to AI. But we all know those first and second year associates, basically, their main job is basically to be a sponge, so that they can learn professional judgment, that they can learn professional skepticism, right? But they also need to be learning how two plus three equals five, and it’s not two plus three equals seven, because if they’re managing AI, they need to be able to understand if the AI kicks out seven, that they go, no, that’s not right. They need to have that understanding, so that when stuff is hallucinated back to them, they can catch it and fix it, and not just pass it up, which is what’s happening right now, it’s everything getting passed up to the managers and the partners and they’re having to fix it. And so they’re saying, well, I could just be doing this with the AI, right? Why do we have these associates if they’re just going to be passing these mistakes through, right? So, very valid.

But who are you going to staff those manager roles with in five years. So you need to have those folks developing that skillset, those soft skills, right? How to talk to clients, how to ask for help, where to go to look and make sure the answer is correct. You can’t just get an answer from ChatGPT and then go ask Copilot. That’s not verifying, which is what a lot of them do. They get something from Copilot and go verify it with Gemini or something like that. But it’s one of these things, like you need to be able to teach them how to do the long division before you give them a calculator. A lot of them are frustrated, they just want the calculators. It’s like, well no, here’s why. And I think this is where the accounting firms that figure this piece out before other folks, are going to be the ones that are winning. That’s going to be the differentiator, because a lot of folks, not just in accounting but across business as a whole, are saying, “Oh great, we can eliminate this rung, we can save money, we can be more efficient,” And a lot of them are slowly realizing, oh crap, and they’re starting to hire back. If you look at Anthropic and OpenAI, there’s a lot of communications folks and titles that they’re currently trying to hire for, and it’s like, well, wouldn’t you just ask AI to do that? I mean, no, you need those senior communication folks to be able to…

And so it’s this process of figuring out how do we bring people in that are one or two years and need to learn, and how do we teach them to use AI, be able to teach them the fundamental stuff, that professional skepticism, that professional judgment, that frankly is the core piece of our profession, that helps build that trusted advisor piece. So that’s kind of the piece that we’re all kind of flailing around, and it’s messy, and we’re going to figure it out, but no one really has the right answer right now. We heard some good stuff, some ideas, and we’ll see how that goes. But I would say for folks to maybe keep an eye on EY, because they’ve got something good in the mix that they’re trying to do.

We did hear about that as well. In fact, I think she’s going to come on the podcast and we’re going to talk about that.

Oh, fantastic.

Yeah, so that’ll be great. Alright, so, and what you said, the ones that figure this out are going to be the ones that are really going to move forward, and I agree with that. I think we all have to figure it out, and the ones that learn how to implement it the best and the fastest, or the right way, are going to have the advantage. But this is the cool thing about the accounting profession is it’s very collaborative and we share knowledge. And so this is something like CalCPA was doing this, I’m sure AICPA is obviously working on this, I do a lot of work with Illinois CPA Society, and this is a topic. So, firms in general, even AI firms, are talking about this as well. I’m going to be at an AI conference next week in New York where we’re going to be talking about this as well. In fact, I’m moderating the panel on leadership in the AI world, so you’re helping me get ready for that, so thank you for that.

Okay!

But so let’s end this with, you know, hey, there’s a lot going on and there’s things that we have to learn, and we know this. I truly believe this is one of the greatest moments of transition in the accounting profession, and I wish I wasn’t 64 so I could see it last longer and see what happened. I think I’ll stick around for a while and see what’s happening. But, one of the biggest things, and we started with leadership and the human part of it, if you had to build the perfect leader to get us through this stage of the profession, the AI age, what would a good, empathetic, vulnerable leader look like, and how would they help everybody in their firm get to this next phase?

Right. I think this leader would, as you said, be vulnerable, be empathetic, ask questions, and not be scared of the answers. Be open-minded to hear the feedback and hear the answers, and be actionable with that information, because a lot of these questions aren’t being asked, a lot of these metrics aren’t being measured, because the results might be kind of scary. And yeah, it’s going to be scary, it’s going to be messy, and we need leaders that lean into that, that want to ask the tough questions, because they’re excited about trying to solve those tough problems. And I think that’s what we really need in our leaders right now.

Alright. I think that’s great. We’ll build that leader. We’ll go work on it. I’m going to ask AI to build that leader and then we’ll see what happens.

Right, right!

Alright. Well, Jennifer, I think that was amazing information, I really appreciate it. It is really an interesting time in the profession and I’m excited about it. Before I ask you about where people can see what you’re doing and the book, this is a question everybody gets at the end of the show. I didn’t prompt you for this, so hopefully I’m not catching you off guard. But it kind of goes to what you were saying, when a lot of people identify themselves as their job, well, I like to try to find out what people love doing when they’re not doing their work. What’s your outside of work passion? So when you’re not being this guru in the AI space with the humans, what do you enjoy doing for fun?

I love spending time in nature. I love going out on hikes and just slow walks through the Redwood Forest. We’re so lucky here in California to have so many different types of forests and deserts and beaches and mountains, and just going and exploring that, and getting lost in the trees. That’s one of my favorite things to go do.

That’s an awesome thing. It’s one of my favorites too, so I love that. And then finally, then, I just teased it, but if people want to hear more about what you’re doing, where would they look? And obviously let’s tell them where they can find the book as well.

Yeah, the book, The Human Cost of AI is on Amazon. And you can find me on LinkedIn, just search for Jennifer Harrity, I pop right up, there’s not many Jennifer Harritys. And jenniferharrity .com, my website.

Nice, and we will put the links in the show notes, so everybody should be able to get to there pretty easily. Well, Jennifer, thank you for coming on the show 48 hours after we discussed it. I really appreciate it and look forward to seeing more of what you’re doing, and reading the book. So thank you.

Awesome. Thank you, Randy, for having me on. It was a pleasure.

Important Links

The Human Cost of AI by Jennifer Harrity

Jennifer Harrity on LinkedIn

JenniferHarrity.com

About the Guest

Jennifer Harrity has spent more than 25 years working at the intersection of communications, leadership, and organizational change, including 14 years at Sensiba LLP, where she founded and led the firm’s Sustainability and ESG practice. That work convinced her that transformation is rarely a technology problem, but an issue centering around trust, particularly in leadership. She now explores that idea as it applies to AI in her book, The Human Cost of AI: Leadership, Empathy, and the Cultural Crisis Hiding Inside Your AI Strategy, and in her newsletter, The Trust Layer, where she writes about AI adoption, culture, governance, and the human side of enterprise change. Jennifer has twice been named one of Accounting Today’s Top 100 Most Influential People.