The Real Reason AI Pilots Stall and Where Leaders Go Wrong with Sania Khana

Michael talks with labor economist Sania Khan about interim findings from their joint research on AI pilots and why many are stalling. Sania shares her background from the White House, Capitol Hill, and the Bureau of Labor Statistics to serving as chief economist at a unicorn AI company, and explains why leaders must consider workforce strategy as AI transforms tasks across industries. They discuss common failures such as whiplash directives, unclear “why,” token-cost surprises, data access and compliance constraints, and a bifurcated tool experience between leaders and employees. They also cover entry-level hiring slowdowns, workforce opt-outs, rising entrepreneurship, and the need to measure outcomes beyond cost savings. Key takeaways include redesigning workflows, choosing better KPIs, and involving workers early to build trust and adoption.

Timestamps:
00:24  Welcome and Setup
01:10 Meet Sania Khan
03:12 AI and Jobs Outlook
06:13 Research Collaboration
08:10 Early Pilot Findings
10:21 Data Access Roadblocks
13:08 Entry Level Pipeline
15:57 Hiring Slowdown and Entrepreneurship
18:21 Measuring AI Value
23:31 AI as a Culture Mirror
27:02 Three Takeaways for Leaders
30:42 Closing


Connect with Sania:
https://www.linkedin.com/in/sania-khan1/
https://www.inflectionpointstrategy.co

Transcript

SPEAKER_02 0:00

Welcome to the Top Voice Podcast, where each week I sit down with leading voices in business, leadership, and transformation to unpack the issues that matter most. Together, we explore fresh insights, bold ideas, and real-world stories from people shaping how we think about change, culture, and what's possible. Hello and welcome to the Top Voice podcast. It's August 11th, and I'm super excited for this conversation with my now dear friend, fellow twin parent, Sonia Kahn. We'll talk about that fun fact in just a moment to talk about AI pilots. We've been doing a bit of research and work in this space. And so today, uh Sonia, we're going to jump into the latest interim findings from our research and what is really working or not in the world of these AI pilots. Before we do that, for those of you who are watching live, please do leave a question or a comment. Let us know where you're tuning in from. We'd love to take those. And oh, as always, be sure to subscribe to the podcast. It's the best way to support the show. Sonia, before we get into the work that we're doing, let's just talk a little bit about the work that you do. I think it's so incredibly interesting in your background. And then we'll jump into what we're learning about these AI pilots going on across the world right now.

SPEAKER_01 1:23

Sure. Yeah. First, thanks for having me, Michael. Uh, it's a pleasure. Um, a little bit about me. I am a labor economist by trade. I spent most of my career in the public sector. So ranging from the White House to Capitol Hill to a decade at the Bureau of Labor Statistics. So I was really diving deep into the microeconomics of labor data, how are certain statistics measured and methodologically found sound. And then I moved over to become chief economist at a unicorn AI company. And I really was taking all of those data points and how are things measured and what's the macro economy looking like, and what should leaders do with this information? And then as I was talking about AI and future of work around the 2022-24 area, I realized that there's a gap that's missing that nobody's talking about yet, which is leaders could be replacing their humans with AI. And essentially I thought nobody is talking about how imminent this threat is. How can we prevent it? Um, you know, and there's like a really destructive path forward with like unemployment, mass unemployment and whatnot that we could be going down. So, how do I save um people from losing their jobs and make it valuable for employers to keep their employees? So that's the work that I'm doing now.

SPEAKER_02 3:01

Yeah, and you were talking about that before. It seems like everyone is talking about that now, and and we're gonna do the same thing. I I'm curious about maybe one perspective over the the the arc as you were talking about this maybe two, three years ago. What was the what's the difference between the expectation or the predictions three, three, four years ago and reality now? Are we are we where we thought you were gonna be or or are we somewhere different?

SPEAKER_01 3:28

I think we are at a good place. I think finally leaders are realizing that this affects their workers and they need to put and understand their workforce and not just shoot for the darts in the dark with a blindfold on with AI pilots. And and so at least they're in the beginning, when I first started doing this work two years ago, it was just, oh, well, the workforce is just an afterthought. And so finally, I'm really glad to say two years later, leaders are worried about their workforce and or thinking about a strategy of their workforce when they bring in AI.

SPEAKER_02 4:12

Yeah. Yeah. And we're gonna talk about some of these results because I think it's it's a little bit of a I have this belief that we're far better as humans at corrective behavior than preventative behavior, which which means it's easier to fix a problem after it's happened for us because we see the results. And I think the last year and a half has been uh a rational, exuberant rush to the AI frontier, and we're starting to see the impacts of that. I guess maybe one other the macro question as a labor economist, where do you put this moment in the arc of, you know, from the dot-com boom to the internet to all of these things that became huge labor force disruptions? I I guess it's obvious to everyone this is a big deal, but I don't know, where do you where do you stack this in the arc of kind of labor disruptions over the last 50, 100 years?

SPEAKER_01 5:07

I mean, I think it's at the top of the list by far. All the other disruptions only disrupted a certain subset of employees of the workforce. But this is across the board, across verticals, industries, across every single job is changing, is transforming. It's some are going to get replaced. There's a large portion of them that are going to be augmented, is what I'm calling it, really. So, and the way that you can think about that is the tasks on your plate are going to change significantly. You might still be a project manager, but all of those things that you were doing, maybe 50% of those you're still going to be doing because they require human intervention. And then there's going to be so many other ones that will be working with AI or just allowing you to do the things that only humans can do.

SPEAKER_02 6:07

Yeah. Yeah. So so let's let's talk about the work be or the work that we're doing. So you and I, we we joined forces because we we look at this problem from the same problem from two different angles. You at this much more structural level of the work that's happening across the workforce. And I look at it from the lens of the individual and how they're adapting and struggling or or or changing in some way as a result of this. So I'll give you the opportunity to talk about the idea that we had and the work that we're doing to describe that, and then we'll jump into some of the findings.

SPEAKER_01 6:43

Sure. Yeah. So I think it's a unique collaboration because, again, you're looking at it from the individual, the micro perspective of change management. And how does change management occur when there's AI transformation and AI pilots at the helm? And then I'm kind of looking at it from a broader perspective of how are organizations doing this amongst uh against the backdrop of the macroeconomy, but then also diving deep into the niche of how are tasks being disrupted? How is the workforce changing? Are they thinking about their workforce? And are they thinking correctly about the strategy of how to go about this the right the right way?

SPEAKER_02 7:33

Yeah. Yeah. So we took that idea and we've interviewed a handful of folks now. We're we're not done. We're maybe a third of the way through. And I'll just maybe share to give you a chance to drink some water there. It was not for attribution. We're we're not going to give away names, companies. We've talked to several people across sectors and some different roles. And the data has been really, really interesting. And people are writing about this at uh at a bigger level in terms of the pilots. But let's let's talk about some of the things that we're, I wouldn't call them final, but some of the things that we're finding. So, what for you has been the most interesting finding thus far and something that maybe you weren't expecting?

SPEAKER_01 8:15

We're seeing this consistent risk pattern across all of the companies that we're talking to. And leaders are essentially sending these whiplash directives, right? It's use AI, you're gonna be measured on it, now scale back, we're spending too much on tokens. And meanwhile, they're saying yes to whatever solution that walks in the door because they want the credit for you know the AI in their team. And so I would say that it's this is not of the appropriate strategy. Um, we talked to a business development leader at a large healthcare institution who's watching his team actually tank on adoption. And the reason's pretty simple. No one told them why this matters or where it actually fits into their right, their actual work. And so the directives are being changed left and right, and the strategy is kind of non-existent. And that's the risk I think we're seeing over and over again. Leaders are making this large mandate and no pl without a plan. And then the workforce gets, they get no story to hold on to. And I think that's why adoption is stalling across the board, um, across most of the companies that we are talking to.

SPEAKER_02 9:41

Yeah, it really speaks to the power of understanding what the individual needs when it comes to it's not just about the tool and it's not just about the technology. And it's so interesting to me because we we seem to never learn this lesson, and we've been doing it for years from cloud-based SaaS solutions to everywhere in between let me just turn the light switch on for your new fancy tool, and everything will get faster, better, cheaper. And I think you're describing a situation of what we've heard that that has been really, really not the case. I think one of the things that I'll add to this that I was surprised by is the data problem. And this is something that I really, I guess oddly didn't think about, but it's sort of staring you in the face, which is there are so many levels of data protection, whether it's Sarbanes Oxley, whether it's personal health information, whether it's, you know, account level data inside of a bank. You can't just go take everything and run with it, even within the boundaries of your own organization across sort of functional lines. What is that something that you we've talked about this a little bit, but I mean, again, from a labor economics perspective, that's like one of these practical realities that I don't know that we always think about until it shows up on your doorstep.

SPEAKER_01 11:03

Yeah, um, I absolutely agree. I believe one of our in our research, we're finding one person in particular said that they um what was it? It was something around the realm of they never knew that this would be a problem until they actually started to use it and that the leadership had access to a certain type of AI, and they assumed that all of their employees had access to it too. So the directive was go use this. This is amazing. You can do X, Y, and Z with it, not knowing that their employees don't have access to this. So I think there's also a lot of lack of transparency in this as well.

SPEAKER_02 11:51

Yeah, yeah. That that was the issue around this also showed up in the research that I've done separately that kind of prompted this conversation between the two of us that leaders are getting the extra sizzle when it comes to maybe certain tools that are spiced up a little bit, they have more capability. Maybe you're if you're in the Microsoft stack, you've got average co-pilot if you're uh a worker bee, and maybe somebody who's a director has co-pilot plus a couple other things, and they show up in the morning and say, Hey, I did all these cool things last night, didn't you do them? And everyone's like, I don't have the same level of permissions. Yeah. And so that's creating this real bifurcated experience between leaders and employees in terms of uh how that shows up. Now, maybe from a again, from a labor economics perspective, when we think about, excuse me, the managers versus employees, how do you evaluate that in broad strokes? Because the economy is made up a lot of different levels of people, and most of them are employees, not managers. Does that make a difference when you look at the broad scale of trends and impacts, or is it just the unit of measurement is just this general employee who we kind of treat nondescriptly?

SPEAKER_01 13:04

Um we're talking about AI transformation broadly, but I do think we're starting to look at entry levels slightly different. We're starting to look at individual contributors, slightly different managers. And then no one's really talking about the CEO angle either of how AI could be affecting their work. But yeah, so finally some research is kind of coming along, is kind of showing up in different levels. Um, of course, there's an entry-level problem, right? So if you stall hiring now, what does that mean with your pipeline problem in the future? Like your shareholders will be happy because your costs are minimized and you know the margins look good, but that's a problem down the road and not too far down the road either. Um, what I do think is a potential solution here that I haven't seen talked about much yet is can you bring in those entry-level workers and let them do the work that they were supposed to be doing for that initial first two one, two, three years in three months so they get used to it? Use the AI to start doing the work that is all repetitive that can be done with AI. Then how do you upskill them to what they would be doing in year three? And so why don't you bring them in not as an analyst, but as you can as a year three associate, right? And so that is a way to get people through the door, create your pipeline so you don't have a problem in the near future.

SPEAKER_02 14:46

Yeah. This brings up two things for me. That someone said this in our interviews, and they said that they work in a bank and they said, Look, I learned this job by keying in data from spreadsheets. And it was part of it was boring and it was repetitive and it wasn't a lot of fun, but it gave me a foundational understanding of how the business works, and that was important. And now, as a uh 50-something year old leader, they were saying, I I have insights and experience because I went through that. And and it, I'm really I'm I like this idea of accelerating it. I guess the question I'd ask is, can you really accelerate it? Right? Can you take three years of experience and put it into three or four months? I guess that remains to be seen. Right. Um, the second thing it brings up for me, and you you mentioned this, is just the the gap in hiring and then the what's happening to that entry-level workforce, because I'm seeing all sorts of articles about how hard it is to find a job, what what sort of the new level of the workforce is going on. I know this isn't research we're doing, but what what else is happening there? Are we seeing those effects happening? Because I've seen real slowdown in hiring.

SPEAKER_01 15:56

Yes. Uh, I think the slowdown in hiring is for entry-level workers. I mean, for all workers, there's a low fire, low hire environment. Um, unemployment rate is staying stable for the most part, even though we're not adding any jobs. Uh, just uh economics tidbit here.

SPEAKER_02 16:14

Um I thought that was a function of people opting out of the workforce, though.

SPEAKER_01 16:18

I thought that was yes, um, opting out of the workforce. And it's actually geared towards 24 to 34 year olds who are opting out in large um in in large numbers. Um, so when you dissect that data, that's that's you know a whole nother conversation here. But um so yeah, overarching low hire, low fire environment. Um, younger folks are being affected the most, large organizations are really pulling back their entry-level hiring. And uh to your question, it's really about these folks, if they don't get in now, what are they going to be doing? Um, what are they going to be doing with AI at their fingertips? And I and then we're seeing new business formation numbers actually on the rise. So, are they starting their own firms? Are they, with the use of AI and lovable and replit, like you could do so much and you can be so creative. Um, so maybe it's going down that path of entrepreneurship. And I think that's probably what we're going to be seeing. Is that what's going to be best for the economy moving forward? I'm not sure, but hopefully these people don't get discouraged and they actually use their time in a worthwhile way until it's time for you know hiring to open up again.

SPEAKER_02 17:46

And um, I I saw a recent article, I saw a recent article about the number of uh it independent LLC's business licenses over the last maybe four years has been higher than the last like 20 years combined or something like that. So it does seem like people are at least taking the chance to go do other things. But from your perspective, I I take your point that what's best for the economy is a that that's a complicated question because I I guess treat me as if I don't and I don't know what I'm talking about here. What is best for the economy? Is it humans being employed? Is it output being generated? Is it productivity? What when when you say that, what is how do we evaluate what is best for the economy?

SPEAKER_01 18:31

So as a humanist and economist, I would say it's um people being employed, whether that's entrepreneurs or employed at a large firm or small firm or at a firm. Um but I think there's also a statistic that, you know, was kind of a 20th century statistic that we're still using in the 21st century, which is all about productivity and efficiency. So for example, we saw the internet come out and we were all economists across the board were saying, oh, this is the time we're finally going to see our productivity numbers increase. But we didn't. And then it was like, oh, well, maybe it's because we're not using the data that we're collecting. We're just collecting all the data now. Maybe when we actually put it to use, and that's when we're the productivity numbers are going to jump. Well, that didn't really happen either. Um, the good news is we are seeing productivity numbers increase since the pandemic. More people can do more uh with their time. And I think AI is going to be helping to increase that as well. But I don't think that's the measurement that employers and leaders should be focused on. I think we are kind of in this new world here where we have to be thinking about qualitative work too. And the way that I'm thinking about this part is, you know, the quality of your decisions and your ability to focus on deep work that moves the needle is going to be very important. And AI allows you to do that as repetitive work is shifted off your plate, but it doesn't have a like a dollar value attached to it, especially not immediately. It might, you might see it in the revenue numbers, you know, four quarters later. So there's this disconnect with how leaders are measuring value from AI. And it's like the old school methods of productivity, efficiency, and and the decreased cost of labor, but that's not what we should be looking at.

SPEAKER_02 20:47

Yeah. That that that came up in the research that we've been doing. And I think this is maybe the the biggest no no-duh finding of our conversations is the rush to meet the efficiency promise and the cost-saving promise of AI has fallen short in two ways. The first one is on the cost side, which it's, and we heard this directly, go use it. People were running up tokens, they were token maxing, and then the bill came due, and everyone said, Oh, wait, slow down. And I think to your point about the the whiplash strategy, or maybe that's complimentary to call it a strategy, the whiplash effect of this go go run and do it. But I think the second thing we're even hearing, and I'd love your perspective on this from our research and more broadly, is the real impact of purpose-driven work for people who just feel like they're now trying to keep up with the machines, which which is tough to do. I mean, my brain can't process as fast as I can get output out of uh you know, out of a different tool. So, so what what what was surprising for you about some of that? What else are you seeing outside of this conversation that we've been running in terms of how people are actually feeling about? Using AI directly.

SPEAKER_01 22:03

So I'll take a step back and I'll say that I think overall AI can help us in so many ways, but it's going to be different for each person. So my sister-in-law, uh, she's a communications professional and she's using AI to summarize research that comes across her desk all the time. And she really needs to just know the big picture, make sure those big points are being hit in in their strategy and in their marketing materials, whatnot. And I thought when she was telling me this, I was like, I can't use it to synthesize research. I really have to be the one as a human to look at it, read it, digest it, put this piece together with this unrelated piece. And in order to connect the dots, I really have to do that reading. So that's something that I can't be putting off to AI. And if I do, it'll backfire for me. So I think it's like an individual use. Some things are really good. Um and I think that's the I think that's what we need to focus on. It's it's it's good for certain parts of people's jobs, but certain parts are going to stay human. And that human part will be different for each per person.

SPEAKER_02 23:26

Yeah. Yeah. I one of my one of my realizations is, you know, for for somebody like me who who runs their own consultancy, a small business, roughly, I AI is just such a multiplier. But when I listen to people that were in these big companies and they have all of these defined processes, they have all of these defined data requirements, they have all of these workflows, and they can't just automate a piece of that without it having ripple effects into other parts of the business. And that was so interesting as a finding. And I'll I'll I'll say it in the way that what we learned is that AI is acting a bit like a mirror. And whatever whatever was going on in your company beforehand, I think we heard this in spades. This is going to reveal it. One example that we heard was that there was not great cross-functional collaboration across this large software company that we had talked to. And it said once AI turned on, it just amplified all of that. Because if somebody created something cool in HR, they weren't necessarily going to go share it with the field sales team. Why would they do that? It's a different thing, it doesn't make any sense, and vice versa. And and so I'm curious about your take on that because this feels like one of the unintended consequences of like, oh, we didn't realize that if we had a dysfunctional culture and we turned on AI, that it wouldn't just magically solve all of that.

SPEAKER_01 24:50

Yep. Yeah. Um, I think we saw most of the companies that were running pilots and doing very similar things, but for their own functions and not sharing. And then I think we only saw one out of everyone we've talked to who did it really well. Yeah. And in that way, they had like a working group, um, they shared resources, they were um everything that someone had that job.

SPEAKER_02 25:17

There was somebody's job to herd the cats, basically. Yeah. Yeah.

SPEAKER_01 25:20

And that was incredible to hear. I think that that conversation really led me to believe a solution's possible here. Someone's figured it out. And I think it's because that person had this particular job to rope everyone's ideas in and make it a collaborative thing. And you'll get so much further if you work together. And I think just from the 30,000-foot view, we're seeing all of these organizations and they're all facing the same problem. And they're just kind of in their own little silos doing it. And what if collectively, and I guess this is like hopefully our research study findings can help people realize this. But if you could all work together, business leaders, you'll find what worked for company A could work for you, company B and C and D and E. Um, so yeah, I'm excited.

SPEAKER_02 26:16

And and and then, but I and I would say with the caveat that even in that instance, what we heard was each of those functions still had their own budgets. They had their own decision-making authority, they were hiring different consultants to help them with implementation. And so even though there was someone who had a job to bridge the gap and do the grunt work, there was still a structural gap in decision making and authority and accountability to go solve your own unique problems, which I think is just amplifying the fact that we're all just running around trying to figure this out in real time. And some companies come up with good ideas and you know, some really do struggle. Uh it we're and we're we're jamming through this. I don't want to give all the all the big points away, but we're we're kind of coming up to time here, Sonia. And I think so, you know, we've talked a lot here about the macro economy. We've talked about labor, we've talked about AI, we've talked about pilots. What what are two to three things that you really want people to, I guess, maybe take away from the moment that we're in and maybe where you think that moment is heading?

SPEAKER_01 27:20

Um so I talk about this a lot as a potential solution. You can't just throw pilots and expect them to work. I think you have to do a couple things. Like one is focus on measuring your KPIs correctly. So again, like what I said earlier, don't just fall back on cost, um, cost of labor decreasing. Think about measuring outcomes. Um, so number two, I would say start with workflow redesign, right? So like create a priority, a prioritization metric with like effort and understand what should take precedence when you tackle AI. Um, so kind of like a bottoms-up analysis. And then number three, I would say involve your workers from the beginning. And the reason why I really say that is you'll only get honest answers if you bring them to the table. Nobody wants to, you know, tell all of the tasks that they do and what can be automated just to be redundant and get laid off. So if you bring them to the table and tell them what your your plan is to move forward, you know, take all the repetitive stuff off your plate, redirect your work to things that can move the needle and drive value and let's upskill you on those areas. Um, then I think they'll actually be on board, start using the AI tools that have been provided to them. And they're part of the solution. They're they're helping you figuring it out. So I think that's all where adoption, that's how adoption begins.

SPEAKER_02 29:11

Yeah. I uh it's a great summary and it's a great reminder that this phrase that I use quite a bit, which is we own what we create. And so when I we saw this, we were seeing this in the research that people have a healthy skepticism, if not outright mistrust, of I know you want me to train this because you're just going to replace me with what I teach the AI engine to do. And that's not always the case, of course, but people they jump to that conclusion and then they opt out of the experience because I'm not going to do that. I don't have any interest in doing that. And to your point, if we if we give them some agency in in saying, here's the universe of the possible, let's go solve a specific problem together, that makes all the difference. And I think that's the other maybe big conclusion that we've heard is this free-for-all of wander through the woods until you find something interesting is not really very helpful.

unknown 30:04

Yeah.

SPEAKER_02 30:04

Go go solve a specific problem. Exactly. And then when you're done with that, go solve the next problem. And if you can do that enough, you'll string together enough wins that ultimately may not redesign the business, but it'll teach your workforce what to do. And I think that's the that's the real goal. But to your point about what CEOs and leaderships are coming in with is hey, go go do this, go save us a bunch of money. And of course, there's, you know, herein lies the tension and the opportunity. So we'll we'll keep going on the research. And uh, I think, you know, that's really we've given a little bit of a preview. Um so I've just loved that we've been able to talk about this together with everyone and start to put some of these big findings. And this is actually so now it's time for your bonus question. As you know, we have a tradition where the last guest leaves a question for the next guest, and they always work out. I just love that it does this. So the question that's been left for you is uh what is something you once believed about your work that success forced you to unlearn?

SPEAKER_01 31:04

Interesting. Um I believed that the solution was okay, like I thought about it from like an economist perspective, like a very analytical way of moving forward. Like, okay, if I just do this, if I calculate this, then I will give it to Michael and he'll do what I say, right? The prescription. And what I found in all of my pilots of testing out uh my solution with companies and individuals is that agency is key. Like workshops matter, bringing your workforce to the table to design the solution is the key for it to actually work. Like you can you can guide the conversation with those metrics and the analytical thinking, but you really have to bring them to the table and make them involved in order for it to actually stick and for them to do what you want them to do. And I think that applies with AI pilots here.

SPEAKER_02 32:09

Yeah, yeah. We just had that conversation last week with Cherry and Coshi about his book, Neurogiving and the emotional part of our human experience, which is look, I think for some people, the rational data is really compelling. And I think for a lot of CEOs, the rational data is a very attractive option when it comes to growth and the next quarter's results. But at the end of the day, real people have to sit at a keyboard and interact with something. And that doesn't always follow the data. It follows the experience of what they're going through. And I think that's a that's a really, really great uh reminder for everyone. So thank you. Uh so Sonia, where where can everybody follow you and your work? You do a lot of posting about a number of issues, so I really do encourage people to follow you. Where can we do that?

SPEAKER_01 32:57

Um, LinkedIn is your probably your best bet. And I have a Substack called Workonomics One, um, where I kind of do a deep dive of what I'm thinking and all of these issues that we're really talking about. So I would love for you guys to follow me there.

SPEAKER_02 33:13

Yeah, and we will include all of the links here. We will we are continuing with our research. I think it'll probably be a couple months before we finish, and we'll we're still deciding where we're going to publish and what that's gonna look like. But uh we're excited to put that out for everyone to take a look at. And uh I appreciate our friendship and collaboration. We briefly mentioned this at the beginning of the show. You are also my inspiration because you have now 16-month-old twins. Mine are mine are 10 months in two days, and you are living proof that I will survive at least 16 months.

SPEAKER_00 33:45

So that's it gets easier. I could tell you that.

SPEAKER_02 33:49

All of a sudden, at like 15 months gets it's this is the hardest thing I've ever done in my life, and there's not even anything close. So uh thank you for all of you who tuned in today for listening to this in into this conversation. Please do follow us to see the the work that's gonna come out. It's really, really not just groundbreaking. I I think it's practical, and that's what we're trying to get at is real solutions for companies that are struggling to find their way. So thank you for doing the work that you do and collaborating with me, Sonia, and for being on the show. Uh, and we'll see you next week on the Top Voice podcast. Thank you so much, everyone.