Dr. Corina Taban | Establishing psychological safety in the AI-powered enterprise

Elizabeth: I stalked Karina down, y'all. She's ex meta, ex-Microsoft, but that doctorate's pretty freaking cool. The doctorate is an organizational change, specifically in adopting AI. Ambiguity has a way of making people fill in the gaps with the worst case scenarios. The uncertainty of not knowing what's going to happen can be as bad on trust erosion as layoffs or restructuring. People don't quit loudly, right? By 2027, companies that don't have a human-centered AI adoption strategy will lose most of their best talent to competitors. Can't learn to ride a bike at a seminar. It's the same for AI. Exactly. Yeah. At the intersection of AI and CX, it's the agentic wave, a shift so huge it deserves its own podcast. I'm Latney Conett, Chief Market Officer at Parlowa. Get ready to hang 20 as we discuss the latest customer experience trends, advances in technology, and most importantly, how you can confidently lead your teams to ride this wave. And Karina is going to talk about that specifically and best practices and how to get our people to adopt AI. I'm so excited to have Dr. Karina Taban on the show today. Welcome, Karina. Let's get into it. Tell us a little bit about yourself and how you developed an expertise in AI adoption. I mean, it it's it's definitely not ⁓ in the curriculum right now. Let's just say that. ⁓ it's not. That's true. I've ⁓ spent about a decade in tech doing negotiations for for Microsoft and for Meta. So I come from the side where things move fast, where I'm tech savvy, where I work with technology and with leaders. And while I was still at Meta, I started doing a doctorate in organizational behavior. And I focused on the relationship between organizations and the workforce in fast moving environments. And then AI hit, and that whole relationship. changed and it was very interesting to explore what it did to organizations to psychological contracts to productivity to peak performance and I started in digging into what was happening not from the technological point of view and I'm sure we'll get into into this because adopting tech is just a piece of it and it's not the main part. What's very important is preparing the processes, changing them and getting People comfortable with adopting AI, with learning, with changing a whole set of mindsets. That's what actually moves the needle. And today I talk to leaders every day about this. I train and advise top-tier organizations on how they should look at AI adoption. Not from the technological point of view, they already have tons of tech, tons of tools. But from the people perspective and from the organizational perspective. So one of the things that you shared with me is AI changes what you call a ⁓ a contract, a a psychological contract between employees and employers. So maybe for our listeners, explain that and and what's happening and what that change entails. When you look at the employment relationship, you have the Official work contract. But things don't stop there. That's just the very beginning. Everything that comes on top of that builds a psychological contract between the company and the employee. What I mean by everything that comes on top of it is the promises, the obligations, what's said by a manager or by company representatives. The everything that Happens implicitly and explicitly beyond that actual written contract. That's a big part of how the relationship would will go. And that dictates and governs whether people will stay, they will be engaged with ⁓ with their work, or they will neglect it, sabotage it, or leave eventually. Let me give you an example. ⁓ I expect from my company. To have fair promotion processes. That's not in my work contract. The company probably never promised or never mentioned anything about it. But because it's in my contract, it's in my expectations, that becomes part of our relationship. And if the company doesn't respect that, or even if I perceive that the processes are not fair, that creates a breach in my psychological contract, meaning that, ⁓ An expectation or a promise was unmet. Would you say that up until now the psychological contracts have been, I don't want to say obvious, but like what you explained is something that is not crazy or out of bounds. And I I think as an employer, you would say, Yeah, you deserve a fair ⁓ promotion and ⁓ you know, bonus and a performance process. That makes total sense. But now AI enters And it's getting rewritten to some degree in real time. So, what are some examples of how AI maybe changes that contract or derails that contract? It's definitely being rewritten because in the past, the deal used to be you invest in the business and we'll invest in your growth and in your career opportunities. Right now, companies are saying, We'll invest in the business and your employability is your problem. Meaning that we expect you to do more with less. We don't necessarily invest in your growth. We've stopped hiring. We stopped promotions. We expect you to learn AI on your own time. We expect you to be innovative, but we don't give you any guarantees that you're going to have any future with the company. That makes it pretty complicated for employees to position themselves. Do you think companies are saying that? Or do you think there's a void, companies are saying nothing and employees are filling it in with fears? Cause I always say i w if there's a void, people fill it. And so cause you know, I think about ourselves as a company, and we've never said, AI is coming for your job. Like we've never said any of those things. We've said, yeah, we want you to adopt AI, we want you to use AI, we want to try to get, you know, more productivity, but we've never said, you know, we're gonna have less people. True. But if you remember, I said that the psychological contract is in big part about perception. So what you said or didn't say might not be as important. What's important is what I perceive and ambiguity. Has a way of making people fill in the gaps with the worst case scenarios. So sometimes not communicating can be worse than communicating potentially bad outcomes. And if people fill the gaps, the uncertainty of not not knowing what's going to happen can be as bad on trust erosion as layoffs or restructuring. So I actually believe in our world. There is a huge opportunity for people that had a pretty tough job today to have what would be arguably a much better job, like better scale, better pay, ⁓ less stress. But they do have to learn AI to go from what they're doing today to that job. And so we we did this study, and what we found is that a lot of folks were very interested and found it very compelling and and wanted to learn AI. but did not have any like opportunity at work. They didn't have any hands on opportunity. There was no real training. And so they sort of felt a little bit like out, like dangling in the wind, so to speak, in in terms of how to go about learning. Is that kind of what you see in your work? It is a big part of what I'm seeing in my work as well, in the sense that people are expected to, but they're not necessarily necessarily supported in learning AI. You also have another piece of it. So if you look at adoption in your personal life, ChatGPT grew in user base in a year, more than the internet grew in over a decade. So people have the appetite to use it. They see that this can save us from a ton of repetitive, soul-crush, meaningless work. The question is, what does it mean for their identity? What does that that mean for their future? Because we do still have this narrative of AI is coming for your job and your kids' job, and then you will have AGI in six months. It's always six months away, and it will, you know, take over the world and turn us into pets or get rid of us. You do have the whole narrative around AI, especially AI at work has been built on distrust. So I think that there's true, yeah. There's like a social narrative about AI that that really impacts how we perceive what's happening at our job. That's part of it. And the other part of it is around incentives and rewards. Cause you are being asked to learn in your own time. But well let me let me double click on that. So are most companies Expecting employees to learn on their own time. Is that the expectation most places? Many of them are because they're not giving people the space and the support to learn. If you look at the productivity curve in economics and in all other fields, ⁓ you usually have status quo, then you have the disruption and productivity go down. So that that that afterwards it Can go up exponentially. Most companies expect productivity to go from status quo to exponential gains, which is not really how things go. So, yes, people get a lot of pressure to deliver productivity immediately without necessarily, well, as you say, they don't necessarily know where to start, what are use cases, what are good ways What the resources, what their colleagues are doing, what's interesting specifically for them. They don't have the necessary skill set and they don't have the necessary mindset. And I what I see at least is that most companies expect them to build that those on their own time while they develop and they deliver exponential results without having the patience for that curve to actually Follow its natural path. Is anyone doing a good job? I think it's easier to do a good job when you're a smaller AI native company. most of what I'm seeing is not necessarily a good job, but a willingness to try to do better, which is I think good. We've gone through a period of experimentation. of throwing stuff at the wall, it's probably time to slow down a little bit and get very clear on what we're trying to achieve and what's a reasonable timeline for that. What do you think about companies that are saying we're gonna track your AI usage? We usually when you ⁓ do that, people find a way around it. So they're using it to build whatever, you know, they find Organize their spice ⁓ Exactly. So you'll end up consume a lot of tokens organizing your spice drawer. It's very, very, very expensive because people will find a way around that and it won't necessarily be the efficient thing to to do. So I think that was an like I I come from tech. I always say to find a good way to do things, ⁓ you need to try different ⁓ ways and then get rid of everything that doesn't work and iterate on the stuff that seems to work. It maybe seemed a good idea at the time. We've tried it. Finally it didn't seem like such a good idea, especially when the bill came in. So now it's time to find other ways that give people a chance to work with. First of all, you do not need AI for everything. I think we need to get that out of the way. The more you consume isn't proportional to the gains in productivity, the way you change your work, the way you reimagine it, and the results you'll have. And then you know it's just a reason to go, ⁓ we have so much money that we're throwing away. So now we need to get rid of people because we're poor. So let's just sit down a little bit with a pen and paper, think about what we're trying to achieve. what the actual business value is, what the core of the business is, and what's the strategy we're going to try, what's the direction, and then what are we measuring? And that doesn't necessarily need to go through AI for every single thing in the world. So you talk about silent resistors or how to identify people that have completely checked out. W maybe, maybe for the listeners, explain to us, you know, how to identify if you've really pushed someone awa away with what your AI strategy and and how you're rolling out AI. I think the best way to understand that is through your managers. Who need training because they are close to the people and they have an actual relationship. They know what the performance looks like in good days. And if they're good man a good manager and they're paying attention, then they will know that somebody isn't performing to their max capacity. They will know what quiet quitting is. So the best way to look at this is through your managers. Unfortunately, people will check out and they will not necessarily leave, which can be a good thing and a bad thing. ⁓ no, I mean quiet quitting I think is what every leader worries the most about. People don't quit loudly, right? I mean they they quit and they're looking for another job or they're kind of doing the bare minimum. I mean that that's like the thing that we're I think we're all most concerned about. I think your A players, the ones who have ⁓ the most possibilities outside of the work, most of them will actually quit. So they will start leaving. And those who quiet quit, the ones who are disengaged, will be the ones who potentially don't have better prospects for the So they stay. Plus they will be disengaged. And at some point that becomes a s a vicious cycle where you're unable to attract better talent because A players want to work with A players. And then everybody who's good will eventually leave. But that's true, AI and not AI, right? So how does AI change that quiet quitting versus call it loud quitting or outright quitting? I think it doesn't change it. I ⁓ Do believe that it accelerates it in the sense that if right now all you're thinking about is how do we get more tech, how do we increase productivity while we decrease human resources? Repeat, repeat, repeat, you're at the end of the day going to align everybody eventually. You have this, I think it was Gardner that came out with the report saying that by 2027, companies that don't have a human. Centered AI adoption strategy will lose most of their best talent to competitors. Which does say something, right? So this used to happen before as well. Yeah. This used to happen in many companies. But now, since everybody is so preoccupied with AI and they're not focused on people at all or almost at all anymore, it's happening at a much larger scale and it's happening much faster. And it will be difficult to recover if everybody quiet quits. So what are the key components of a human-centered AI strategy? Top three, in your opinion. Okay. So first let me just quote some interesting research that places 30% of adoption on tools. data and infrastructure and then 70% on people and on ⁓ processes. So the question is first, are you creating trust, psychological safety, and support for people to experiment? And I'm going to develop, well I'm going to give you these three and I'm going to develop how I'll talk about how you develop them generally. So in order to adopt AI, People need new skills, and even more so, they need new mindsets. Okay. They need to move into a mindset of uncertain tolerance to uncertainty, into a mindset of constant learning, into a mindset of experimentation, of trial and error. And you need all of these things, meaning the trust, the support. And the psychological safety in order to build that. And do you are you investing? Are you talking to people? Are you training your managers and your leaders to use AI? Because the people who are deciding what the AI strategy is, in many cases, they don't use AI themselves. So you have the leaders who are dictating the AI strategy. But they are not empathetic to what it means to actually learn, to try, to test, to make errors. And it's a whole different thing when in the morning at work your leader says, Hey, I tried this use case, and you know what? This is what went well, and this is what didn't go well. And the fact that they do that, the fact that they show you how they failed and what they did, that gives everybody. the possibility to experiment in a visible way and it give them the space to make errors. So that's the part about psychological safety. Are you investing in their actual training? So okay, you want them to learn these new things. Are they getting the space to to learn? Are they getting ⁓ the the resources? Are they getting the time? Are they able to see their colleagues do actual things to to build actual actual use cases, build actual agents, or use AI in a way that is interesting for their specific work? In most cases, that's not the answer is no. So that's a great way to show people okay, we're all in this together and we are invested in you. Getting these skills while you work and developing those new mindsets. All right. So psychological safety, training. What was the third one? So it was psychological safety. The second one was support. And you can see the support through our leaders supporting you. Are they showing you how to use it? Are you get getting training? Are you getting investment? And the third one I would say is around trust. And you build trust by building a narrative about a common future together, which is not the case, right? What we were saying before is that companies prefer not to say anything oftentimes because they don't necessarily won't know what's going to happen or they don't want to commit. Well, building and I'm going to give you a fourth one afterwards. ⁓ They they are they don't necessarily want to commit, but then people are filling in those gaps. So the best way is to clarify things around how what's going to happen, that they have a place in the company. And yeah, if you can give any reassurances, that's even better. And I'll give you a fourth one. Is your reward system incentivizing embracing AI? Think about. Let's say Mary is an analyst. Mary figures out a way to ⁓ do a report that used to take her three days and three hours. Mary now has three options. Number one is tell her manager. And what that ensures is that she'll get many more reports because now she has time. Number two is say nothing to no one. So she does her report in three hours. And then uses the rest of the two days and a half to chill. And then the third one is tell everybody, and then maybe her company will decide that maybe it doesn't need that many Marys anymore because look at all these productivity gains. So in these three scenarios, number two is a reasonable response. But what if we thought about the more subversive fourth option where Mary gets tell Mary tells the company, and the company says, okay, Mary, now you get your Friday's afternoons off to spend with your family, or you get a significant raise, or you get some new promotion opportunities. The company still sees the significant gains, but Mary has actual incentives to come forward and to rewrite and reimagine her work. But today the reward system for using AI well is usually either more work and less stability. So it's interesting. There's this ⁓ law called Parkinson's Law, and it it basically talks about how the work just expands to the time allotted. And I and and I mean this is a very old, this is very pre-AI. And so I think that what we're seeing with Token consumption up, AI adoption seemingly up, but limited maybe productivity gains is a version of that. Like I don't know if people are chilling necessarily, or if it it maybe took them a lot longer to AI that first report. And so that like maybe we're still in that trough. But I certainly don't don't feel like people feel that measurable productivity gain. And to your point, when we don't feel the measurable measurable productivity gain, how do we really incent marry? Yeah, I mean when that actually happens and that shows other people why it's interesting to actually reimagine the whole time the work you're doing and the way you're doing it. And you are right about Parkinson's law. And even with AI today, you can do so much more stuff. that you're forgetting about what the actual work that you're doing is. Well, I think it actually is a little bit like a video game. Like like it has an addictive quality to it. Like, hey, do you want me to go ahead and do this? And maybe now we should do this. And, you know, three hours later, again, you're organizing your spice drawer with with AI, which isn't really needed. Which isn't which isn't really needed. But again, I mean I think if Doing the work really differently and making a meaningful contribution had an actual reward that mattered, people would would start thinking about it differently too. And that's also what would incentivize the whole mindset switch because it's not enough to just use AI as a search partner, which is the case for most people today. You can do so much more. But it takes learning, it takes skills, it takes feeling. And not feeling that that identity threat around your profession. So there's there are all these switches that we need to try to see where they take us. Do you think that the reliability is still a concern? And and what I s what I mean by that is one of my first experiences with AI was doing a bunch of research, and it was wrong. And, you know, I put together a whole strategy based on research that was wrong. And so that was like, ⁓ my goodness, this is not good, right? And and I think that sometimes we over rotate to negative versus positive reinforcement. And and so do do you think that reliability or people getting burned by it not working is still hindering adoption? Are we past that? ⁓ well I think ⁓ there are two things. I think that certain people don't necessarily use ⁓ their judgment, sometimes don't even read what AI is telling of them and just take it as a given, which leads to a lot of AI slop and a lot of people. But Karina, it's amazing. It's gonna make us so efficient. Can do the work of fifty people. But I I mean I have the same issue sometimes as you do because it takes I feel like it takes a lot of editing. I create a lot of content and even now that I've tweaked everything, I still need to edit everything. But it still takes me less time than it would if I had to write it from scratch. Come up every time with new ideas. So I think you have certain cases where reliability is not necessarily that great, but it still moves us forward in terms of productivity. So does it does it ⁓ hinder adoption? ⁓ I believe that the more you play with it, the more comfortable you get with the these sorts of things, so not necessarily. I'd work first on ⁓ the skills, on giving p people permission to experiment, and I think that would help them see things in a different way, not as, ⁓ this is not good at all, or this is a hallucination, so I'm not going to use it at all, but rather ⁓ okay. Let's take this from the beginning or okay, let's tweak this or change that or give me that other that other angle. So, you know, we've got an interesting mantra here at Parlowa because we've got getting our customers to adopt AI and building agents, right? So using using our solution to build agents. And we're trying to make it obviously as easy as possible and as natural as possible. For folks to be able to build and deploy AI agents. So that's part of our adoption challenge. Then the other side of the adoption challenge is their customers, so consumers, need to trust and want to use the AI. So what ⁓ maybe in your research, I know I'm I'm taking you in a little bit of a different direction, but is anything in your work helping employees adopt AI also relevant to helping your Customers adopt AI. So you the the way you you are somewhere in the middle and how you get your customers or how you can help your customers help their customers adopt adopt. Yes. Yeah. Well, I I think that the thing with agents right now is that well, one switching from a Search partner to an actual agent who does something on your behalf is first and foremost a question of control, right? Do in a in a time where so much is changing, or where everything feels uncertain and unpredictable, we'll want to control the things that we can control. So delegating control. Becomes a little bit complicated for some people. But let's say there is the people who do not want to delegate control and feel uncomfortable with that, but you will still still have a piece, a chunk of people who say, okay, I'm I'm fine with this. Now the question is: what's my incentive? For example, in my case, I am a small business owner. So I I kinda I don't need agents, but agents can really help me improve my productivity, improve what I can do in in more little time. So I'm really incentivized to actually go learn or use these these things, be it on my free time or while I work. Does everybody have that? I'm not sure. And then from the people who actually have the need or could use agents. It's a question of skills, especially if you're going to build your your own. So what are the use cases? How does this help me? And how will I do it? And how what's a simple way to for me to actually understand and to build a way that doesn't take a lot of time for me? Because It's a trade-off. The time that I will spill will spend building this agent is a time that I won't spend on my business or I won't spend up with my family, with or with my hobbies. So ⁓ it's a question of you you won't get everybody, but that's probably not the point. I don't think everybody is ready to adopt agents. It will come, but it might take some time, and it's also a question of control. But People need to understand exactly what it can do for them. For me, the first time I jumped on my first webinar to build ⁓ an agent. So it I already had this interest of, okay, how do we do this? Let's do it. My question was: but what am I going to do with it? What is it going to help me with? So I think it's having very vivid and clear use cases of. How is this going to help you and how can you use it? And for people to say, ⁓ wow, yes, that's that's something I really need, that's something I can really do. And then be like, okay, either here's your agent or here's how you built it. So everything needs to be, I think, easy, but also you need to understand very, very specifically what it's going to do for you. And the way to help maybe your partners. sell them further on their side is to get them this training around, okay, this is how you paint a picture of the future with an agent or with a fur workforce to your clients. Yeah. It all comes back to that woofum, right? What's in it for me? Like everything ever. Like everything ever. Yeah. Yeah. It's so hard, but it's so easy at the same time. we did something Here with our team, the marketing team, that I gotta say worked really, really well. What we did was we actually had a full day. And everyone came to the AI day with a use case that they wanted to build. So we had them take the time and think about the use case that they wanted to build. ⁓ so they didn't just show up and we said, ⁓ build an agent, right? ⁓ and and so every team kind of came with their thoughts on and had had been thinking about it. And then we had like a coach. So we had a webinar open with a coach. And at any given time, you could check in with the coach and say, I'm having trouble, you know, w it's not doing this, and how do I get it on the right track? And so I think that it got people A excited about what they were gonna build, because they had time to really process. And then B, we didn't just say, well, where's your darn agent? We said we're gonna give you a day to do it. And we're gonna have this, you know, AI expert that is gonna help you get there. And and I that was really eye-opening for me. And I think we're gonna continue to do it because you're you s hit spot on just saying build agents, build agent, but when? Like it's not like you took all this work off, right? So creating the time and space. them being in control of what they build and why they build it, versus me saying, these are the agents I want. No, you you what would be better for you? What's in it for you? What what agents do you want? And then having the expertise literally on demand so that it doesn't feel like I've taken all these trainings. I mean, we all we've all taken trainings and then you're like, okay, I checked the box, I passed the test, but I think so much with AI is doing it, doing it, doing it. And so that worked really well. So for listeners, like, you know, set up this AI day structure. It I I feel like it really worked well for us. The thing with adult learning is that we process by doing. I mean that's ⁓ you know, you have a seminar and you make them or you give them videos and you make them listen through. If they don't engage with what they're learning While they are learning, it's not gonna stick. And they're going to say, ⁓ okay, I'll do it, I'll do it later, but then life happens and never do. So it is important to give the space, the time, and the support and the resources so they can do it in the moment and learn by doing. And the thing with this is the more you do, the easier it gets. So you can move to more complicated things, you can Reimagine as well what the technology can do because now you know how to use it and what it can actually augment or replace or automate. One of my favorite lines is ⁓ you can't learn to ride a bike at a seminar. It's the same for AI. Exactly. ⁓ so that's a good analogy. Well, I loved having you on the agentic wave. Thank you so much. I know your time is precious all the way from Paris, which is one of my favorite places to be. So next time we're going to do it together in Paris. I hope that's okay. Okay. Of course. ⁓ I want to summarize really quickly and let me know if I missed anything. We talked about psychological safety and the fact that companies aren't talking a lot about the new employment contract. And what AI does to the new employment contract. So people are filling in the blanks and it's it's they could be filling in the blanks with some pretty dark, demotivating stuff. So really thinking about how we are proactive about maybe talking about that new site ⁓ like contract that we have, employee to employer. ⁓ we talked about, you know, there's a formula, there's creating the psychological safety, but there's also training, there's giving people time. ⁓ there's giving them support and there's giving them clarity on AI expectations and and and usage. I thought that your story about what is the reward system was really interesting. Like the the way that we're a lot of people are rolling it out right now, there's not a lot in it for employees. Like almost zero. So kinda and that I that I feel like that goes back to the contract of like what happens when when I crush all these use cases with AI. Like let's play this ball forward. So I think we probably need to be better about communicating what happens when when AI is working amazingly. What what happens to Mary? ⁓ and I I like that idea of Mary like thinking about who Mary is. That makes it so much more real and visceral. ⁓ and then last but not least, control. You know, that that it it feels out of control right now. And there's a lot of things with AI that you know, that make it harder because it's not like traditional software where it's gonna perform the same way every single time. And so helping people understand that that that balance between control and output, I think, is ⁓ is really critical. So those are some of the things that I took away. Did I miss anything that our listeners have to hear? You did not. I'm going to add one last thing related to psychological contracts and the relationship between employees and employers. Know that it's always based on reciprocity. So individuals will look to establish reciprocity. If they look like they're if they feel like they get they are getting a good deal from their company, meaning supports, incentives, trust, psychological safety, they will go above and beyond to repress reciprocate and meet those standards. However, if they feel like they're not getting a good deal or something changes, they used to be getting a a good deal and now they're not anymore. They will adapt their attitudes and behaviors to match that. And that could mean neglecting the work, sabotaging the work, work, bad mouthing the company, or eventually leaving. And even resisting AI or backlash or not doing things as well as they could, that can be that can be part of it. So investing in the people is a way to make sure. that they are also giving their best and they're supporting this adoption. I love it. I love it. So it's it's a value exchange at the end of the day. And how do we give that value to our employees so they wanna to to lean in and perform and perform in this new era, perform with AI, not instead of. So ⁓ so thank you again for coming on. The agentic wave. I learned a lot. I'm sure our listeners learned a lot. Have a great day. Thank you, Karina. Thank you, Latsani. Have a great day. That was amazing conversation. I hope you enjoyed it. Don't forget to like and subscribe.