Ep 250: How AI Erodes Trust on Teams

AI tools create a specific kind of trust problem on teams, and most leaders aren't sure what to name it or where it's coming from. In this Practical Field Notes episode, Mimi and David react to a Harvard Business Review piece by Jayshree Seth and Amy Edmondson on psychological safety and AI integration. They connect the article's ideas to what leaders are actually navigating in their organizations, including how to build learning loops that keep AI use transparent and how leaders can model the kind of openness that makes it safe for teams to question, challenge, and learn. Humanergy is a leadership development company that helps leaders and teams work better under real conditions. Learn more at humanergy.com.

(Note: The HBR article is behind a paywall. HBR offers free monthly access, if you're out of free reads, you can still follow this conversation without it.)
Link to article here:
https://hbr.org/2026/02/how-to-foster-psychological-safety-when-ai-erodes-trust-on-your-team

Learn more about Humanergy's work: https://www.humanergy.com

Join the Humanergy community on LinkedIn.

Sign up for our FREE leadership workshops. 

Transcript

Mimi Mitrius (00:10) Well, hey everybody. Before we get into this episode, I just want to introduce myself. I am Mimi Mitrius with Humanergy. And I also want to introduce you to something new that we're doing. We're calling this Practical Field Notes. It's going to be a recurring series where we take a recently published article on leadership — and it doesn't actually have to be recent. It could be an older one that's very evergreen.

We're going to bring some Humanergy coaches in to talk about it, to react, to maybe push back, or to connect it to what they're seeing actually happen in the organizations they're working with. So really, it's just going to be a conversation about the ideas that are going on in the world of leadership right now.

This was a good place to start, so I'm going to give a little origin story about how this article came across our desks. A couple of weeks — well, months ago — I was working on one of our free leadership workshops. If you haven't signed up for those, you should. I was working on a presentation centered around one of our frameworks, the Four Choices, and how to use it as a practical tool when working with AI. I was about 90% done with the presentation and about two days out from delivering it, and then this article popped up in my LinkedIn feed.

I had kind of a sinking moment after I read it, because it was a really great article, but I was also worried that I had somehow accidentally plagiarized. I had a moment of freak-out. So I sent it to David and I was like, "Help. Did I do something wrong? Are we in good shape?" And so I'm going to hand it off to you, David. But first, right before I do that: the article is "How to Foster Psychological Safety When AI Erodes Trust on Your Team," authored by Jayshree Seth and Amy Edmondson. It was published in the Harvard Business Review on February 4th, 2026. We will link to it in the description. It's a great article. David.

David Wheatley (02:11) I appreciate that, Mimi, and thanks for the intro. I think if I build on your story, one of the things I shared is it's always kind of cool when your thinking aligns with someone else's thinking and you have that moment of, "Have I read this before? Did I just..." In reality, it's just because this is part of the conversation that we're having with people.

AI has really been a big topic in all of my coaching work. Your presentation in Episode 240 of the Humanergy Leadership Podcast was very timely, because it really slots into what leaders should be doing, mapping AI use over the Four Choices — there are potentially lazy ways of doing it and impactful ways of doing it. This article helps validate that, in my mind.

I actually shared it straight away with a couple of clients, because it's been part of the conversations we've been having. How do we manage AI, and how do we make sure it's not being used in a lazy way? How do we make sure it's not undermining the ability of the team to create? That's really the heart of what Seth and Edmondson are writing about.

Mimi Mitrius (03:33) Yeah. Just for our listeners, a quick one-sentence summary of the article: AI tools aren't just a technology challenge. They create specific, hard-to-name trust problems on teams, and leaders need to treat AI integration as a team learning challenge to address them.

So with that, David, how are you seeing this play out in the organizations you're working with today?

David Wheatley (04:02) I want to start with a line from the article at the end, actually. One of the closing paragraphs says, "The future of work isn't about choosing between human intelligence and artificial intelligence. It's about building teams that allow both to contribute to their fullest potential."

Way back in January, we did a podcast episode looking at how a leader's job is to make sure they're not delegating leadership, judgment, or empathy to AI. Where Seth and Edmondson pick up in the article is: let's also make sure AI isn't getting in the way of building a team.

I think about this in terms of how we operate at Humanergy. We've been virtual since 2019, so you've never known us to have an office. One of the things we do is build in structured time for us to interact. Part of that structured time, these days, is building a learning loop around what we're discovering about AI so that it's shared. We've been doing that for a few years. And then this article comes out and basically says: do the same thing.

If you're adopting AI, create rapid learning loops. Amy Edmondson's book "The Right Kind of Wrong" talks about building agile learning systems so we can fail, learn from it, and apply it at the next level. That's integrated into this article — if you're using AI, build learning loops, and create separate spaces for people to have the conversation about what they're learning, both positively and negatively.

Mimi Mitrius (05:48) Yeah. To give a little color to how we do that: once a month, Humanergy has a team meeting we call the Academy. It's about an hour and a half. The first agenda item is AI, where we all come together and share what we've been learning, what mistakes have come through — like, "This was an 'oh crap' moment when I used it and it spit out the wrong information."

It's been super valuable. Framing it as a learning tool helps create psychological safety within the organization. At least for me, it made me feel very comfortable sending that article to David and saying, "Did I do something wrong?" I knew he was going to guide me in a way that felt like we were doing something right together. So just creating the framework and structure around it matters.

One of the other things I've thought about as this has unfolded is the team charter work you do, David. I'm wondering if you could see an AI team charter, or some kind of structured framework around AI, that might help create this psychological safety and bring it from theory into practice.

David Wheatley (07:27) I think that's a good idea. As we build team charters using the What Great Teams Do Great model and people start overlaying AI on top of that, if you use the article's approach and build in rapid learning cycles and spaces for humans to interact around what they're learning, that should be a natural part of it.

What you said about not fearing bringing this issue up tells me we've been relatively successful in creating a culture where mistakes are actually rewarded in some ways. If you identify a mistake, we get it out there and recognize it. Where it becomes a problem is when we either cover it up or repeatedly make it.

That's where I go back to the article. We're in a rapid learning process. Let's make sure we're tracking it, not repeating those mistakes, and sharing the learning in a way that gets the most out of both the human and the AI side.

One example came to mind recently. I was working on a tool of ours, trying to evolve it into the next generation. I'd run a bit long on a slide, so I was squeezing content in. Google keeps teasing you with the "Beautify this slide" button, so I thought, let me try it. The first time, it eliminated a bunch of things. I went back and said, "Don't eliminate anything." The second time, I thought, "That's perfect. Beautiful, really tight, fits."

And then I noticed a typo. I'm not Mr. Detail, but something caught my eye. I went back to the original and there was no typo. AI had introduced it. And I think what this article is talking about is exactly that moment: now I've got questions about how I trust AI. That becomes a judgment check: I'm going to produce something, but then I'm going to evaluate it in detail before it goes out.

And here's the other side of it: if something like that does get out, it undermines trust in me. People start wondering whether I'm throwing everything into AI without applying any judgment. That's really what Seth and Edmondson are getting at — how quickly trust erodes if we don't have basic systems in place to make sure we're talking about it and learning from it.

Mimi Mitrius (10:37) Yeah. The article touches on something interesting: when a person makes a mistake, we can chalk it up to being human. We can repair it in real time, within a group, with each other, assuming there's already some level of psychological safety. But when we put something out that was AI-generated and either don't disclose that, or do disclose it and it still has an error, the team starts to think: "AI was wrong, and the person who put it out didn't double-check it." We're creating this weird loop of discomfort and probably extra work for everybody.

David Wheatley (11:34) Yeah. And that's what they're talking about when they say AI erodes trust. The psychological safety piece comes from knowing we can admit mistakes and move on. We should be learning from it and sharing it, building even more psychological safety.

There's another line in the article that stuck with me: teams must operate in an environment where everyone feels able to question and challenge performance. And they add that the same should apply to AI's performance. We should be able to question and challenge AI just as we would each other's work. If you can create that culture, AI slips in seamlessly and becomes an asset. If you can't, it has the potential to undermine relationships and slowly degrade the power of the team.

Mimi Mitrius (12:26) Yeah. So in that presentation I did — which David mentioned you can find for free on the Humanergy Leadership Podcast, Episode 240 — that's one of the things I talk about. If you're familiar with the Four Choices framework, we make four kinds of choices regularly: destructive, passive, productive, or transformative. Creating psychological safety around AI is a transformative choice. It usually has to be built into culture to feel successful. Those kinds of things happen when the conditions are right.

David, what do you see as the right conditions for creating psychological safety within an organization?

David Wheatley (13:20) Well, the article says: first, let's talk about it. What would it look like? You and I have been going back and forth on what our AI governance looks like, what our expectations are. One example: we don't share confidential client information with AI unless we know for certain it's locked down. That becomes a question we need to answer as a team. How do we know that? How do we play with that?

If we have that conversation upfront, then set expectations, and then run rapid learning cycles to refine and align, people start to feel: "I can bring this to the team. I can bring this to colleagues. I'm not going to be judged for it." Amy Edmondson's "The Right Kind of Wrong" goes much deeper into that, but the core idea is creating an environment where people feel able to ask questions and challenge.

If the team doesn't feel able to do that, you don't have psychological safety. How do you build it? It takes a focused effort: set the expectation up front, then model it in how you respond when it happens.

Going back to your example: I would hope we set up front that we expect transparency. If you know about something, tell us early — we can help. If you tell us late, we just end up judging you, and we don't want that. And then our response the first time you actually do it is what generates the psychological safety. If you said, "Hey, I think I screwed something up," and I responded well, that's what tells you it's safe to do it again. I might still be kicking myself for the mistake, but nobody else is.

At Humanergy, I'd hope we're getting to the point where if I'm doing something wrong, someone tells me. Don't just let me sit there because I've been here for 26 years. I'm still fallible, and that's what we want. We have this hugely powerful resource in AI, but it lacks judgment, empathy, and context sometimes. The structure we create around it is what allows people to play with it, learn from it, and share that learning so it builds psychological safety rather than eroding it.

Mimi Mitrius (16:24) Yeah. From an employee standpoint: the more my leaders bring something to the table and say, "Hey, I screwed this up, here's what happened, here's how I dealt with it" — that creates safety for me to do the same. So to any leaders listening, whether you're leading a team or a whole organization: the more you make your learning visible to your employees, the more you're going to foster a culture of psychological safety.

David Wheatley (16:57) Yeah. I had a great conversation with a leadership team we've worked with for a while, where the CEO was pushing the idea of rapid learning cycles. Let's try different things, understand what AI is capable of, play with it. There was some caution from other people on the team. We ended up having a really good conversation about what was generating the caution and what was generating the excitement.

Where they landed was: we need to keep experimenting, but we need to be thoughtful about how we use it. That creates the sense that we're not just throwing everything at AI, but we're understanding what it can do before it goes out into a public or team setting and potentially undermines what we've built.

Mimi Mitrius (17:55) Yeah. So for a leader who reads this article — and we hope everyone will click the link and read it. It is behind a paywall, but HBR is gracious and gives free reads every month. If you're out of free reads, just wait until next month. For leaders who read this and think, "Yes, this is exactly what's happening on my team" — where would you suggest they start?

David Wheatley (18:20) A few things come out for me. First, normalize some of the anxiety. We're entering a new world. We're going to be anxious. If we can openly discuss that and understand what's out there, that's what starts to build trust. Let's talk about what's causing the anxiety, and then address it.

Second, create smart learning cycles. How often are you using something and then evaluating how it went? We tell our clients to use after-action reviews — that's U.S. Army terminology for rapidly learning from the gap between where you planned to go and where you actually ended up. Do that with AI.

Third, encourage people to critique it. What I learned from my own example is that I will not put something out from AI without thoroughly reviewing it. It replaced the word "to" with "too." I have no idea why. It didn't make sense. It was such a small thing, but potentially so impactful.

And finally, maintain a focus on collaboration. It's easy to say, "I've got AI, I'll go collaborate with AI." The value of our Academy sessions is that we bring it back and say, "This is what I did with AI, and I'm looking for feedback. What are other people building on?" We keep the expectation that we're stopping and talking about this together. We're not losing the human element.

Mimi Mitrius (20:27) Yeah. One of my favorite lines from all of this was from your presentation, which I think is what inspired mine: "When we outsource our judgment to the algorithm, we've outsourced our humanity." I think that's the heart of it.

David Wheatley (20:45) Well, there's an irony to that. As I was writing that piece, I was iterating it through AI to polish some of the language. That sentence is probably not one I would have come up with on my own. But as I ran different iterations through the process, it helped sharpen it. And then I took it away and was able to polish a sentence that said, sharply, exactly what I was trying to say in my long-winded way.

Mimi Mitrius (21:17) So the moral of the story is: AI can erode trust and psychological safety, and when used properly, it can also enhance the work we're doing. There's a balance.

David Wheatley (21:29) Exactly. And if you enjoyed this conversation, I'd encourage you to go back to Episode 240 and listen to Mimi's presentation there. She labels all of this in really practical terms.

Mimi Mitrius (21:41) Great. Thanks, David. Thanks for doing this with me.

David Wheatley (21:45) Appreciate you bringing this one up. It's a good article. Looking forward to the next one.