Name: Human in the Loop: When to Trust the Machine
Uploaded: 2026-06-01
Duration: 17 min 58 s

Agentic systems are now fully embedded with workflows, making decisions, flagging risks, and 	acting autonomously at scale. So the question keeping COOs up at night isn't whether to automate. It's something harder: when do you trust the machine, and when do you take back the wheel?  This session will examine real examples of organizations navigating this balance with intention: where they've deployed agentic systems, what they've learned about when to trust the machine and when to pull back. Hear what works, what doesn't, and how to build operating models where human judgment and AI capability harmonize to unlock true competitive advantage. 

Sarah Franklin, Chief Executive Officer, Lattice
Francine Katsoudas, EVP and Chief People, Policy and Purpose Officer, Cisco
Moderator: Kristin Stoller, Editorial Director, Fortune; Co-Chair, Fortune COO Summit

Human in the Loop: When to Trust the Machine

Conferences

So who here in this room has, um, had to pull back an AI system after giving it more autonomy than you thought it could handle? Only want to, OK, alright, we got some people, so we're gonna be talking today about where a human really matters, where a human should be in the loop. You know, most people, Fran and Sarah, when they talk to me, say it's obvious firing requires a human decision in the loop, especially. We're talking HR, but I think that's the easy answer. Is there anywhere else, um, that you two have found a less obvious one, maybe, one that maybe you've piloted letting an agent take over and then had to think, you know what, no, this should actually be a human. And Fran, let's start with you. Oh gosh, um, yes, so it was about 3 years ago we piloted having an agent review communication between 2 employees and the communication had gone south. And what the agent was very quickly able to do was just to look at, hey, you keep asking the same question over and over and you're showing frustration when the question comes in and it was interesting because it showed us the capability. But at the end of the day we believe that's a conversation that a leader needs to have. And so there can be insights that you decide not to scale, but that was a great test for us a few years ago. It was an aha of what was possible. That's a really good one, Sarah. Um, yeah, I'd say anything that really does need the judgment of a human and especially when you're connecting with other humans. So you take a general thing like with marketing. And it's been very common to say just have the AI write the emails, write the social posts, write the SEO, everything, and uh when you do that you get an initial bump, but it, it tailors down because the AI tends to go towards the average so. And, that's where we've seen that it's, it's maybe you have an initial bump because of activity, but then the judgment of what's really resonant and understanding human nuance is not as, as easy to decide. Yeah, yeah, that makes sense. Conversely, is there a time where an AI agent uncovered something that you thought no human could ever, ever do? And Fran, we'll go back to you. There have been so many things where we see possibility and potential uh from a Cisco perspective right now we're refactoring millions and millions of lines of code in a time frame that candidly you wouldn't have thought was possible and so I, I think it's amazing for us to see those types of examples. And then on the flip side I would say that um we're about to launch an AI coach for every employee. And I think understanding that there are a lot of different use cases that we're gonna learn from, we're gonna see how it works, we're gonna make it tailored to each role and so I think there's great surprises on both sides, yeah, and, and Sarah, what about you, Alais? I mean, that's what our product does, which has been super empowering and I'm very proud of the impact it has on employees, um, from coaching to performance to really giving true workforce in. Intelligence so that you understand and you're putting in AI into the workforce and it can help you understand better what is happening in the nuance of the conversations and like one on ones for example or in weekly updates and just giving you more eyes and ears and deeply understanding what's happening in the organization because just the systems of record telling you what your revenue is or what your costs are don't tell you truly what's working but. Conversations that you can have and the updates that employees are sharing has been too hard to get the important nuggets um from before but now you can so so you have that workforce intelligence, uh, which is very powerful um in the in the world today. Yeah, absolutely. So Francisco just reported record earnings thanks to AI which was incredible congratulations, but I know, you know, also in the news is that. They also announced 4000 layoffs, um, as part of an AI-led restructuring and I'm hoping, you know, as the chief people officer, how do you ask, you know, the remaining workforce to both lean into AI in this exciting new time but also deal with the realities that yes we are doing layoffs and yes this might replace you. It's, it's really hard as you all know, and we had a very candid conversation with our employees because we know that roles are gonna continue to evolve and we know that there are places within the company where we're gonna invest quite a bit from a resource perspective and we may not at any given time have the the skills or the people to focus in those new areas. I think for us the ability to get in front of and to have assumptions around how work is changing. Is part of the contract that we have with our employees and so something that I want is from a transparency perspective to be able to say hey these roles are the ones where we anticipate the biggest change and now here's the training that we need you to get into right away. What we have found is that when we've had previous restructurings together with the training that we provide but also using recruiting internally and placement externally. We've been able to place 75% of the people that are impacted. And so just imagine if that became 85% or 90%. It would make people feel a lot less worried because they know they're going to land and so it's what we're working through today, but it's tough. What do you think it'll take to get to that to that 85 to that 90? I think, uh, clarity on how roles are changing. So right now Cisco leads, uh, an AI consortium. We work with about 8 other tech companies. We're sharing all of our data together and it's something that's available. We're sharing all of our training as well. We're using that to start to say how are the skills shifting for a cybersecurity analyst. And if people understand how that's changing, I think it'll allow them to prepare. And then I think you're just gonna have to invest as companies in helping your people land well. Yeah, that makes sense. So I wanna talk about Lattice a little bit. Um, so a few years ago, for, for those in the audience who don't know, when Lattice tried to give AI agents employee records a few years back, there was huge backlash. You got a lot of hate for it, but now so many people are, are bragging about it, doing the exact same thing. Um, how does it feel now? Do you think it's, it's just kind of come full circle, and why weren't people willing to accept it back then and, and now they are? It was something that, uh, so yes, it was hard and especially because the reason why we did this was because we saw this moment coming we saw that people were going to be taking headcount dollars and wondering if they should spend it on technology. It's a very hard place to be in and something that needs to be done, a decision made with, uh, a lot of transparency, accountability, and responsible use of the AI. So it was, um. It was early, but it doesn't mean that I'm happy to say I was right. I'm actually sad that I was right because it means that we are where we're at, where we have all these questions, so much uncertainty, and we need to focus on how we make people successful and how we help people in AI work together, um, as you're saying, which is incredible at Cisco, like how you're investing in skills, how you can invest in skills and not just severance, how you. Bring people into the future of people AI because people aren't tokens. Token usage is not a proxy for performance. It's far more nuanced than that. If you ask somebody at a company who sends the most emails, you wouldn't say they're the most productive person. And so this was the thing was that we have processes in place for humans in the workplace, how we on board them, how we train them, how we, uh, manage them and and grade their performance. We should do the same for AI. We were chatting earlier about your home. Uh, we all have homes that we, if some person knocks on the door, you don't just let any person into your home, talk to your children, eat your food, sleep in your bed. You ask them who they are, why they're there, you know them, and it's the same thing with our work. We don't just let any AI in. We need to have clear guidelines and clear guardrails around what happens when you bring AI into a company. Yeah, I love that metaphor. I'm going to turn to the audience in just a second for questions, but first I want to ask you both about something I'm really sick of hearing. About which is any time I talk to, you know, CHRO about AI they're telling me, well, you know, we're using AI to free up our humans, uh, for more meaningful work, and I'm like, but what does that mean? Who decides what is, is meaningful? Who is the judge of, of meaning, uh, Fran, how are you looking at that at, at Cisco? Oh, it's, it's, uh, a great question because I think more and more leaders have to have a much better sense of what is valuable, especially as it relates to work flow and uh in the past, you know, leaders were guided to just let your team go delegate the work, check in and see how they're doing and I think what's happened is over the years there's quite some distance between leaders and workflow. And I think in order for us to make those decisions and to understand the value in a process you have to be close enough to the work and so as an example of something that I'm doing at Cisco is I'm talking to my leaders about becoming more of a player coach specifically as it relates to workflow and processes and guiding them to say what three workflows are. Most interested in and and what is the work that you're doing to ensure the value is there because I, I think you're right. I think if we all define value in a different way we're not gonna get the results that we want. Absolutely. Is there anything to add? it's an easy question because Lattice's mission is to make work meaningful and for the past decade that's what we've been, uh, focused on and it comes down to as a Human deeply understanding not just what you're doing but why, what is the reason that you are there and how does it connect to the greater purpose and the greater, um, objective of a company? And when you can connect those two things, not just in work, it's in anything you do, it's so important that as a human we understand why we're doing what we're doing, otherwise there is no meaning that you feel in it. And so that is why when you speak of processes, a process that is deeply broken in, uh, corporations is the performance management process. It is cyclical. It is once, maybe twice a year you check in, and as you were sharing, oftentimes there's drift. You set up your OKRs at the beginning of the year, then 6 months in, priorities have changed. Focus has, has changed, not that that's bad, it's that the performance process hasn't kept up with the business. And that's our focus is making that real time and in the flow of work so that you can find meaning and that you don't have this drift because a process didn't keep up with the reality of the business. Makes sense. Any questions? Anyone out here? See one here up front, and we have a microphone right coming to you. Yeah, um, thank you both. Sarah Buckler, Chief operating officer of 50,000 ft. I'm wondering, and I think the two of you are uniquely positioned to talk about trust in the age of AI. Um, particularly trust with all of our employees as we are asking them to do things that intrinsically as tribal people feel like they're going to take something away from us. How are you thinking about trust in this moment and what are you implementing to build trust? Um, I'll start by sharing that, um, I spent 15+ years of my corporate world at Salesforce and, uh, worked directly for Mark Benioff and Brett Taylor, and what was deeply rooted in me is you build trust through transparency and I will say it as a CEO, um, I have, uh, biweekly all hands. They're pretty organic and I share. Very openly how we're doing, uh, what we're doing with AI, taking our EPD organization through a transformation to becoming builders, what that means for their careers and having very candid conversations and that transparency. Has been monumental at building trust. It doesn't mean employees always like what we have to say, um, you know, just like my kid doesn't always like to eat her broccoli, but, um, it's important that I share with her why eating vegetables is important and it's, it's very similar just. Your employees trust you when they can trust what you're saying is true and that you will be there when you have questions, but I would just say the short version is just transparency. It's interesting because something that we saw about 9 months ago was that On the teams where they were using AI very effectively, um, trust started to drop within the team and so what that taught us is that we just have to invest so much more and I agree with Sarah. I, I think it's really all about transparency and I think that we have to share with our people what we know, what we don't know, some of the assumptions that we're making, the risks that we believe in, and then you marry that trust with values, um, that you consistently. Um, lean on as you navigate, but I think more and more as you go through change, trust is harder. Any other questions? Mhm. Hi, uh, Erin Fenton with GNG Insurance. I was curious on your perspective on, uh, using AI within culture and driving culture. Uh, we hear a lot about increasing revenue or driving efficiencies and cost savings, and I know Cisco and other companies are best place to work are really. Well known for their culture, are there any tangible examples you can use how you've used AI to improve the employee experience or enhance that culture? I think you mentioned individual coaching, uh, agents, but what other tangible examples could we potentially learn from, you know, from your guys' great culture? That's a great one. Yeah, so I'll, I'll share too. I, I think you were right to, to hit on the learning side, and I think more and more AI can help us customize learning to our, uh, employees based off of where they wanna go and the job that they're in and then the perception that we have around change. I think that's easy. The other thing that we've been doing for so long, which has been so helpful, is you can use the technology to really get the sentiment of your teams and uh when we're in all hands meeting, uh, it's so much more important that I hit the hard hard question or that I acknowledge the sentiment that is coming through from technology and so we're analyzing chat. As the meeting is going on after the meeting and those insights allow me not to have rose colored glasses in some cases I have to hit exactly what I'm seeing. Those are things that are important, but I would absolutely say that I think we have to continue to build use cases around culture. I, I think it's there, yeah, I, I + 100 I guess I would say, um, you know, we have also built the tools for the individual. Coaching the agent that can join your one on ones that can help you not just understand what's going on but can practice with you. Maybe it's a difficult conversation maybe you overspoke somebody and people were afraid to share that information with you. What's interesting is that we've seen, um, they're having AI in a way can help us be more human because we don't feel so judged by the AI if it gives us feedback. And the other thing I would say is that, uh, yes, AI being having more eyes and ears, as we said earlier, it really helps people feel like wow, like she understood this or they listened to that because you can now and so it having it in your culture and then also employees knowing that you're investing in skills, not just. You, you want to give them the AI skills because you care about their careers. Maybe they are going to have a career with you for many years. Maybe they're going to shift to another place. But having that loyalty and then feel like, you know what, I worked at Lattice, and Lattice is where I got that groundswell of opportunity because I learned the skills. They feel like you're investing in them and they deeply care about that. And one quick last question for you both. If you could redesign one thing about your organization today to be ready for AI as a true colleague, does maybe in the next 10 years, what, what would you wanna focus on and, and why 30 seconds or less? I know that's a quick one, Fran. OK, quickly, I would not look at AI as a colleague and so that I, I would say, and it's so funny because I had this debate with one of my peers. I just don't think we should look at it that way. I think we should look at AI and agents as part of the workflow but not a colleague, and I think the sooner we land that, the more confident our people will be. I love that answer, Sarah. Um, what I would say is that I would actually have people do more in-person bonding than just AI. I would do more touch grass and then more activities, more softball. Yeah, I, I would. I think AI is everywhere and I think that it will never replace that human connection. I love it. Well, thank you both so much for your insights. Thank you all for the question.