Name: AI IS YOUR NEW EMPLOYEE
Uploaded: 2025-05-21
Duration: 24 min 10 s

AI-driven digital labor is reshaping the workplace—not by replacing humans outright, but by transforming how work gets done. The future of work is increasingly collaborative, dynamic, and augmented by AI agents and intelligent systems -- and organizations that embrace this shift thoughtfully will lead the next wave of innovation. Hear from two leaders who sit at the intersection of innovation, people and technology about the strategies and tools to lead this successfully. 

AI IS YOUR NEW EMPLOYEE

Conferences

Oh come on now. It's not always. Don't know. Yeah. Yeah Please take your seats. Our program will begin shortly. I. like so OK Yeah Yes. Oh Please take your seats. Our program will begin shortly. I Do nothing seems to change. No. And the street of the dreamers, but she often. Let them take what you believe, she said it's better. Another confusion, so the two things are. Yeah. Get a think about. Yeah Give me a. Please welcome Fortune's next to lead editor, Ruth Umo. All right, I hope you all enjoyed your lunch. Thank you and welcome back. Before we begin our next session, a few words from our sponsor calm on the importance of the mind body connection. Calm Health is focused on bettering your whole health, and here's how I want you to think about that. You see, your mind and body are connected. We're here to help you understand the relationship between your mental health and your physical health and how each influences the other. On this journey with us, you'll be empowered with tools and insights to improve both. So let's get started. Let's get started. Indeed, common thinking is that AI at work shouldn't replace real human intelligence. It should enhance it. This means creating powerful partnerships between the two, with AI working on routine tasks and people bringing human elements like creativity and emotional intelligence. We even the two together successfully take strategy and planning. Fortunately we have two guests, real live humans, not AI, to help with that. Mahi Bareti is chief executive officer and co-founder of the global HR technology company Penom and Andy Valenzuela, EVP and COO of Employee Success at Salesforce. They joined Fortune Brainstorm editorial director Andrew Nesca. Please welcome them to the summit. All right, hey everybody, let's give it up for humans, am I right? So we're here to talk about machines. Um, let me, let me lay, let me, let me lay the land a little bit, OK? Um, so there was a poll last month of tech leaders, that's important, tech leaders, OK, not HR leaders. 48% say they are adopting or deploying agentic AI. 50% believe that AI deployment itself will be autonomous in the next two years. So we're going to make the AI deploy more AI, right? Lovely. 43% say more than half their AI budget today is for Agentic. How many people in the room know what agentic AI is? Raise your hand. OK. Now I've laid the land as to like where the mindset of the tech leaders are now in the same breath, recently, Carnegie Mellon and Duke researchers decided what happens if we just make an entire organization out of agents? And they did that. They created a small software company. Are you guys familiar with this study? created a small software company and they staffed it with agents in every function, every function of a modern corporation. And They discovered that the best worker, the best agent. Only completed 24% of its tasks and managed to do so in about 30 steps. So a lot of optimism, some hard realities in all of this. Andy, I'm going to start with you. How should we in this room be thinking about restructuring our organizations? For that. Yeah, look, I mean, the first thing is you should be rethinking it. I think everybody, look, time is now. Everyone should absolutely be one thinking about what are the skills of the future, what are the, how are you rethinking your jobs? every job we talk about it within my work today, every job should be rethought about. You should be focused on investing in talent agentic talent. We have jobs today within our own organization that didn't exist, you know, 60 days ago because we're, we're thinking about this today. Um, and so from an organizational standpoint, I think it is really about creating orgs that are flexible, creating orgs that um are gonna allow for this agenttic capabilities. Um, and leaning in. I mean, I think that's most important. We don't have all the answers today, but I think it's all about leaning in right now. And just to bring the whole room with us, could you very quickly, since your company makes some of this stuff, uh, explain to the room what agentic is vis a vis generative AI and some of the other cases it's not all AI is the same. So just very briefly, could you, yeah, I mean think about Agentic is one step further in the AI revolution. It's actually helping. Taking task, it is smarter AI. It is task. It is able to do things that that plain AI couldn't do before. And so I like to think about it as broader task automation and decision making that can happen via agents, right? Thank you. So Mahe, I have been thinking a great deal about how AI feels very one size fits all, and of course it's a lot more nuanced than that, as you well know. Um, could you please tell me your perspective about how people in different industries should be thinking about this? Your company works across many different kinds of industries, geographies, and so forth, and it's not one size fits all, is it? Yeah. So I'll give you two different variations of how you segment. Yeah. The first one is you have to look at work zones. We actually look at work in 5 different zones, zone 1234, and 51 and 2 are more like. Frontline workers, 4 and 5 are more like knowledge workers. The way you deploy agenttic infrastructure is entirely different based on where you really. Play out within your ecosystem. No company is just zone 1 or 2, just frontline, just knowledge worker. You have a combinations. So now when you're really deploying, how do you think about deployment in an effective format? I'll give you examples of our own customers. Frontline workers, automation is the last 4 years is actually ramped up to a different level. And but like knowledge worker ramp up on automation is limited. The reason is when you're hiring a knowledge worker, the most important thing is not just automation. You have to think about the intelligence piece of it. So we look at every spectrum in 5 different levels, level 1234, and 5. Level 0 is what we call a stone age. Uh, there are many companies who actually live in Stone Age in recruiting and retention and growth. Now, how do you move it to 1234, and 5? So in terms of our research, there are a good number fall between 0 to 1. There are a good number fall between 1 to 2. There are very few who actually fall in level 3. So now when you sit as a company, you have to evaluate that and then you think about should I move one step forward or two steps forward. And like if you're really moving like 2 or 3 steps ahead, like let's say you're in level 2 and if you want to go to level 5, your change management is over the roof and your organizational structure has to take it in a very effective format, otherwise, it creates a lot more complexity. Your leader should be inclined into that particular equation. So, uh, in that particular fashion, agents will work in a beautiful format, but you should be aware of where to deploy when and what particular level. Right. Now one thing I think about is, you know, I said that study at the top about the researchers who created essentially in a vacuum and agenttic company, uh. And it still didn't function very well, but, but you know, we, we don't all have the option to just wipe the slate clean and start over, right? We have living breathing organizations. We have to integrate these what's what's the thing we like to say? We build a plane while we're flying it, right? We all say that now, um, so, so. How should we be thinking about What humans should do versus the agents, what requires automation versus collaboration, Andy. Yeah, I think it's it's extremely important. We think about you have to refocus the right human work, and it is find the opportunities where agents can do incredible things. Like I have a great story. We launched a new pay planning system at the company recently and it's 11,000 planners in the company. Within a matter of days we actually built an agent that did the entire enablement for those 11,000 people. So rather than Days of um meetings and sessions and training sessions to teach this entire workforce we we enabled an agent and literally real time every manager could go engage with this agent and it enabled them on an entirely new system, something we hadn't even thought about doing through, you know, the, the life cycle of the project and two weeks before we go, we're gonna, we said, wow, we're gonna think about enablement completely different and it was. Um, one of those rethink where where's my team broader focused wasn't on that enablement piece. It was, let me refocus them on another part of the work. And so I think it is constantly evaluating where can I get wins and refocus my talents on more critical opportunities. Yeah, Mahe, do you have a different perspective? The most important critical thing what we have seen is use cases, which use cases you can really apply agents. We look at agents in two different spectrums. One is experienced agents, the other one is persona. Agents. When we call experienced agents in the flow of talent ecosystem where all experienced agents can influence the candidate experience can have a con agent, onboarding can have an onboarding agent. Employee engagement can have an employee engagement agent, and you can really use personal action on all those spectrums. So your workflow is much more effective throughout the experience what you're delivering for people who are outside the company. Now you come into the other side of the equation, which is the persona. And we have a recruiter. So if you take a recruiter task, where the recruiter spends most of the time, sourcing is one aspect recruiter spends. The second thing is intake of like what they should talk to a hiring manager, what they should talk to an employee and all that stuff, then scheduling, then you can really think about like screening. So for all those we actually created agents specifically to do that. Then we have on the other side, which is HRBP. Where you really look at like, OK, what are the agents we can build for the HIVP primarily on the side of If you want to do one on ones or feedback loops, can you do like what do you call as performance management and all that stuff? How can you schedule employee and the manager meetings quite often? What's the frequency? So all that you can really do it with a lot of agents, and that's what's effectiveness. If you ask what is the single most important organization like organism. Which is super organized in the, on the planet Earth. What is that? What is the most organized? Infrastructure on planet Earth. What's that? The human body, I think it's ants. So what, the the most important thing is, agents are like ants. They take the ants work. And that's what we'll delegate to. It's repetitive, constantly really uh like which most of us really irritate to ourselves. We don't need to do it in that same. Diengaged passion. So part of it can be really engaged with an agent. That's what is the biggest aha the impact what it can create. Yeah, yeah. I want to come to the audience for questions because I know a lot of folks are thinking about agenttic, but I have another for you. So. We've been talking a lot in the last two days about skills, right? Upskilling, reskilling, this skilling, that skilling, you know, just skilling for AI, and I'm thinking a lot about that vis a vis agenttic. Uh, we had a conversation yesterday in which one of the speakers from Microsoft, Colette raised, uh, the, the, the idea of the concept of managing teams of agents, like a new skill that we will, we, we will all have to learn, um. What are your thoughts about that, Andy? Is that where we're headed? How should organizations be thinking about skilling for an a agentic future? Yeah, I think it is absolutely where we are headed. This will be a world of do they file expense reports? OK, just, just checking if they do, hopefully they're really small. Yes, yes, I hope I hope we automate that too. Yeah, yeah, yeah. All right, great. So no, we should absolutely be hyper focused on it. I mean today we are, we are. Looking at in our service centers or in um ES operations we are absolutely looking at how do I balance workload between agents and humans score we're building scorecards to evaluate that and look at it. So we are in the midst of doing this today, so I do absolutely believe it is going to be a core part of the future. And we, we absolutely have to build the skills so that someone knows how do I do that? How will it operate differently? How is I as a manager going to operate with both agents and and with humans? We got to, we have to help our folks work through that. Yeah, daunting. All right, let's go to the audience for questions. Hands if you don't mind. If not, I certainly have more of them. All right, we've outlined our, our robotic future and they're terrified. Um, so, so. We can't just deploy teams of agents in our very complex, uh, uh, uh, you know, revenue driving regulated industries without a little bit of concern. So Mahe, uh, I want to talk to you a little bit about Something the technology folks love to talk about frameworks, governance frameworks. How should we be thinking about deploying these agents in a responsible and ethical and Cogent way in an organization today. So the most important thing is why are we deploying agents? It's about autonomy. That is the most fundamental fabric what we want to really get to a system to really be autonomous. So now, when you really look at autonomous, there are 3 types of Autonomous deployment you can think about. One is operational autonomy. The second one is functional autonomy. The third one is a hierarchical autonomy. Now if you really think about operational autonomy, it's about tasks. What task for which personas I can give it to agents so that it can be autonomous. Now you go to functional side of it, it's about which particular business unit or a functional unit. I want to operate in which particular format and in that's like what is my agent is going to do for what particular section, either by automation or augmentation of intelligence. And the third one is hierarchical. This is where actually it falls flat. Every organization is super hierarchical, whether we agree or not. In a hierarchy, where is the agent operating at what particular spectrum? Let's say a company has level 7 or band 7 to band 0. If you're really operating, are the agents operating in particular bands or combination of bands? Is it crossing the limits? Because What happens is the control on the power changes when the band changes. And that is how companies are organized. Now, in that particular infrastructure, how do you deploy agents is very critical. Now, The previous point about what I really said about frontline versus knowledge worker. But when you really take front line, retail businesses, hospitality businesses, logistics. If you really look at like anything to do with uh high volume hiring. In those spectrums, you basically can think automation much more heavily. You can go to an advanced automation in terms of both. Operations and also what you call as um functional. But the minute you come into knowledge worker space, the most important thing you have to think is hierarchical organizational structure is very important. And how do you really deploy agents in that particular framework is critical because power shifts and the decision making is very expensive. Are you picking up what Mahe is putting down? I'm, I'm listening. I'm learning at the same time. Yeah, yeah, I, I, the thing, the thing that I'm, it's a very intriguing way to think about this. The thing that I'm stuck on is if we deploy these agents in our organizations, how do we ensure that they act? The way we expect human workers to, uh, that they are not biased or or overly biased in a way that hurts the organization, that they are accountable, right? What happens if an agent, if if an agent is working for a payment processor, how do you make sure that agent just keep processing fraudulent payments and actually, you know, uh, uh, are there ways Andy that we should be thinking about. These, these concerns that we would have for normal humans, but you know it's a bit abstract with the agents. Yeah, go back to the framing of its humans plus agents. Yeah, I mean the idea here isn't that they're going to completely go off and do everything on their own. We've got great examples with our customers. One that I always think is super impressive is the loan process. And the amount of paperwork and amount of checks and balances that have to happen in a loan process that takes a tremendous amount of human time than an agent can do in a matter of seconds. And so you know we have customers deploying agents in that space, and an agent can look and evaluate and make a recommendation on whether somebody is viable for a loan or not. But it's still completed with the human. The human has to come in and validate what the um what the viewpoint is of that agent. And so I think that's where the balance comes into place is if you are making core decisions or you're leveraging um critical information that. There is some human element to it to do that agent validation. Yeah, I'll give you an example. So Thermo Fisher is one of our customers who deployed our automation infrastructure specifically to automate 20% of their workflow, specifically recruiting manufacturing units. So they have like 30 manufacturing units and that they have to recruit constantly. They deployed the complete automation, which is about like level 4. It's all about agents and how it is deployed. And they thought like the recruiters are not required until there is an awful lot really popping up. As things are really going in, there was a huge influx saying like the part does not work. And the product is not working primarily because it's not about the recruiter not working out. The managers who are actually sitting on the manufacturing units. Don't know how to interact with an automation infrastructure. They're only used to really working with pure recruiters. So in that particular format, what happens is they actually converted the recruiters into relationship managers and deployed them in the local manufacturing units. So that kind of change nobody had expected, but the leader actually made that change, saying like, The fundamental flaw is not about deploying. It's not working. It's about there is a nuance which I have to really bring in, and if that nuance works, then things will work out. Well, gentlemen, thank you for giving us a glimpse into our agentic future. I feel like we could go much, much deeper, but we do certainly don't have time. Thank you so much. A round of applause for these guys. Thank you. All right, our robo.