October 15, 2026 Three UW-Madison deans on AI and the future of work Written By: Doug Barton Categories: Event | Faculty Twenty-eight years later, the same question On the evening of Oct. 5, 2026, three UW-Madison deans sat down with a room of Wisconsin business leaders at Monona Terrace in Madison and talked for 40 minutes about what AI will do to work, to companies and to the people who do the work. It was the centerpiece of the executive reception that closed our executive summit day, the night before the UW E-Business Consortium‘s 28th annual conference, and it was a powerful conversation. UWEBC exists because of a moment like this one. Twenty-eight years ago, a new technology called the internet was going to change everything, and nobody could say exactly how. Six UW deans and a handful of Wisconsin companies decided the sensible thing was to figure it out together. That is still the whole idea: The university and industry learning side by side, so that what gets built is actually useful, and so that the benefits reach the people of this state. It is the Wisconsin Idea with a business plan. So it was fitting that Adrienne Hartman, EVP of Marketing at J. J. Keller and chair of our advisory board, put four plain questions to Dean Vallabh “Samba” Sambamurthy of the Wisconsin School of Business, Dean Devesh Ranjan of the College of Engineering, and Dean Soyeon Shim of the School of Human Ecology. What changes in the next three to five years? What skills should we be building? Why do so many companies get stuck in pilots? And what happens when efficiency costs us something we care about? What follows is what they said, with as little of my commentary as I could manage. What changes in the next three to five years? Dean Samba’s answer came from a thought exercise his school ran over the summer: The class of 2030 arrives on campus in a few months, so what kind of work are they walking into? The business school’s conclusion is that the job shifts from managing human output to orchestrating human and AI capability together. He used the image of a conductor who knows when the violins come in and when the brass should wait. Agents can execute fast, but somebody has to decide when they are invoked, to watch what they do and to design the end-to-end process they sit inside. Most companies have spent decades optimizing processes for human efficiency. That design work has to be done over He also reminded the room that the last time everything was supposed to change, it did not change the way people predicted. In the late 1990s, the catchphrase was “the internet changes everything,” and the forecast was that nobody would ever walk into a Best Buy again. The conversation came back to the middle then—and Dean Samba expects it to come back to the middle now, with human agency on top and AI as an assistant that people have to know how to use well. Dean Ranjan looked at the physical world. He sees the next five years driven by physical AI: robots doing inspection and safety work in places people would rather not go, like nuclear facilities, and cobots on manufacturing floors working alongside the people already there. Then he raised the topic most AI conversations skip: Every token runs on a chip, every chip needs power and nobody wants the power plant in their backyard. Public investment in infrastructure, as a share of the economy, is a fraction of what it was in the 1950s and 1960s, and the economy riding on it is far larger. Grid, transmission, renewables, fusion and the resilient infrastructure under all of it are engineering problems, and the country is already short of engineers to solve them. AI adoption will make that shortage worse before it makes it better. Dean Shim has spent four decades in higher education and has watched the overhead projector, the CD and the internet each remake the classroom in turn. AI, she said, is doing it a hundred times faster. What she and her faculty were discussing a year ago is already out of date. Her read on the technology was the most sober of the three: Any technology is good or bad depending on how we use it, and AI will be no different. It will make us faster. Whether it makes us better is up to us. 30% faster is not 30% better Dean Shim put the problem bluntly. She wants her students and her staff to use AI. What worries her is that every answer is starting to look the same. So she asks two things of anything that lands on her desk. Is it accurate, and can you explain the logic behind it? And is it authentic—meaning is it actually your voice and your judgment? “You can submit anything you want,” she said. “Are you able to defend it? I’m looking for you.” Using it unethically is off the table entirely. Beyond that, she said, you have to be smarter than the tool. The deans kept circling back to the same word: verification. Dean Samba called it algorithmic discernment. If you run a supply chain, you need to understand the supply chain well enough to quiz the model and catch it when it is wrong. Garbage in, garbage out has not been repealed. The lesson of analytics, the internet, and now AI is the same lesson: Trust it, and still verify it. The skill companies need to build is the verifying. Dean Ranjan, who’s a mechanical engineer, said the most important part of writing code was always verification and validation. That discipline, he noted, is slipping: Very few people understand what is behind today’s models well enough to check them. When an AI —and so most of us are consuming, on trust, what a small number of people put out. His image was a traffic signal. Right now every light is green and everyone is driving. When a light turns red, who in your organization knows to stop? Dean Samba gave the Midwest some credit here. He visits West Coast alumni every three months and hears conversations that swing wildly from one quarter to the next. Wisconsin, he said, dampens the oscillations, and that is an advantage worth keeping. Why companies get stuck in pilots Adrienne asked the question most people in the room were quietly asking themselves: What separates the companies stuck in pilot mode from the ones seeing real business value? Dean Samba called it a leadership conundrum. With a new technology, you want people to experiment, and experiments are local by nature: What can I do with this in my job? You hope enough people form their own mental models that a wave builds. But companies get stuck when they treat AI as a technology rollout instead of an experiment in how the business runs and how it makes money. The hard call for a leader is knowing when to stop encouraging local tinkering and start building organizational infrastructure. Productivity gains are real, but if you stop there you have sub-optimized the whole thing. Then he told a story worth writing down. Twenty years ago he consulted with a global hotel company as it tried to become an e-business, and asked what stood in the way. Data, process and mindset. Ten years later he asked a major automaker the same question about analytics. Process, data and mindset. It is the same answer today. The companies that get out of the pilot trap fix those foundations while the experiments run, and they pick the KPIs that matter, whether that is customer relationships, manufacturing or product innovation, so the effort is a journey and not a pile of disconnected experiments. Dean Ranjan added a question for anyone running a pilot right now: Just because you can do something with AI, is AI the right way to do it? Every query has an energy cost. Some companies have started giving employees a limited budget of tokens and asking them to spend it wisely, and he suspects that discipline will change how people think. He also put a question to the room that deserves an answer: Most companies made sustainability commitments a few years ago, so how do those commitments and the AI investments now underway fit together? In his view, one has to help the other. Adrienne pushed on a related point: What about when the efficiencies we build hurt the customer, the employee or the community? Dean Shim’s answer was to ask what efficiency is for. In her school, getting routine work done faster has bought time for the things that were getting squeezed out. She is making phone calls to students, parents and donors instead of sending letters, and she finds it refreshing. Technology, she said, is going to make everyone roughly equal in what they can produce. The difference will be the human relationship and the quality of thinking behind the work. Preparing people, starting where they are On upskilling, the deans were practical. Dean Shim attended what she described as Harvard Business School’s first executive program on AI last spring—the only person there from higher education—and walked in sure she would be the least prepared person in the room. Everyone else felt the same way. Her takeaway was that it is not too late for anyone. Back home, she found her own faculty and staff spread across the whole spectrum and learned that pushing does not work. Meet people where they are. Give them space, resources and permission to be a beginner. Dean Ranjan had just spent two days with his own advisory board, where members asked what the College of Engineering has ready now for reskilling the workforce. His answer was that upskilling is about mindset before it is about tools: Treating AI as additive to what you already do, and never relying on it without questioning whether the answer is right. He described a course in the college’s Interdisciplinary Professional Programs (InterPro), taught by an industry practitioner, built half on the fundamentals that cut across industries and half on business-school-style cases in the learner’s own domain. He also recalled an AI teaching assistant developed at his previous institution years ago that went on to become a widely used product. The technology, he said, was the easier part. Understanding how people come to trust it was the real innovation, and that is what the college is trying to build into its programs now. Dean Samba added that as humans move up the value chain, the skills that gain value are the old ones: problem formulation, curiosity, cross-disciplinary thinking, a willingness to experiment and take risks. AI literacy cannot stop at knowing how to use your favorite chatbot. He called it old wine in a new bottle, and said that is not a bad thing. The last question came from the audience, from the parent of a high school junior who is choosing a college: “If the technical skills a student learns today evaporate in three to five years, and the return on a degree is not what it was, how do the three schools together rebuild the value proposition?” Dean Ranjan answered with what engineering has always taught: design thinking from the first semester to the last, a capstone project with real industry partners, and the convention of working on teams across disciplines. Those three hold up in any era. Dean Samba described the business school’s version, which he calls a growth mindset: You run toward change, not away from it. Problem-solving agility through an entrepreneurship boot camp and a required capstone. Learning agility, which means learning how to learn and unlearn, because today’s AI is tomorrow’s something else. Leadership agility through experiences like Leadership at Lambeau and a new sales challenge with the Milwaukee Bucks. And lifelong learning beyond graduation, over a 25-year career in which the half-life of a skill keeps shrinking. He and Dean Ranjan have each said the university should become the Netflix of lifelong learning. The contract with students, he said, is simple: We cannot predict the world, but we will be with you and we will prepare you. That is as good a description of the Wisconsin Idea as I have heard lately. The work ahead I left the room optimistic, and I want to be careful about why. Not because the deans promised that AI will work out. They did not. Dean Shim said we have to be very cautious. In the same breath, she said she has never been more excited about her work. Dean Ranjan spent a good share of his time talking about power plants, transmission lines and a shortage of engineers. Dean Samba told us that the barriers today are the same barriers as twenty years ago, which is not a cheerful thing to hear. And the question most people ask first about AI—what it means for their own job—did not get a full answer in 40 minutes. It will take more than one evening—and it is a question UWEBC will keep working on with our members. I am also optimistic because of how the deans talked about AI. Three deans from three very different schools gave the same answer in three dialects: The technology is the easy part. Process, data, mindset, trust, verification and the human relationship at the center of the work are the hard parts—but they are the parts a university and a state’s employers can work on together. That is what UWEBC has been for since 1998. The internet did not turn out the way the forecasts said, and it also did not turn out badly, because people in rooms like this one kept asking what it was actually for. We will do the same with this technology. My thanks to Adrienne Hartman for leading the conversation, to Deans Samba, Ranjan and Shim for their candor, and to the members and guests who filled the room. If your company is working through any of this and would like company, that is what we are here for. Doug Barton is director of the College of Engineering’s UW E-Business Consortium, or UWEBC, a university-industry partnership with more than 80 member companies across diverse industries. He has held senior leadership roles at UBIX.AI and IBM. He also an accomplished speaker, having presented globally on design thinking, entrepreneurship, disruptive technology, AI and innovation. Featured image caption: Deans of the UW-Madison College of Engineering, Wisconsin School of Business and School of Human Ecology shared their perspectives on AI in education and in the workplace with business leaders who attended the Oct. 5, 2026, UWEBC executive summit. Submitted photo.