Wharton study co-author Jeremy Korst talks:
•Shifting perceptions about the ROI of AI among managers
•Which organizations have most effectively scaled AI
•The technology's potential impact on jobs
Artificial intelligence seems like it's rapidly approaching omnipresence in corporate workflows, but what do today's deployments look like in practice, and where does travel management fit among corporate AI priorities? The University of Pennsylvania's Wharton School for three straight years has conducted surveys and research about corporate AI deployment.
BTN managing editor Chris Davis spoke with Wharton study co-author Jeremy Korst, a longtime tech executive, a member of the Wharton executive advisory board, and CEO of AI strategy consultancy and incubator Mindspan Labs about his findings on current corporate AI philosophies and investment strategies as well as the technology's effect on future employment. Edited excerpts follow.
BTN: Now that you've done this for three years in a row, what is the overall corporate philosophy regarding AI deployment and adoption?
Jeremy Korst: Over the past three years until now, it's come to the point where we now see a large percentage of business leaders—managers, directors, and above—become at least weekly users, and now almost half are using AI daily. Consistency through our study period shows that leaders have been generally positive with their outlook and perceptions of the technology.
This year we added a question around their perception of ROI. Approximately three-quarters are seeing at least moderately positive ROI on their initial investments. One reason we think we see that is, we intentionally took a more broad-based question around their perception of ROI versus other studies like McKinsey or MIT, which have been trying to find actual, tangible financial ROI. We did this because it is early in the adoption cycle and may be hard to find measurable ROI at scale currently. We wanted to understand their perception because these leaders are the ones making budget and deployment decisions. We found that they do have intent to increase budget [with] an expectation of increasing returns.
Another point is how we're looking at skills augmentation or enhancement versus skills replacement. Over 80 percent say they expect AI will enhance or augment skills, while over 70 percent also say they think some skills will be replaced. I'm saying "skills" specifically, not jobs yet. This year we added a question about whether they were worried about the impact on skills. Leaders said that about 43 percent were worried that the effect of replacing some skills could be detrimental to the organization. Each organization will likely make that decision differently based on their strategy. In travel, you can imagine an online low-cost player versus a high-end approach where you expect a human hands-on approach.
We also found a gap between executives—VPs and above—and middle managers. The most senior executives have really positive perceptions of current ROI, whereas middle managers had less so. When asked about the relative speed of adoption, twice as many executives said they were ahead of the game compared to middle managers. We surmise this is because middle managers are closer to what's actually happening, having to deal with the hallucinations and the day-to-day challenges, whereas executives are foreseeing the future.
BTN: It’s still a relatively new technology that's been widely adopted quickly. You mentioned the executives' personal use. How much of that drives the organization? If a CEO’s personal experience with ChatGPT is positive, does that reflect how the organization approaches AI?
Korst: The intent is there because executives are seeing some of the value. The tasks executives do, like document summarization, content creation and research, lend themselves to current AI capabilities more than task-driven and function-driven roles lower in the organization. Those executives also have been ahead of the curve because most organizations allowed them to test the technology first due to governance and security concerns. So they're a bit ahead in the adoption cycle. This year, I expect to see agentic AI—AI that can do its own autonomous task-level workflow work—come to the forefront, which is more applicable to the work that actually gets done in the rest of the organization.
BTN: How much discrepancy is there between different functions or roles, like support roles versus customer-facing or revenue-generating, in terms of adoption and philosophy?
Korst: Interestingly, marketing leaders have been some of the lowest and slowest adopters over our three years of data. This is a function where McKinsey estimates there may be the most impact, yet there is a gap between current use and expected impact. Anecdotally, marketing has been inundated with new technologies for the past 10-plus years, so maybe they are in a wait-and-see mode. Perhaps there's been some resistance as well.
On the other hand, the technology and engineering space has seen heavy adoption of coding agents in the last six months. Procurement also ended up being a top-tier adopter because they are in the center of it, evaluating AI tools for other functions. Operations were laggards initially but have started to catch up. Same thing with HR.
BTN: Our survey of travel managers found that while most are using it in some form, most say their organizations are still just testing it or using it in a few operational processes, with very few deploying it strategically. How does that compare to other areas?
Korst: That is a very common point that our survey corroborates. Recent studies from McKinsey, BCG, and Deloitte all show less than 10 percent. Approximately 5 percent of organizations have scaled any AI within their broader organizations.
Where are the rest of the organizations [in the] 90 percent? The difference [between] excited executives to full-scale deployment is in the pilot process. Many pilots won't work, which is the nature of innovation. Finding the right use case and workflow is important, but human capital change management also is exceedingly important. Organizations must understand their strategy and how to equip people with the right skills and workflows, rather than just bringing technology to an existing workflow.
BTN: For companies that have moved to a more strategic deployment, are there common threads or themes?
Korst: They have a clear vision and a playbook for the way AI should be rolled out in the organization. I hesitate to say "AI strategy" because AI is a tool, not a strategy. It's about taking a strategy and applying AI in authentic ways. You need top-down vision and resources, but bottom-up enablement. Organizations [succeed] by embracing, encouraging and enabling the mid-management layer and their teams to experiment and fail and find ways that they can win and rethink their workflows. Most managers already spend less than 30 percent of their time on strategic work, so you have to give them space to do that.
Also, in my experience it's a lot easier to rally the troops when I talk about how we're going to use this to do new things for our customers, enter new markets, find new customers and have new products versus the CEOs who have been trying to talk to Wall Street about all the cost savings they're going to get from AI.
BTN: To go back to ROI, that was a sharp difference in our data. The Wharton survey showed 72 percent of respondents measuring ROI, while only 21 percent of travel managers in our survey said they were formally measuring ROI in their travel program. Is that discrepancy typical or is there something deeper here?
Korst: My hypothesis is that the levels within the organization are skewing that. The higher you are in the organization, the more likely you are to say you are measuring and seeing positive ROI.
BTN: Regarding internal investment, you found increasing investment through 2025. Is that a rising tide lifting all boats, or is there a difference in how that investment is partitioned between support roles like travel management and customer-facing or revenue-generating roles?
Korst: I don't have the data on that, but I suspect it depends on whether the company's focus is cost efficiency or strategic growth. For support organizations, they are likely looking for tools to increase overall productivity.
BTN: Respondents in our survey were very strong in saying AI is an enhancement, not a replacement for a human travel manager. How realistic is that? How safe are pure knowledge worker roles going forward?
Korst: We're very confident that those who know how to use the technology will replace those who don't. I expect agentic technology to play a very large role in transforming some of these jobs in the coming years in the travel industry. The net impact on the number of jobs depends on whether the technology causes more growth. If [AI] helps make it easier for people to travel more for business, then the net impact [would be] less, if at all.
I am dubious that AI is replacing many jobs today. But a year or two from now? It may be capable enough to replace some full roles. However, look at Walmart—their CEO said almost every role will be changed by AI, but he expects to have more people, not fewer, because they will use it for growth. The jury is out, and we'll have to watch the net impact.