Yes, dreaming up your next getaway can be exhilarating. But booking it? Often less so. For many customers, the thrill of travel risks getting lost in logistics, says McKinsey Senior Partner Jules Seeley. On this episode of The McKinsey Podcast, he joins Global Editorial Director Lucia Rahilly to discuss new research on the rise of AI in the travel industry, including the advent of AI agents that could design end-to-end itineraries, rebook disrupted flights, and tailor recommendations to each traveler’s tastes—changing the calculus for customers and travel employees alike.
The McKinsey Podcast is cohosted by Lucia Rahilly and Roberta Fusaro.
The following transcript has been edited for clarity and length.
Travel in the age of AI
Lucia Rahilly: ‘Tis the season for holiday travel, and for some of us, that means confronting a slew of logistical decisions: deciphering carry-on fees, deliberating over cost vs. convenience trade-offs, deciding whether to splurge for a seat upgrade. McKinsey Senior Partner Jules Seeley says these practical to-dos sometimes diminish the delight that travel otherwise inspires.
Jules Seeley: The typical customer says they like the travel research experience. They like to explore and think about where they want to go. But they don’t enjoy the practical side of converting that into a specific trip itinerary.
Lucia Rahilly: Agentic AI just might be the travel assistant customers have been looking for.
Jules Seeley: We’re excited about agentic because over time, it could be a very different way to manage travel that lets customers focus more on the excitement and not as much on the friction and pain.
Lucia Rahilly: So when it comes to travel companies, what’s the state of AI adoption?
Jules Seeley: We’re seeing significant adoption already, particularly in internal experimentation—travel companies using it to help improve their own business.
Lucia Rahilly: And have companies using AI seen heartening results, or is that still TBD?
Jules Seeley: I would say it’s still TBD. There’s a bit of the AI paradox, where we see about 80 percent of companies actively using AI in some form but only 20 percent saying they can see any direct impact, particularly in the P&L [profit and loss] of the business. That reflects the early stage we’re at with AI. We’re confident that over time, we’ll see the impact flow through both to the economics and to the customer experience.
Humans and agents at work—together
Lucia Rahilly: You’ve said that the most significant area of adoption has been internal. Suppose I’m a frontline travel worker. How could AI affect what I do and how I do it?
Jules Seeley: Across a whole set of roles within travel companies, AI helps employees become more efficient by doing more of their manual work and freeing up time for the added value they can bring. We think that’s an early path to adoption and one that there’s quite a lot of excitement around.
In addition, there’s the other side of AI: In automating many tasks, it might also change the employee landscape. To that point, our research indicates that skill sets might need to change, but that over time, these evolutions will help organizations because employees will be freed up from managing systems and more able to deliver meaningful service.
Want to subscribe to The McKinsey Podcast?
Lucia Rahilly: Let’s say agentic becomes capable of managing some processes from end to end. What kind of human-only skills will remain vital to the travel experience?
Jules Seeley: Travel is a very personal and emotional area; customers are looking for an experience that inspires and excites them. Even when they have to travel—for work, for example—people look for what we call the magic of the travel experience. That means there’s always a role for people and physical touchpoints to help make sure customers are getting the best from that experience.
Lucia Rahilly: Walk us through the kinds of operational processes our research suggests might be largely or even fully delegated to agentic AI within the next few years.
Jules Seeley: These areas include marketing and how you think about targeting your marketing; some elements of customer service featuring AI chatbots, AI interaction through a website, and even AI digital voice engagement; and broader sales activity, sales origination, and sales operational support. AI can code on behalf of an organization as well. It’s supporting more and more internal activities, automating some parts of them and making them even better and more powerful.
Making headway on the agentic journey
Lucia Rahilly: Suppose I’m a travel leader, and I’m sold on the benefits of investing in agentic. What’s the right starting point?
Jules Seeley: One is internal processes because organizations have visibility and control over them. They can make changes that are less visible to external stakeholders, which gives them a bit of a canvas to experiment on without getting tripped up by external challenges with third parties. It also can drive meaningful cost efficiency opportunities over time, which in lower-margin businesses can be an attractive part of AI deployment. Something has to help fund the technology that goes into making these changes.
One caveat: While those are very valuable capabilities, it’s not clear that being agentic AI enabled in your HR or finance processes or in a call center is going to provide long-term strategic advantage. We’d expect many companies to similarly enable their internal processes with AI.
In addition, we encourage leaders to look at their company and think clearly about what they want to be known for. What’s the magic of their organization? How do they think about AI’s role in that part of the organization? If that’s what gives them a competitive advantage, AI could play a big part in making it as amazing as possible.
Lucia Rahilly: Does the travel workforce have the skills necessary to adopt agentic at scale? Does piloting agentic internally help develop these skills, or do leaders need to put formal upskilling programs in place?
Jules Seeley: In general, I’d say none of us has the deep capabilities today to harness the full power of agentic. Nor do we really know where it’s going. It’s a learning journey for all of us, and we should experiment both at work and in our personal lives. Are you using agentic to explore your own travel journeys? To shop? To research topics or synthesize materials, inside and outside of work? Encourage your team, from top to bottom, to learn about agentic AI and what it can do.
We also encourage formal learning journeys. We often go on these with our clients. We run some executive immersion sessions where for two or three days, we’ll have the entire executive team or board immerse with us and other organizations to really get a flavor for what the technology can do. Then over the next six to 12 months, they explore how those cutting-edge capabilities can become more present in their own organization.
Lucia Rahilly: Learning takes time, and internal experimentation takes capacity. Is our belief that the efficiency benefits of agentic AI are immediate enough to help bridge that capacity gap?
Jules Seeley: Part of that is prioritizing or reprioritizing current activities. Many efforts happening in organizations might now be less important because of agentic opportunities. We encourage organizations to look through their roster and decide whether all their efforts still have the ROI they expected, given the newer agentic capabilities. We also encourage reprioritization to ensure there’s enough capacity and bandwidth, both technical and managerial, for agentic opportunities.
Managing risks—and reluctance
Lucia Rahilly: How should companies introduce agentic to customers who might err on the side of reluctance? Are there some interactions that might enable a wade-in-the-water approach?
Jules Seeley: There are ways to experiment without the commitment hurdle of having an LLM [large language model] plan and book an entire trip. For example, imagine customers starting on a journey of physical purchases, perhaps of smaller commitment levels, and then getting more comfortable with committing larger dollars and with the complexity travel involves.
And while an LLM is never perfect, there will come a point when it’s as good as the customer, who also doesn’t know anything about that particular hotel in that particular region or about the choices of airlines or activities. So travelers will put some trust in the thousands of other travelers who’ve been on that journey and the LLMs’ ability to synthesize that and create something for them.
Lucia Rahilly: Are there governance or oversight mechanisms that should be in place for travel companies, given that customers’ money and personal logistics are on the line?
Jules Seeley: There are a few elements travel companies will have to think about. First, it likely becomes less transparent who the ultimate booker of the travel is. If I’m using my favorite LLM to research a trip to Paris, it’s not clear whether the LLM is using the airline website, the hotel website, or if it’s intermediated through an online travel agency [OTA] or perhaps even buying a package behind the scenes from a package travel provider.
That may create questions for customers, as well as differing risks for providers depending on whether they are the merchant of record, the transacting party, or an intermediary along the way. With that are many elements of payments, like currency and exchange rates, and whether customers’ money is held securely if their trip is refundable. It will take some time for either the norms or, in some places, the regulations and rules to account for this level of complexity.
Lucia Rahilly: Forgive me for bringing catastrophic thinking into the mix, but suppose there were some kind of agent-related imbroglio that affected a lot of customers. Do leaders need to do anything specific to prepare for larger-scale logistical or reputational risk?
Jules Seeley: When agentic means that I, as the customer, have not chosen exactly which pathway is being booked and who my counterparty is, it begets questions: Do the individual companies involved in that research process that ultimate transaction? Are they clearly aware of the risks and issues they have to manage?
There are a lot of questions about the ethical use of AI and how pricing works. There’s a lot of sensitivity around AI and personalized pricing, for example. There’s a lot of risks when it comes to sending intellectual property, as well as personal and financial data, through the chain. What will be really important is making sure risk management processes are fully up to date.
What agentic AI means for brands
Lucia Rahilly: Suppose travelers start interacting primarily through AI agents. Will brands need to start thinking differently about the way they maintain relevance and loyalty? And should leaders be thinking about that now?
Jules Seeley: LLMs could provide an incredible amount of detailed nuance and tailored information without a brand’s stamp. This does raise the question: What is the future value and relevance of a brand, and how do you keep that deeply connected to customers? Our sense is that brands that have a powerful, clear value proposition will continue to have that connection because customers will seek it out. For example, customers may say, “This is the city I want to stay in.” But then they could also say, “and these are the brands I like.” Or the LLM may already be aware of the brands they like. Brands that are weak or lack differentiation are likely to suffer more in the future.
Lucia Rahilly: Is there a way for smaller travel players to get some of the benefits of AI or participate in the AI-enabled ecosystem without being left behind?
Jules Seeley: In an agentic world, where there’s a large availability of information about different experiences, you could imagine an LLM would start to surface those that resonate with like-minded consumers—which could create a higher level of visibility for some boutique brands or boutique experiences that haven’t been as visible in the past.
On the flip side, it’s likely to be harder for those organizations to develop the sophistication to participate as actively as larger organizations can. That’s where large organizations will still have some of their edge—their ability to understand where the customer is and what their expectations are and then to adapt their own businesses accordingly.
Travel in an agentic future
Lucia Rahilly: Say more about a world where there are an increasing number of agents participating in travel through LLMs, but also bespoke agents through individual companies.
Jules Seeley: A hotel brand or an airline or a cruise or an OTA will likely have its own agents you can interact with and develop its own vertical expertise on how to use agentic in travel, in addition to the horizontal expertise of the LLMs. That’s some of the debate going on right now—the extent to which travel will end up as a vertical agent, given depth of experience and complexity, or whether the horizontal LLMs will play a bigger role due to their broader wealth of information about customers.
Lucia Rahilly: If we fast-forward, say, three to five years, what do you think would be most surprising about our experience of travel in an agentic-enabled future?
Jules Seeley: We’ll likely see a few meaningful shifts. First is this overall experience of travel research and travel booking. We’ll likely see more passion and emotion and less frustration around that booking experience.
Second, we’ll see the leading travel players focus on a small subset of areas where agentic capabilities will make the most difference. What are those domains, and how do you completely reengineer them? I’d say we’re still in a world where AI is being applied to steps in a process or steps on a journey. As the capability becomes more mature, we’ll see that entire process or journey getting reengineered from end to end.
Third, there is a lot of research now on voice assist and voice engagement. To date, it hasn’t been as helpful for travel beyond trip research. Once you get past the description of a particular destination, using a voice assistant to list 50 hotel choices is very inconvenient. You still want to display the product. One of the things we may see, though more complicated, is the ability to personalize to the individual customer—based on the data, the insight, the information, and the LLMs’ processing capability—to narrow that list down. Suppose an LLM can find you a hotel, or a choice of three hotels, that really resonates with what you’re looking for in terms of price point, experience, location, and amenities. That would significantly change how you interact with a travel company and become much more like a rich conversation and much less like an index and database search.
Related Articles
