How Stephanie Bowers’ approach, shaped by years of diplomatic advance work. Reveals the gap between a generated itinerary and a trip that actually holds
Stephanie Bowers spent a year putting together a 60th birthday celebration in Japan.
The centerpiece was a private Buddhist temple in Kyoto — not a guided tour of a temple, not a reserved seat at a ceremony, but exclusive access to a specific site not normally open to the public. Alongside that: the former executive chef of Nobu, brought in to prepare the meal, and a particular tuna sourced from a single vendor at Tokyo’s Toyosu fish market who then coordinated its transport to the venue. Getting from idea to booking required a chain of introductions that stretched from a prominent Japan-based correspondent Bowers reached through a colleague, to a long-term American expat living in Kyoto, to the monk, to the market vendor. Each connection unlocked the next.
None of it was searchable. None of it was bookable through any public channel. And none of it came together until close to the event date. “A lot of the pieces, frankly, didn’t come together until right before,” Bowers explains.
They did, however, come together, because Bowers had built enough structural redundancy into the plan that late confirmations didn’t cascade into failures. The family arrived. The temple was reserved. The chef was there. The fish was there.
Now ask an AI to replicate that trip.
Usage of AI tools for trip planning doubled in nine months between October 2024 and July 2025, from 11% of all travelers to 24%. A separate survey found that 33% of travelers plan to use AI for travel in 2026, primarily for destination research (75%) and itinerary suggestions (70%). The pitch is real: a coherent multi-week itinerary, tailored to stated preferences, delivered in seconds. For a long weekend in Lisbon or a first trip to Tokyo, it’s a reasonable starting point.
For a multigenerational journey to Japan costing more than half a million dollars — one that includes a private Buddhist temple and a chef whose name you can only obtain through three degrees of introduction — it is not a starting point. It’s a liability dressed up as research.
The problem isn’t that AI produces bad itineraries. It’s that AI produces good-looking ones — organized, logical, full of recognizable names — and the gap between a good-looking itinerary and a trip that actually works is precisely where ultra-luxury travel lives.
What AI Recommends (And Why That’s the Problem)
AI trip planning draws from what exists on the internet. That means it defaults to properties and experiences that have been written about extensively: the hotels with aggressive PR, the restaurants with thousands of reviews, the tour operators who invested in digital visibility. Simon Cameron, CEO of Lightfoot Travel, put it plainly in January 2026: AI “will not unlock doors to premium, behind-the-scenes experiences” and cannot step in when last-minute changes are required.
The properties ultra-high-net-worth travelers actually want — the sole-use safari lodge that will run entirely on the family’s schedule, the 12-room inn in Kyoto with a waiting list that operates on referral, the African lodge that doesn’t list availability on any platform because it doesn’t need to — are often exactly the ones that don’t appear in AI outputs. Their discretion is a feature, not an oversight. For the travelers who seek them, visibility is the opposite of what they want.
Stephanie Bowers works through Singular Sojourns, affiliated with Virtuoso and with direct partnerships at brands such as Four Seasons, Mandarin Oriental, Rosewood, and Belmond. Those relationships aren’t decorative. They’re how a general manager at a property gets a message about an arriving client before the client’s flight lands. They’re how a room block gets held for a multigenerational group traveling across time zones when the official confirmation window has technically closed. Some clients travel with private security. Others are discreetly well-known, and when hotel staff recognize their names the effect tends to be either inflated pricing or unwanted visibility. Managing that requires trust built over years, not a booking platform.
AI can compare published rates and inclusions across a dozen luxury hotels. It cannot text the GM.
The Redundancy Clients Never See
Stephanie Bowers’ background is unusual among luxury travel advisors. Before founding Singular Sojourns, she held senior director-level positions across the National Security Council and the State Department’s Executive Secretariat Operations Center. In those roles, she supported overseas travel for the President, Vice President, Secretary of State, and First Lady at the level where failure would be seen by the entire world and there is no graceful recovery.
Presidential advance teams arrive at a location weeks before the principal. They walk every route, identify the alternates, confirm which elevators work, and install communications infrastructure from scratch. Every contingency gets a plan before it becomes one. The scale is not comparable to private luxury travel. Private travel and presidential travel are not the same thing. But the underlying discipline is identical: solve problems before they happen, build redundancy into every layer, and never leave anything to in-the-moment judgment that could have been resolved in advance.
“I am very, very used to putting together extremely complicated itineraries that have to be perfect and down to the minute,” Bowers says. “Attention to detail is super important: understanding how long it realistically takes to have a meal, how long it realistically takes to see a museum, how long someone actually realistically needs for downtime.”
That last point matters more than it sounds. Most itineraries that look airtight on paper fall behind by noon on the first day. The distance between a scheduled experience and a realistic one is where plans fail. Bowers’ calibration for that gap was built in environments where falling 90 minutes behind was not recoverable.
The Japan birthday trip illustrated how that discipline works under pressure. “A lot of the pieces, frankly, didn’t come together until right before,” Bowers says. She had built enough redundancy into the structure that when individual elements confirmed late, nothing collapsed. The clients arrived. The temple was reserved. They never knew it had been anything other than certain.
When something goes wrong on a trip she’s planned — a driver who doesn’t show, a restaurant that loses a reservation, a weather event that closes the preferred route — there’s already a resolution pathway that activates without the client knowing there was ever a problem. That’s not customer service. That’s architecture.
AI doesn’t build redundancy. It builds an itinerary.
Knowing What Clients Actually Want
There’s a layer of ultra-luxury travel planning that has nothing to do with relationships or logistics and everything to do with information quality. Bowers has described it as one of the consistent problems in working with high-net-worth clients: the person who communicates on behalf of a wealthy traveler is often not the traveler.
“Just like senior government leaders, there’s often junior staff around them that make assumptions on their behalf that are actually inaccurate,” she explains. “Actually getting to know the clients personally is really important to me: understanding their true preferences and not what people project as their preferences.”
For a multigenerational family trip, the problem compounds. A son tells the advisor his father “loves wine.” Does that mean Burgundy Grand Cru vineyard access or a good bottle at dinner on a terrace? Completely different trips. A daughter says her mother wants to relax. Does that mean a private beach, a spa schedule, or five days with nothing on the calendar? You need to ask the mother.
An AI trip planner processes stated preferences and generates corresponding suggestions. It has no mechanism for distinguishing between what a client wants and what someone else wants on their behalf. It cannot follow up with the person who actually matters. It takes the input at face value.
For a weekend getaway, this produces a minor mismatch. For a journey anchored to a milestone — a 75th birthday, a 50th anniversary, a year when a grandparent still has their health — the stakes for that mismatch are not minor.
What AI Does Well (And Where That Ends)
None of this argues against AI in travel. Advisors across the industry are using it to draft marketing emails, generate first-pass itinerary frameworks, and organize research that once took an hour into something manageable in ten minutes. The Luxury Travel Report found that AI reduces email response time from roughly 40 minutes to five to ten minutes for routine correspondence. Advisors describe it handling “80–90% of the work” on standard administrative tasks.
That’s a real efficiency gain, and it’s where AI is strongest: the part of travel planning that is essentially information organization. Research synthesis, first-draft communications, comparison tables across published options.
What it cannot do is build the relationship that gets the temple reserved. It cannot apply years of firsthand destination knowledge — Bowers has traveled to all seven continents, and has been posted to Madagascar, South Africa, Spain, Iraq, and The Bahamas — to know which property has recently changed management and which hasn’t, which route looks shorter on a map but adds two hours in practice, and which confirmation in Japan you should build a structural backup for because it won’t arrive until three days before the event.
The traveler using an AI-generated itinerary doesn’t know what they’re not getting. The gaps are invisible by definition. You don’t know the temple existed if no one with access told you. You don’t know the lodge had availability if your inquiry never reached the right person. You don’t know there was a problem on day four of your trip if someone already solved it before you arrived.
That invisibility is part of what Bowers is selling, and it’s the part that no algorithm can replicate. The question for travelers at this level isn’t whether AI has gotten good at travel planning. It has, within a specific and bounded domain. The question is whether the things it can’t do are the things you can afford to find out about the hard way.
For most trips, the answer is yes. For a sole-use safari lodge with a 90-year-old grandmother and four generations who haven’t traveled together before, or a year-in-the-planning 60th birthday in Kyoto with a chef, a monk, and a fish market vendor — it isn’t.
