Picture your team's next brainstorm. Someone opens ChatGPT, types a prompt, and thirty seconds later you've got twenty new campaign ideas, a dozen possible visitor personas, and a fresh angle for the shoulder-season push. It feels like progress. It moves fast.
But look closer at those twenty ideas. How many of them made someone at the table say, "wait, why haven't we done that already"? Or did they mostly sound like a slightly nicer version of what you were already planning?
That gap is exactly what a group of marketing researchers from Harvard, Wharton, Columbia, and Northwestern set out to explain in a recent Harvard Business Review piece, "Research: The Innovation Problems AI Can't Solve." Their finding is worth sitting with: some teams using AI report a genuine creative renaissance, while others get a flood of homogenized, forgettable ideas that all sound like they came from the same person. The difference isn't which AI tool they used. It's whether they understood what was actually slowing their innovation down in the first place.
That distinction matters just as much for a five-person visitor bureau as it does for a Fortune 500 product team. So let's translate it into destination terms.
The researchers point to four stages in the innovation process where AI tends to create new blind spots rather than closing old ones. Here's what each one looks like from a tourism office chair.
Ideation. Ask an AI model for new experience ideas, and it will reliably hand you back the statistically likely answer: the "hidden gem" angle, the "bucket list" itinerary, the golden-hour photo op. That's not a flaw in the technology. Large language models are built to predict the familiar. Left unchecked, they nudge your team toward safer, more conventional ideas right at the moment you're trying to stand out from every other destination running the same playbook.
Screening. Once the ideas are on the table, someone has to choose which ones get funded. A polished AI-generated pitch deck can look more convincing than a scrappier one, even when the underlying idea is weaker. And if you're using AI to help rank or recommend which concepts to pursue, it's often quietly reflecting the assumptions already baked into your organization: the market segments you've always targeted, the partners you've always worked with.
Visitor research. Simulated visitor personas can test a concept in minutes instead of weeks. That speed is genuinely useful. But real visitors are wonderfully irrational. They take the detour because a local mentioned a great taco stand. They booked the tour because a front-desk staffer's enthusiasm was contagious. A simulated persona won't tell you that — only a real conversation will.
Post-launch learning. AI can summarize a thousand survey responses into a tidy paragraph in seconds. The risk is that the summary confirms what you already believed rather than surfacing the complaint you didn't want to hear. Fast synthesis isn't the same thing as honest learning.
The researchers' core recommendation is a simple diagnostic question, and it's one every tourism leader can borrow: before deploying AI on a problem, figure out whether the real bottleneck is informational (you genuinely don't have enough data), judgment-based (you have the data but struggle to decide well), or incentive-driven (your team is rewarded for looking decisive or busy rather than for being right).
Those three bottlenecks call for different fixes, and AI only solves one of them cleanly.
If your bottleneck is informational — you don't know what first-time visitors actually want to do on a rainy Tuesday — AI-assisted research can genuinely help you gather more signal, faster.
If your bottleneck is judgment — your team has plenty of visitor data but keeps greenlighting the safe, familiar campaign anyway — no amount of AI-generated options will fix that. That's a decision-making and courage problem, and it needs a different kind of intervention: a clearer decision framework, a devil's advocate role in the room, leadership that rewards the ideas nobody else would have proposed.
If your bottleneck is incentive-driven — staff get more credit for shipping something this quarter than for shipping the right thing — AI will just help you produce mediocre work faster and with more confidence.
Here's where this connects to something Learn Tourism talks about constantly: your frontline staff, your volunteers, your tourism ambassador programs, and your residents are not just delivery mechanisms for a strategy someone else designed. They're sensors. They're the ones standing at the welcome center desk, riding the trolley, working the front desk when a family asks a question no persona ever anticipated. That direct contact is exactly the "grounding" the researchers point to — the thing that keeps a destination's decisions tethered to reality instead of to whatever a model predicts is statistically likely.
If your team is using AI to move faster on itineraries, campaigns, or visitor research, that's not a problem. The problem is assuming speed alone means you've solved something. Before your next AI-assisted idea goes live, run it past one real conversation: a frontline staff huddle, a visitor center chat, an ambassador debrief. Ask what a person noticed the tool lacked.
Pick one innovation project on your desk right now — a new tour package, a redesigned welcome experience, a campaign refresh. Ask your team honestly: is our bottleneck informational, judgment-based, or incentive-driven? Write the answer down before you open any AI tool. Then decide, on purpose, whether AI is the right fix for that specific bottleneck or just a faster way to avoid facing it.
That five-minute exercise will tell you more about your innovation process than any prompt ever will.
If your team wants to work through this diagnostic together — as a live workshop with your actual projects on the table — that's exactly the kind of session we love building at Learn Tourism. Reach out and let's talk about what that could look like for your destination.