The One Question That Reveals Your Next AI Project
This is part sixteen of our Putting AI to Work in Your Tour Business series. Watch the full series here and read part one here.
Most tour operators pick their first AI project the same way. Whatever annoyed them most, this morning or last week. A slow inbox, a clunky spreadsheet, a review that took forever to answer. Three or four weeks later, they’ve got a clever little tool, maybe a bit of time saved, and a growing suspicion that AI is overhyped.
Picking the wrong target caused that outcome.
“The problem wasn’t necessarily the AI, it was choosing or prioritizing what to use the AI on.”
The Trap of Reactive Prioritization
If you’ve already built your AI Brain, the folder of everything your AI needs to know about your business, you already know the next problem. Everything suddenly looks automatable. Every task on your list looks like a candidate. And if you haven’t built it yet, that’s still step one, but the wall shows up either way. Most operators respond by grabbing whatever annoyed them most recently. That’s the gap this framework closes.
Start With What Breaks
Before scoring anything, sit with one question: If your bookings doubled tomorrow, what breaks first?
“Not what’s annoying or what takes a lot of time, but what would actually break.”
For most operators, the honest answer sounds like: I wouldn’t be able to respond to inquiries fast enough, I don’t have enough guides, I already have reviews piling up unanswered. Whatever your answer is, that’s your constraint, the wall between you and growth.
The best first AI project isn’t the task that annoys you most. It’s the task standing closest to that wall. Fix the constraint and you don’t just save time, you free up capacity across the whole business.
Watch for This Trap Before You Score Anything
A lot of AI spending in general goes toward sales and marketing, the flashy stuff. But most of a team’s actual hours sit in the unglamorous back office work: inquiry handling, scheduling, reporting, follow-up, reconciling numbers. That’s where a lot of the payoff hides. Tour operators fall into the same trap. Everyone wants the AI marketing machine, and the money is often sitting in operations instead.
The Four-Question Scoring Framework
Once you’ve got a list of candidate projects, score each one from one to five on four questions. Add up the total. The higher the score, the closer that project should sit to the top of your list.
How often does this happen?
Daily tasks score a five. Something you touch twice a year scores a one. Frequency is where the compounding lives: a task you do every day pays you back every day. A task you touch twice a year barely moves the needle, no matter how frustrating it feels in the moment.
Could you write down how you do it?
If you can explain your steps to a new hire, you can teach them to an AI. Score a five. If every case depends on judgment only you or your leadership team can make, score it lower. Get clear on your own logic first, and the automation gets a lot easier.
What are the stakes if it gets this wrong?
Low stakes mean more autonomy for the AI. High stakes mean a human needs to stay in the loop. If an AI drafts one guest email with a small error, that’s still worth a higher score, because the downside is contained. If it’s managing your ad budget or emailing your entire list, the downside is bigger, and that pulls the score down.
Does the payoff beat what it costs to build?
You don’t need a spreadsheet. Just a rough number: how many hours this saves, multiplied by how often it runs. Two hours back a week is a real prize. Two hours back once a quarter usually isn’t worth the build. This question stops you from spending a weekend automating something that was never going to pay you back. Once you’ve picked a winner, the next step is turning it into a repeatable AI skill.
Running Real Projects Through the Framework
Take responding to online reviews. It happens after nearly every tour, so frequency scores a five. Most operators already have a pattern for tone and brand voice, so teachability scores high too. A slightly stiff reply is low stakes, since you can edit it before it posts. And the payoff is close to immediate. That’s about as close to a perfect first project as it gets.
Now take inbox management. In season, inquiries come in constantly, another five for frequency. Most of those questions have already been answered a hundred times before, so teachability scores high as well. Stakes sit a little higher, maybe a three or four, since you’re talking to individual guests rather than your whole list. But faster, more consistent responses tend to raise your conversion rate, so the payoff more than covers the build.
Compare that to something like an annual brochure. It happens once a year, so frequency scores low. The template barely changes, so teachability scores fine. But the stakes are high enough that a person needs to review it regardless. Add it up, and it rarely earns a spot in your AI brain as a skill, even though it might still be worth some AI assistance along the way.
What a Skill Is
Once you know which task wins, write down exactly how you do it, step by step. That written process is what gets taught to your AI as a skill, a documented process your AI can run every time you need it, without you re-explaining it from scratch. Most AI tools, including Claude Cowork, have a built-in way to build and save that skill inside your AI brain. The same idea applies outside your AI brain too, using tools like Make.com or n8n, where a change somewhere in your booking software or CRM can trigger the AI layer directly.
What To Do Next
Come up with ten candidates for your next AI project or skill. Score each one against the four questions above, total them, and rank the list. Take the winner and write down exactly how you do it today. Build that one first. If you want a second set of eyes on your list, our free strategy call is a good place to start.
