Operators keep asking Tourpreneur co-founder Pete Syme to automate their customer communication, and he tells them to stop.
Pete Syme and Mitch Bach open a new monthly series, Pete & Mitch Are Angry, with AI, whose noise has grown too loud for a small operator to sort through without hours to spare. Pete’s reason for saying stop is that talking to guests is an operator’s biggest chance to learn what they want, and that knowledge is where the upsell, the cross-sell, and the second and third booking come from. An operator who hands that seat to a chatbot learns none of it. Supplier work, invoicing, and pricing are where the automation belongs. Requests for a workforce of AI agents get a similar reply: agents built on a broken business break it faster.
AI amplifies whatever business it is pointed at. A solo operator with 12 tours wanted them all automated, and Mitch’s advice was to cut the list to three, aim each at one specific customer, and skip the tech needed to serve 12 tours in 12 locations with different vendors and transport. Pete reaches the same place from the other direction: AI agents on top of broken processes speed up the breakage, so the foundations get fixed first. With sound foundations, the question that matters is what AI cannot do for the business. The answer differs by operator because geography, market, and economics differ, and no consultant can supply it. Automate the repetitive work around that answer, leave the answer itself untouched, and start every AI decision there.
Key Takeaways
- AI rewrites of listings and brochures strip out what makes a tour distinctive. 02:24 to 04:10 OTAs are using AI to turn a listing a guest might connect with into a generic private tour with lunch. Feeds fill with AI-made brochures for tour companies, and the sense of design disappears from them. The first wave of AI travel planners followed the same pattern, reducing itineraries to easy problems and delivering weak results.
- An AI badge on a tool tells an operator nothing until the tool passes an audit. 04:29 to 06:16 Much of the stack is the same set of tools operators used years ago with an AI wrapper attached. The audit question is whether the tool is the old one with AI features added or a new tool built from the ground up that does more than the old one. Any new tool has to deliver benefits that justify the transition cost and the cost of running it.
- Dig into how much an “AI expert” knows about tours before hiring one. 04:29 to 06:16 Tours carry constraints, as every sector does, and an AI expert without deep sector knowledge cannot design around them. The sector filled with AI experts remarkably quickly, and a lot of them lack that knowledge. The builders to trust hold deep domain knowledge and are learning AI alongside everyone else, applying it to what they already know.
- A demo that works for one or two operators is a long way from a tool that works for the industry. 06:25 to 07:43 Inboxes and LinkedIn are full of demos that look amazing. A big gap separates what a demo does from what it takes to run in production. The gap comes from an industry that is complex and fragmented, and clever AI leaves that untouched.
- Planning is part of the trip guests are buying, so automating it removes some of what they enjoy. 06:25 to 07:58 AI travel planners treat trip planning as a problem to solve, and the problem is already solved. Many people like the friction of planning a trip. Repeated psychological studies show that anticipation, meaning thinking about and planning the trip, is a large part of its enjoyment.
- Keep guest conversations human and point AI at the back end. 07:58 to 09:10 Communication with guests is hospitality, and no other activity teaches an operator more about customers. Handing it 100 percent to technology and agents means learning nothing about guests and missing the highest conversion rate. Supplier, invoice, and pricing automation is where AI goes to work, while the conversation stays human because it opens the upsell, the cross-sell, and the second and third booking.
- Automate information, and keep a person at the moments a guest feels seen. 09:10 to 12:26 At 35,000 day tour tickets a year, nobody should hand-send the meeting location, because that message is information. A recent hotel booking produced an email from a concierge who opened with “I’m a real person emailing you.” The email cited the folio notes, noticed the trip was a special one for the booker’s mother-in-law, and offered personalized touches, and it landed with the guest it was written for.
- Design tours by hand, because AI settles on the average of everything written about a place. 09:10 to 12:26 Fed every competitor, Wikipedia, and every other source, AI finds the most average, normal stories about a place and puts them on a shelf, which makes more “cheap cans of soup.” Personal perspective, personal access, and off-the-wall ideas are what AI cannot replace, however hard it is pushed to be weird or original. Those ideas become an operator’s IP, and constantly innovating on them becomes the business moat.
- An operator with broken processes who builds AI agents gets a company that breaks faster. 12:26 to 13:59 A business model and company that are not working at their best get worse when AI speeds them up. The fix comes first: get processes and systems working at their best, then add agents. Agents can lift productivity and scalability substantially, and the foundation decides whether that holds, since a house built on weak foundations falls to the first wind.
- A confusing tour lineup needs a cut before it needs automation, because AI amplifies whatever is already there. 13:59 to 15:01 A solo operator with 12 tours wanted all of them automated, and the underlying problem was a customer confused by 12 choices. Narrowing to three tours aimed at a specific customer removes the need for tech built to serve 12 tours in 12 locations with different transport and vendors. A personal AI assistant goes off and automates all 12 tours without flagging the problem.
- Hold two opposite views of AI at once, and design both with it and against it. 15:01 to 16:39 AI will be good enough in many cases, and prospects already send itineraries an AI planned, a trend that will grow. Operators have to watch where AI changes the market, their customers, and their own business, then design against the negative changes and build tactics that use its strengths. Neither the AI-pilled camp nor the doomer camp holds up, because customers are affected by AI either way and operators have to deal with those customers.
- Write personal AI principles that fit the business, because absolute statements about AI are usually emotional reactions. 16:39 to 19:25 Where an operator lands on the spectrum depends on the business and the person, and probing why something feels like the greatest thing since sliced bread exposes the reaction. AI has sat inside daily life for a decade, in Uber and Spotify, so the field is wider than generative AI and LLMs. Absolute prohibitions deserve the same suspicion as absolute endorsements.
- If an AI-guided version of your tour would be good enough, differentiate on what that version cannot copy. 16:39 to 19:25 Picture a food tour run with self-driving cars and an app that drops guests at restaurants that already know their tastes, with AI commentary on the history of each dish in between. That format alternates information and taste, the same rhythm as a food tour, so if it is as good as a guide, or good enough, the tour has to do something different. Much of the AI being built follows what humans already do, cheaper and easier, and produces no new experiences, which leaves creativity as the ground to lean into.
- AI belongs inside a strategy, and a separate AI strategy bolts something onto the business. 19:25 to 21:13 Treating AI as its own strategy produces a bolt-on that works poorly. The correct sequence is a real strategy first, then the strengths of AI backfilled into it. The effort goes into having the correct strategy.
- The number one question is what AI cannot do for your business, and the answer is the North Star. 19:25 to 21:13 The answer differs for every operator because geographies, markets, and economics differ, so no single answer covers a whole vertical. Build back from that North Star, use AI for the annoying, repetitive work around it, and never touch the North Star itself. No expert or consultant can name it, since only the owner can decide what AI cannot do.
- Talk to an operator who has used a vendor, and check its web traffic, before buying. 21:13 to 21:59 Promises are cheap, and it is easy to spin up a website and overpromise and underdeliver. Ask to speak with an operator, or a couple of them, and hear how the vendor helped. Checking the web traffic of a platform raised in the Tourpreneur Facebook group turned up nine visitors in two months.