Episode 30
AI That Actually Knows Your Tour Business, with Kelsey Tonner
AI isn't giving you generic answers because it's limited. It's because it doesn't know your business.
Learn how to build an AI that actually knows how your business runs.
In this episode of the Tour Operator Growth Podcast, Nikki is joined by Kelsey Tonner, founder of Guest Focus, who has coached tour operators since 2015 and now leads a team of more than 20 coaches. This isn't a tool roundup or a list of prompts to try. It's a map of how AI use actually progresses inside a tour business.
Kelsey walks through three levels: uploading files into a single chat, which works until that thread degrades and starts missing the mark; projects built on a real business profile and ideal guest avatar; and a full AI brain, hundreds of text files on your own computer that remember your business, self-correct when you give feedback, and move with you if a better model comes along. He covers why built-in memory features fall short, how connectors turn AI into the central place you get work done, practical ways to keep token costs down, an honest look at privacy, and the tasks operators already run overnight. He closes with exactly where to start if you're overwhelmed.
Grab Kelsey's free resources, including the AI business profile builder, the customer avatar builder, and the AI brain building prompt and workbook:
https://guestfocus.com/using-ai-business-resmark/
Frequently Asked Questions
-
Why does AI give me generic answers about my business?
Because it doesn't have enough context about you. It doesn't know where you operate, what you sell, who your guests are, what tools you run on, or who your competitors are. Kelsey calls this context engineering, and it's the single biggest reason results feel off.
-
Why do my best AI chat threads get worse over time?
Long threads hit the limits of the context window and start compressing what they remember, so quality quietly deteriorates and you find yourself repeating things you already uploaded. Reusing one long thread is also expensive, because the model rereads the entire conversation every time you hit enter.
-
Does AI memory solve the context problem?
Only partly. Memory features are useful, but you can't see or control much of what gets remembered, and it blends everything together. If you use one account for personal goals and multiple businesses, the profile it builds of you gets confused fast.
-
What should I put in an AI project?
Two documents belong in almost every project: a business profile and an ideal target guest profile. Built properly, each runs 12 to 15 pages. Add whatever else the workflow needs, like your policies, FAQ, or 50 sample responses in your voice, then start a fresh thread every time you use it.
-
What is an AI brain?
Think of it as a company wiki that improves itself. Instead of static context re-uploaded per project, it's a growing set of text files on your own computer that hold everything about your business, expand as conditions change, and correct themselves when you give feedback. Because it's plain markdown, you own it and can move it between models.
-
Which tools should I connect to AI?
As many as you can. Email, calendar, and cloud storage are native connections. Booking software like Resmark now connects through MCP servers or API connectors, and Zapier can bridge smaller or industry-specific tools. Google Analytics and your CRM are two of the highest-value connections most operators skip.
-
How do I keep AI token costs down?
Keep your instructions file short and use an index so the AI only pulls the context a task actually needs. Write longer, more considered prompts instead of dozens of quick back-and-forth replies. Start new threads rather than extending old ones. And schedule heavy tasks for off-peak hours or overnight.
-
Is it safe to put sensitive business information into AI?
Everyone lands in a different place, but Kelsey's take is practical: read the terms, turn off data training in your paid plan settings, use strong unique passwords and two-factor authentication, and recognize that most of your sensitive data already lives in the cloud. Withholding financial and personal context has a real cost in how useful your AI can be.