Why Your AI Bill Keeps Climbing, and the Five Habits That Fix It
This is part eleven of our Putting AI to Work in Your Tour Business series. Watch the full series here and read part one here.
If you’ve been building an AI brain with us in this series, there is a good chance you have already hit a wall. The usage bar fills faster than you expected, the bill creeps up, and you’re left staring at the screen wondering what you did to deserve it.
You didn’t do anything wrong. You just haven’t been told how this actually works. So let me pull back the curtain.
A short back and forth chat is more expensive and will go through your usage faster than one long prompt.
This is a Strange Moment in AI Pricing
Let me be honest about something. Right now the big providers, Anthropic, OpenAI, Google, even Grok, are burning investor money to get you onto their platforms. Prices are unbelievably low compared to what it actually costs to run these tools. Look at the investment going into data centers and energy, and you start to see the gap.
For a tour business owner, that is a gift. A hundred dollars a month today gives you access to intelligence that didn’t exist at any price a few years ago. But it’s not unlimited. Usage caps exist, and I expect they’ll keep getting more aggressive, because the cost to run this stuff is real. This is not a scam and it is not a cash grab. It is simple math. So my take is this. Let’s take advantage of the window while it is open.
Start with Tokens
A token is roughly three quarters of a word. The word “tour” is one token. The word “itinerary” breaks into a couple. Punctuation often counts as its own token. So a long document runs into thousands of tokens.
Both sides of the conversation count. The context you upload, the words you use in your instruction, and everything the AI says back all use tokens. Every message you send, every reply it gives, every file it reads, all of it adds up.
Here is the Part That Surprises People
The context, meaning everything already in the conversation, gets reread by the AI every single message. So if you’re 20 messages deep, message 21 doesn’t just send one new question. It sends the whole history back so the AI keeps track of what you have been talking about.
That is counterintuitive. For my first couple of years using ChatGPT, I had no idea this was happening. But that one fact explains almost everything about your bill. Ten quick three-word replies means the AI rereads your entire conversation 10 times over. One long message where you ask for everything at once is a single reread, and far fewer tokens. The rhythm that wastes the most money looks exactly like the way most people use AI right now. Quick question, quick reply, quick follow-up.
Message twenty-one does not just send one new question. It sends the whole history back.
Five Habits That Cut the Waste
Nothing fancy here. Think of it like AI habit hygiene.
- Start fresh more often. If you have moved on to a new topic or a different challenge, open a new chat. Do not let one thread swell for hours. A fresh chat starts with zero context to reread. There is a bonus too. As a conversation grows, the AI has to guess which parts matter most for your latest question, and you start seeing more hallucinations. A clean start gives you cheaper costs and better answers at the same time.
- Pick the right model for the job. We did a whole session on choosing between the powerful reasoning models like Claude Opus, the everyday workhorse models like Sonnet, and the smaller models built for high-volume tasks. One of the smartest moves is to have your most powerful model build the plan and act as your reasoning partner, then take that plan into a new session with a more cost-effective model to do the work. You do not need the top tier for everything.
- Batch your questions. Instead of firing off six tiny messages, write one longer message with everything you need in it. Your AI handles multi-step requests well, and it is great at reading the natural language we use every day. This was a hard habit for me to break, but now I am a lot more thoughtful before I hit enter.
- Learn your reset windows. Most tools now run on usage windows. In Claude that is often a five-hour session. Every time the window kicks over, your usage resets. I have a scheduled task that fires every morning at seven, earlier than I actually need it, just to start my five-hour window. By midday I have my full allowance available. That one timing change made a big difference in how much I can get done in a day. My wife and I share an account, and we lean on idle overnight hours too, when the rate is more favorable.
- Mind your file formats. Not all files cost the same. Screenshots and PDFs are the most expensive, because the AI has to run optical character recognition to read them, which means a lot of image processing and a lot of tokens. A brand kit saved as a PDF can cost you a noticeable chunk every single time you use it. A Word doc sits in the middle, easier to read but still carrying formatting bloat. Markdown and plain text files are the cheapest and fastest. Your AI reads them instantly, no parsing, no overhead.
The brain is a folder of plain text files. When your AI reads it, that is the cheapest, fastest input it will get all day.
Why the AI Brain Pays for Itself
This is exactly why our AI Brain Builder prompt is built with Markdown at its heart. The instruction files, the memory files, the skills, all of them are plain text. Compare that to re-uploading the same four PDFs every session, and you can see why the brain saves you money on usage alone. You get more context, delivered in the cheapest possible way.
It goes further with connectors. Your AI does not load everything at once. It pulls what it needs when it needs it, reaching into your email, your drive, your booking software only when a task calls for it. That selective loading keeps your costs down and your results sharp.
How I Actually Use Mine
For full transparency, I am on the 5x Max plan from Claude, which runs $100 a month at the time of recording. You can go up to two hundred a month for twenty times the usage if you are a real power user. But with these habits, scheduling heavy work for cheaper windows, watching my context, and keeping my brain in Markdown, I get a lot done on the hundred-dollar plan. Some months I have even switched back down when I knew I would not need as much.
The target keeps moving, but these habits will help you make the most of what you are paying for. If your usage has been climbing and you cannot figure out why, start with one of these this week and see what changes.
If you have been frustrated by usage, or you just want to compare notes with someone who has been playing with this, I would like to invite you to book a free call with one of our tour business coaches. We would love to share what we are learning across our community of operators.
Book a free call: https://guestfocus.com/checkin
