Last updated: October 30, 2026
Revenue management software isn’t cheap, and most independent hotels don’t have a dedicated revenue manager to run it anyway. So it’s worth asking the obvious question: with tools like ChatGPT and Claude now writing code, building spreadsheets, and reasoning through logic problems on command, can you just build the thing yourself?
Turns out, yes — mostly. Here’s the full five-step process, start to finish.
Step 1: Teach the AI Your Hotel’s Pricing Logic
Before anything gets automated, the AI needs to understand how your hotel actually prices rooms today — your base rate, your seasonality, and the local events that move demand. Open a new chat and lay it all out in one prompt:
“I run a 24-room boutique hotel in [your city]. My slow season runs January through March and peak is June through September. There’s a local festival the third weekend of July that fills the whole town. Build me a rate-setting framework: a base rate, and clear rules for when to raise it, hold it, or discount, based on how far out we are from the date and how fast rooms are booking.”
Both tools will come back with something genuinely usable — tiered pricing bands, a lead-time multiplier, a note about not discounting festival weekends. Spend a session going back and forth, correcting anything that doesn’t match how you actually think about your market. This step usually takes one focused afternoon.
Step 2: Turn the Logic Into a Working Formula
A framework in a chat window doesn’t do anything on its own. The next step is converting it into something that actually calculates a number. Ask directly:
“Turn that framework into a Google Sheets formula. I want a column for occupancy percentage, day of week, and a manual field for ‘upcoming event,’ and have the formula output a suggested rate for each date.”
You’ll get back a real formula — nested IF statements, usually, sometimes a cleaner approach using lookup tables. Paste it into a sheet, build out the next 90 days of dates down the left column, and fill in your current occupancy by hand for each one. At the end of this step you have a spreadsheet that, given the right inputs, suggests a rate.
Step 3: Build Your Competitor Tracking Routine
Pricing logic means nothing without knowing what the market’s doing. Pick four or five comparable properties in your area — the ones guests actually cross-shop you against — and check their live rates on Booking.com, Expedia, and Google Hotel Ads each morning. Ask the AI to help structure the log:
“Help me build a simple daily tracking sheet for competitor rates. I want to log today’s rate for each property, note any big swings, and flag if more than two of them move the same direction in one day.”
Fifteen minutes a morning, every morning, gets you a real comp-set view feeding directly into the formula from Step 2.
Step 4: Automate the Daily Output
With the formula built and comp data flowing in, the last piece is getting a suggested rate delivered to you automatically instead of opening the spreadsheet by hand. A short script does the job:
“Write me a Google Apps Script that runs every morning at 7am, reads the suggested rate column from my sheet, and emails it to me so I can review and approve it before it goes live.”
Google Apps Script is free, lives inside Google Sheets, and both ChatGPT and Claude can write a working version of this in one or two tries. Set the trigger, test it for a few days, and the suggested rates start landing in your inbox before you’ve had coffee.
Step 5: Review and Push the Approved Rates
The final step is the quality-control layer. Each morning, open the email, scan for anything unusual — a group inquiry that came in overnight, a competitor’s rate that jumped for no obvious reason, a holiday the formula didn’t know to weight — and then log into your PMS and manually update the approved rates property by property.
Five steps, one afternoon to set up the framework, and a 15-to-20-minute daily routine after that. By most measures, that’s a real revenue management process, built for free, in a weekend.

Here’s What Actually Happens by Week Three
The framework holds up. The formula holds up. What quietly stops holding up is you.
Step 3 and Step 5 were never automated — they were just renamed. “Build a tracking routine” was always a person checking rates by hand, every single morning, for as long as the hotel exists. “Review and approve” was always a person doing the actual pricing judgment the spreadsheet couldn’t do. The AI wrote a good formula. It didn’t remove a single hour of daily work — it just gave the daily work a more official-looking home.
And the two steps that look automated, Step 4 especially, aren’t connected to anything real. ChatGPT and Claude have no live link to your PMS, your channel manager, or your booking engine. The script in Step 4 reads a spreadsheet — it doesn’t talk to the systems that actually hold your rates, so every “automated” update still ends with someone typing the number in by hand. The two hours a week this was supposed to save never actually show up.
This isn’t just a hotel-specific problem, either. MIT’s NANDA initiative studied 300 enterprise AI deployments and found that AI tools bought from a specialized vendor succeed about 67% of the time — versus roughly 22% for tools built in-house (MIT NANDA, “The GenAI Divide: State of AI in Business 2025,” via Fortune). Their explanation lines up almost exactly with what happens above three weeks in: “Generic tools like ChatGPT excel for individuals because of their flexibility, but they stall in enterprise use since they don’t learn from or adapt to workflows.”
Create your own RMS using AI — pilot success rate, buying from a vendor vs. building in-house
There’s a sharper problem underneath all five steps, too. Every one of them involves typing real hotel data into a public chatbot — occupancy, rates, sometimes details tied to a specific guest or group. Ireckonu’s CEO put it bluntly: hotels are “experimenting with generative AI without the necessary checks and balances,” and in the worst case, that data can land on external servers in jurisdictions with weaker privacy protections than GDPR requires (Travolution). And when the daily review step in Step 5 eventually gets skipped — a busy morning, a day off, a long weekend — nothing catches the mistake before it goes live. Hospitality Net documented exactly this: an AI tool drafted a guest-recovery email promising a $75 credit nobody had approved, and because no review step existed that day, the hotel was on the hook for a commitment it never signed off on (Hospitality Net).
This Is Exactly What Ramsi Does Instead
Everything the five-step version was reaching for — a real pricing framework, live competitor data, a system that updates itself, something trustworthy enough to not need a daily manual check — is what Ramsi actually is, not what it’s pretending to be. It reads live occupancy and competitor rates directly instead of waiting for someone to check Booking.com every morning. It pushes rate changes straight into the systems you already run instead of emailing you a number to copy by hand. And it runs inside its own guardrails continuously, instead of needing a person to catch its mistakes before they go live. See how that pairs with the PMS you already run →

FAQ
Can you actually build a working RMS with ChatGPT or Claude? You can build a real pricing formula and a working spreadsheet — both tools are genuinely good at that part. What you can’t build is the live connection to your occupancy, competitor rates, and PMS, so two of the five steps stay manual forever, no matter how good the formula is.
Why do homemade AI tools fail more often than AI tools you buy? MIT’s NANDA initiative found AI tools purchased from a specialized vendor succeed about 67% of the time, versus roughly 22% for tools built in-house — mainly because general-purpose AI tools don’t learn from or adapt to one specific business’s workflow the way a purpose-built system does.
Is it safe to type hotel guest data into ChatGPT to get pricing help? No. Hospitality-tech leaders have warned that public AI tools aren’t built to handle sensitive guest information, and that data can end up on external servers in jurisdictions with weaker privacy protections than GDPR requires.
How much time does the DIY version actually save? Less than it looks like on paper. The formula and the automation script remove the math, but the daily competitor check and the daily approval review — the two steps that take the most time — stay manual the entire time.
What’s the real difference between the DIY version and a revenue management system like Ramsi? The DIY version produces a number someone still has to gather data for and manually apply. A revenue management system reads live data continuously and applies the rate itself, with nobody re-entering numbers every day.