Every independent hotel is already running an AI product.
It is just made of sticky notes, Gmail stars, highlighter on printed arrivals lists, a group text between the GM and the housekeeping lead, one encyclopedic night auditor who remembers everyone’s dog, and a shared Dropbox folder nobody under 30 has the password to.
We spent most of this past quarter sitting on the wrong side of the front desk at twelve independent hotels, three-star urbans, a coastal resort, two converted-warehouse boutiques, a mountain lodge, several family-owned small chains. We came to understand why personalization at scale is failing at the independent level. What we found was more interesting than that.
These are the field notes.
1. The GM’s most important system is her sticky-note monitor
At four of the twelve properties, the GM’s primary operational dashboard is a square foot of sticky notes stuck to the bezel of her monitor.
Yellow ones are VIP arrivals. Pink ones are complaints in progress. Green ones are repair items waiting on the engineer. A single blue sticky with three exclamation points means the Thompson anniversary suite needs the rose petals, the sparkling (non-alcoholic for her, the Piper Heidsieck for him), and a handwritten note from the owner.
The GM does not call these sticky notes. She calls them “the system.” When we asked one GM why the PMS’s VIP-flag feature was not doing this work, she looked at us with a particular mix of patience and pity and said: “Because the PMS loses them.”
This is the most clarifying sentence we heard across twelve hotels: because the PMS loses them.
2. The night auditor is the CRM
At three properties, the returning-guest memory lives inside one specific human being. At one, he is named Thomas. He has worked the 11pm-to-7am shift for nine years. He remembers that the couple in 408 lost their dog last spring. He remembers that the man in 214 pays in cash, always, and wants his receipt printed, always. He remembers that the Korean family in 606 prefers the shower in 608.
Thomas has never logged into the CRM. He does not know what a “guest profile” is. He is the system the CRM was supposed to be.
What happens when Thomas retires? Nobody in the building can answer that question without looking uncomfortable.
3. The Moleskine in the top drawer
One GM keeps a leather-covered Moleskine notebook in the top-right drawer of her desk. It contains every guest dietary restriction, allergy, preference, and anniversary that has ever been communicated to her personally by a returning guest.
She started it in 2019. It is two-thirds full.
It is also, quite literally, the hotel’s most valuable piece of data infrastructure. It is not backed up. It is not GDPR-compliant. It is handwritten. She knows this. She has been asking for a better system for six years.
4. Gmail stars are a taxonomy
At five of the twelve properties, the reservations inbox runs on Gmail stars as a de facto ticket system.
- Yellow star = VIP booking, needs owner attention
- Red star = complaint or refund in progress
- Blue star = group booking in negotiation
- Green star = confirmed, needs pre-arrival email
- Purple star = anniversary / special occasion
The GM did not invent this system. The previous GM did, and trained the current one. The one before that used a different color code. Nobody has ever written this down.
When one of the five GMs we shadowed onboards a new reservations coordinator, she does not teach the star codes with a document. She teaches them by sitting next to the new person for a week and saying “that’s a yellow, the Goldbergs always get the corner room.” It takes two weeks before the trainee is trusted to handle the inbox alone.
This is not a failure of management. This is what works. It is what has been selected for across a decade of operational iteration. The problem is that it does not scale, does not survive turnover, and does not transfer to the night auditor.
5. “Spreadsheet” is a personality type
At every single property, we found a revenue manager who maintains what she calls “the master spreadsheet.”
The master spreadsheet is not the PMS report. It is not the RMS output. It is not the CRS channel mix. It is her own, private, lovingly-versioned Google Sheet that reconciles all three, computes commissions manually, flags rate parity violations, and tracks year-over-year at a room-type level.
Most GMs do not know the master spreadsheet exists until they walk past her desk and catch a glimpse. One GM, when we showed her the spreadsheet her revenue manager had been maintaining for three years, said quietly: “This is the only honest number in the hotel.”
6. The WhatsApp group is the operations layer
At nine of the twelve properties, the real-time coordination layer, the one that actually runs the hotel minute-to-minute, is a WhatsApp group.
The groups vary: “Front of house.” “Ops team.” “Amy’s hotel 2024.” They contain the GM, the assistant GM, the maintenance lead, the head of housekeeping, and usually the owner. They move roughly 40–200 messages a day. They include photos of broken things, audio clips of irate guests, videos of late-arriving shipments, and occasional celebratory emojis.
The WhatsApp group is the actual operations platform. The PMS is where you enter transactions after the operations happen. This inversion, which operators experience as obvious and vendors experience as counterintuitive, is the single biggest gap between how the hospitality software industry thinks the back office works and how it actually works.
7. Everyone has a lie they tell about channel mix
At three of the twelve properties, the GM gave us one number for direct-booking share before the engagement and a different, smaller number after we had sat with the revenue manager’s spreadsheet for a week.
The gap was consistently 5–10 percentage points. Always in the optimistic direction. No one was lying. They were quoting the number they most believed, or most wished was true, or the one from the last board deck. It just happened to not be the number the spreadsheet produced.
This matters because Phocuswright’s 2025 data shows OTA share at independents has hit 61% and is still climbing. The average independent GM, in our sample, thought their direct share was meaningfully higher than it was. The drift between perception and reality is itself an operational risk.
8. The 2.3-profiles problem is real, and nobody mentions it
We ran a small exercise at each property: pick one return guest; count the number of profiles in the PMS. Every single property had multiple profiles for at least one returning guest we checked. At one property, a couple who had stayed six times across three years had five separate profiles, with the allergy note on one, the loyalty email on another, the correct billing address on a third, and a pet-friendly flag on a fourth that was last updated in 2022.
This is not anecdotal. dailypoint’s analysis of 4.5 million stays found the average returning guest has 2.3 profiles inside the PMS alone. Our sample of twelve properties was consistent with that number.
What nobody mentions, in any PMS vendor pitch: this is a structural feature of how the PMS ingests bookings from multiple channels, not a training problem. The receptionist is not the cause. The ingest is the cause.
9. The labor shortage has a face, and it is exhausted
By the end of the quarter, we stopped asking “how are you staffed?” and started asking “how many double-shifts has the front desk covered this month?” The average answer, across twelve properties, was 11. The high was 23.
AHLA’s January 2025 survey puts 65% of U.S. hotels at staffing shortages. The BLS JOLTS data puts leisure and hospitality monthly separations at roughly 5.8%, an annualized turnover rate north of 70%. Industry benchmarks from ROAR put the replacement cost of a single hospitality hire at $3,000+.
What the data does not capture: by the time we arrived for week two at one urban property, the receptionist we had met on day three was gone. Two weeks at a time is the current tenure distribution at the bottom of the labor market.
10. The operational information is locked inside people
BCG and NYU SPS, in their March 2026 AI-First Hotels report, made one quantitative observation that tracks with everything we saw:
Nearly 50% of hoteliers report difficulty accessing critical operational information.
BCG + NYU SPS, AI-First Hotels, March 2026
This is the single most important sentence in the report, and it understates the problem. The critical operational information is not “hard to access.” It is not in the system at all. It is in Thomas’s head, in the Moleskine, on the sticky note, in the WhatsApp group, in the Gmail stars, in the master spreadsheet nobody else has a copy of.
The J.D. Power 2024 North America Third-Party Hotel Management benchmark found staff service to be one of the top drivers of guest satisfaction, and noted that past a certain staffing threshold, scores drop ~50 points. Every property we visited was at, or past, that threshold. Every one of them was running on Thomas’s memory.
What we took away
The hospitality industry is not short of software. It is short of memory, a consistent, unified, always-on layer that captures the operational knowledge currently living in heads, drawers, and WhatsApp threads, and makes it available at the moment of guest contact.
We did not design FlowStay in a conference room. We designed it on the wrong side of twelve front desks, watching the same gap open and close, twelve nights in a row.
The product we’re building is not, fundamentally, about AI. It is about respect for what is already working. The sticky notes work. The Moleskine works. Thomas works. They should not have to keep working that hard.
The better question is: what would it look like if the hotel’s memory did not retire when Thomas retires?
That is the question we are answering.