Six layers, one hub — the minimum stack for the new PMS era

What an independent 100-room hotel actually needs to compete in the agentic commerce era

The previous article in this series argued the PMS era is ending. The focus has moved from operations to commerce, and the architecture has to move with it. This piece answers the natural next question: what does the new architecture actually look like inside a hotel, and what is the minimum bar for participating in it? For the 100-room independent, the answer is six specialist layers, one PMS acting as the commerce hub, and a clear line between what is required and what is optional.

A note on the names below

The names below are the focused players — founded in the last decade, built for the architecture the previous era did not have, and shipping faster than the powerhouse vendors can keep up with. The powerhouses are not absent from these categories. They are absent from this list.

What is deliberately not on this list: loyalty programs, POS, BI tools, housekeeping, and energy management. They are real systems hotels need. They are not what the commerce hub is. Marketing — metasearch, paid acquisition, content and generative engine optimization — is not included as a layer here either. It is genuinely commerce, but it is its own subject and will be the focus of a follow-up piece.

Layer 1: The PMS as commerce hub

The PMS is what every other layer reads from and writes back to. For the 100-room independent, this means the PMS holds three things nothing else holds: the live state of the property's inventory, the canonical record of every guest and every reservation, and the financial truth of what money has moved. Everything else — pricing decisions, distribution updates, guest messaging, payment processing — is a specialist function that reads from the PMS and writes back to it.

For the 100-room independent, the technical bar this implies is concrete. Specialists read and write through real-time, bidirectional APIs, not nightly batch syncs that leave the property's data stale between updates. The PMS pushes updates out the moment a reservation is created, modified, cancelled, checked in, or checked out — so the rest of the stack reacts immediately rather than working from data that is hours old. AI agents act through the PMS via MCP exposure rather than around it. And the commercial team opens a reporting view every morning, not a BI tool nobody logs into.

The PMS market segments along this line cleanly. Vendors building for the commerce hub model are the ones investing in API depth, MCP exposure, and real-time event coverage. Vendors building for the unified-suite model are the ones acquiring adjacent modules to bring functions in-house. The 100-room independent betting on the commerce hub model is betting on the first kind.

Two names stand out as the canonical examples of that first kind. Apaleo (Munich, founded 2017) is the most explicit commitment to the commerce-hub model — a PMS core with a marketplace of specialists the hotelier picks from, rather than a bundled suite of acquired modules. Customers include citizenM, easyHotel, limehome, Miiro, and Numa. Stayntouch (US-based, cloud-native, mobile-first) offers 1,200+ integrations through its open API and has built no adjacent modules of its own — the marketplace is the way functionality gets added. Customers include Zoku Amsterdam, Pod Hotels, and the TWA Hotel. Both are accessible at the 100-room scale.

Layer 2: Channel manager — agentic-ready distribution

The channel manager is no longer just about pushing inventory to Booking.com, Expedia, and the long tail of OTAs. The 100-room independent now also needs its inventory exposed to AI agents through MCP — readable in real time by ChatGPT, Claude, Gemini, and whatever consumer AI surface emerges next.

Two names have shipped real product against that bar. Lighthouse launched Connect AI on the MCP standard in June 2025 and shipped the first direct booking app for hotels inside ChatGPT in March 2026, via its sister company The Hotels Network — the focused player that moved first. SiteMinder followed in April 2026, with Demand Plus extending hotel discoverability into ChatGPT, Claude, and Gemini through MCP, and Channels Plus opening the same inventory to AI-native intermediaries on the same protocol.

For the 100-room independent, the practical question is which channel manager fits the property's distribution mix — and which has shipped real product, not just a press release. The names above have shipped product. Most of the category has not.

Layer 3: Revenue management — ML-based, total-profit-oriented

For an independent at 100 rooms, the RMS is not assisting a revenue manager — it is the revenue manager. The commercial team is too small to run pricing analytics by hand, and the property is too big to leave pricing to gut feel. The RMS has to make granular pricing decisions continuously, across room types and stay patterns and segments, and adapt as the demand environment shifts week by week.

The buyer side adds urgency. As AI agents take a growing share of the shopping behavior the RMS has to price against, the gap between machine learning and rule-based pricing widens — the ML-based system reads the structured signals an agent buyer is reading; the rule-based system does not. For the 100-room independent, the choice of RMS is increasingly the choice of who is doing the pricing work the property cannot do internally.

Three focused players are worth knowing in this category — all of them specialists in revenue management only. FLYR Hospitality (formerly Pace, ML-first from inception, customers including Ennismore) and BEONx (Spanish, explicit RevPAG positioning, customers including Barceló) were built around machine learning from the start. RoomPriceGenie (Swiss, founded 2017) is built specifically for independent hotels without revenue management teams.

Layer 4: Guest CRM — vector architecture, agentic-ready

The 100-room independent's guest CRM has two jobs: convert a one-time stay into a direct-relationship guest, and respond to the increasingly conversational, increasingly AI-mediated way guests interact with the property. Email automation handled the first job well enough for the past decade. It does not handle the second at all.

The replacement architecture is vector-based and agentic. The focused players in this category are unusually international: Bookboost (Sweden, AI Agent launched March 2026, trained on five years of real hospitality guest interactions), Duve (Israel, dynamic personalization that pulls live PMS data, guest profile data, and reservation context into every response), HiJiffy (Portugal, multilingual AI chatbot built specifically for hotels), and Canary Technologies (US, AI Voice deployed globally with Wyndham, omnichannel guest service across calls, text, and web).

Each takes a different angle on the same operational problem — handling guest interactions in natural language at the volume and velocity the modern guest now expects. Akia (San Francisco, ex-Facebook engineers) and Runnr.ai (Utrecht, WhatsApp-first, founded by ex-Booking.com employees) are credible adjacent options for properties whose guest mix or channel mix fits their angle better.

The 100-room independent picks the vendor whose product surface fits its operating model. The wrong choice in this layer is the guest CRM module bundled with the PMS suite — built for the email-automation era, not the conversational one.

Layer 5: Payments — integrated, not owned

The 100-room independent does not need the PMS vendor's bundled payment gateway. It needs a direct relationship with a payment processor that is being built for the new agentic rails — Stripe's Machine Payments Protocol, Visa's Trusted Agent Protocol, Mastercard's Agent Pay, AWS Bedrock AgentCore Payments.

This is the layer where the right answer is a powerhouse — specifically, the powerhouse that is building the agentic protocol layer rather than competing against it. Stripe and Adyen are not focused players in the sense this article uses the term. They are the foundational infrastructure the focused players themselves are building on. A 100-room independent on Stripe or Adyen inherits the protocol work being done at the foundation. The same property on a hospitality-specific PSP, or on the PMS vendor's bundled payment module, pays twice — once for the payment service, and once in the form of every protocol upgrade it does not get.

The architectural rule for the 100-room independent: the merchant account is the hotel's, held directly with Stripe or Adyen. The PMS integrates with it the same way it integrates with the channel manager or the RMS — as one connection among many. The PMS vendor is not the hotel's relationship with the payment processor. The hotel is.

Layer 6: Direct booking engine — AI-search readable

Once the AI agent has found the property, something has to convert the interest into a booking. That is what the direct booking engine does — except the conversion job has changed shape. The guest may never land on the website at all; the AI agent may complete the transaction inside ChatGPT or Claude itself. The direct booking engine has to handle both paths: the traditional website conversion the property still depends on, and the in-AI-app booking flow that is growing month by month.

This is the category where the focused players are most clearly winning. Two are worth knowing. The Hotels Network sits across this layer and Layer 2 — Lighthouse's distribution architecture handles the AI-surface exposure; THN handles the booking conversion on the hotel's own channels. Together they give properties of any size a direct presence in conversational AI search and a conversion-focused direct channel underneath it. Hotelchamp (Amsterdam, founded 2015) focuses entirely on direct channel conversion, personalization, and booking engine — a focused player in the original sense.

Both are accessible at 100-room scale. The alternative is the legacy direct booking engine — including the one bundled with the PMS suite — which is not building for the agentic surface at anywhere near the same pace.

The hub that connects them

Six specialist layers do not make a commerce hub on their own. The PMS turns them into one — by being the place every other layer reads from, the place every transaction settles back to, and the connection surface that joins them in real time. Without the hub, the stack is six tools the commercial team logs into separately.

The integration discipline is where the architectural choice actually lives. A commerce hub PMS makes the six specialists feel like one system. A unified PMS suite makes the six specialists feel like marketplace integrations that almost work. The commercial team sees the difference every morning, in whether the data they need is in one place or six.

What this means for the 100-room independent

The MVP is achievable today. None of the six layers requires custom development. Each has multiple credible specialists already in market at the 100-room scale. The work is integration discipline and architectural commitment — choosing to run a connected stack of focused players rather than a single suite that does everything badly.

The hotels that make that choice in 2026 will have an architecture that keeps moving. The hotels betting on a unified suite are betting their vendor will keep moving for them. The diagnostic is straightforward: against each of the six layers above, what shipped in the last six months, and what ships next quarter?

by Markus Busch, Editor/Publisher Hospitality.today

Read also: The PMS era is ending

Enjoying this analysis? Hospitality.today delivers daily insights on hotel distribution, AI trends, and travel commerce — straight to your inbox. Subscribe for free at Hospitality.today →