Seba BlancoI help independent hotels sell more effectively and operate smarter by combining hotel technology with sales and marketing.
Do not index
Do not index
Somewhere in your inbox this month there is a vendor deck with the phrase "agentic-first commercial operating platform". The pitch behind it reads something like: faster innovation cycles, deeper integrations, a conversational approach to hotel tech, the kind you would expect to work inside ChatGPT or Claude rather than despite them.
That sentence is doing a lot of work. This post unpacks it word by word, shows what it looks like on a normal Tuesday at a 20-room property, and ends with three things you can check before you sign anything.
Three words, three shifts
"Agentic" describes software that does things, not software that answers. A chatbot tells you your occupancy for next weekend. An agent reads the PMS, notices Saturday is at 40 percent, checks what the two hotels down the road are charging, drafts a rate change and a message to last year's October guests, and waits for your yes.
"First" is about who the product is built for. Most hotel tools today were designed for a person clicking through screens and had an AI feature bolted on in 2024 or 2025. An agentic-first platform flips the order: the primary user is an assistant acting on your instructions, and the screens exist for the moments you want to look at something yourself.
"Commercial operating platform" sets the scope. Not housekeeping boards or maintenance tickets, but the part of the business that brings money in: rates, distribution, direct booking, guest communication before and after the stay, reputation. In other words, the work you currently split across a PMS, a channel manager, a booking engine, a mailing tool and Google Business Profile.
Why "inside ChatGPT or Claude" is the important part
This is the line that matters most in that quote. The vendors saying it have accepted something the industry took a while to admit: hoteliers are not going to open a seventh dashboard. They are already talking to ChatGPT or Claude about their marketing, their pricing and their reviews. The question is whether your hotel systems can join that conversation.
The plumbing that makes this possible is a connector standard called MCP, the Model Context Protocol. Think of it as the USB port for AI assistants: a PMS that exposes an MCP server can be read and, with permissions, written to by whichever assistant you use. OTAs and several PMS vendors have been moving in this direction since 2025, and an assistant that can see your availability and rates directly is a distribution channel in its own right.
For you, the practical difference is this. An AI feature inside your PMS only knows what the PMS knows. A PMS that opens itself to your assistant lets that assistant combine your data with your competitor set, your email list and the forecast for the long weekend, in one place, in one exchange.
A Tuesday morning, before and after
Before: you open the PMS to check pickup, switch to the channel manager to compare pace with last year, skim a rate-shop email you never fully read, decide to drop the midweek rate, change it in three places, then remember you meant to email the couples who stayed for the wine festival. It is 10:40 and you have not looked at reviews.
After: you write, in your own words, "How did we pick up overnight, where are we soft in the next three weeks, and who on the past-guest list matches those dates?" The agent returns a short answer with numbers. You reply "Drop midweek 8 percent on those dates, keep the weekend, and draft the email." It prepares both, shows you exactly what will change, and executes only when you approve. Twelve minutes, one window.
The approval step is the whole point. Agentic does not mean unattended. A well-built platform logs every action, asks before anything touches a live rate or a guest inbox, and lets you set the rules once, in plain language: never below a floor rate, never email a guest twice in 30 days.
What it does not fix
An agent inherits your data. If your guest profiles have no email addresses, your rate history lives in someone's head and your PMS tags are inconsistent, the agent will be fast and wrong. Cleaning that up is your job, and it is worth doing whether you buy a platform or not.
It does not replace commercial judgment either. The agent can tell you that Thursday is soft. It cannot tell you that Thursday is soft because the regional airline cut a flight, and that this changes your whole shoulder-season plan. That part is still you.
Two more things to watch. Vendor lock-in gets more expensive when your workflows are conversations built around one platform's tools, so ask how you would leave. And "agentic" is already being stuck on products that are, underneath, a chatbot with a nicer logo. The test is simple: can it take an action across two of your systems, with an approval step, without you opening either one? If not, it is not agentic-first.
Small hotels do not have to wait for a vendor, by the way. A Make.com scenario that pulls the morning report from the PMS into Notion, hands it to Claude with your pricing rules and returns a draft rate change on WhatsApp is a working prototype of exactly this idea. Plenty of independents run something like it already.
What to do on Monday
Start with an inventory, not a demo. List every system that touches revenue or guest communication, and next to each one write whether it has an open API or an MCP connector. If your PMS has neither, that is the constraint every other decision hangs on, and it is fair to ask the vendor for a roadmap date.
Then pick one recurring commercial decision, ideally the one you make every week and half-resent, and write it out as a conversation: what you would ask, what data the answer needs, what action follows. That one page becomes your evaluation script for every platform that promises to be agentic-first, and a spec for anyone you hire to build it.
Finally, when a vendor uses the phrase, ask two questions. Which assistants can talk to your platform today: ChatGPT, Claude, Gemini, or only your own? And which actions can an agent execute, as opposed to recommend? The answers separate a commercial operating platform from a dashboard with a chat box.
Agentic-first is less a product category than a change in where the work happens: out of screens and into a conversation your systems are allowed to join. The hotels that benefit first will be the ones whose data is clean enough to trust and whose owners have decided which decisions they are willing to delegate. Do the inventory this week and the demos will get much shorter.
If you want the Make + Claude + Notion version of that Tuesday-morning workflow, the post "Automate Hotel Morning Reports with Make, Claude and Notion" walks through it step by step.
Written by
Seba Blanco
I help independent hotels sell more effectively and operate smarter by combining hotel technology with sales and marketing.