Zaplar Wants AI to Run the Hotel Back Office - and Let the Humans Keep the Judgment

A hotel is a machine that never stops running, and most of the people who run it will tell you they spend the day staring at screens instead of guests. Bookings live in one system. Payments in another. The events team keeps a spreadsheet. The revenue manager keeps a different one. Somewhere in the middle sits a human copying numbers between tabs at 11pm. Zaplar, a four-founder startup out of Stockholm, looked at that scene and asked a plain question: what if the software just did the copying - and everything around it - by itself?

That is the short version of the company now sitting in Y Combinator's Summer 2026 batch. Zaplar calls itself "a frontier agentic operating system for hotels." Stripped of the jargon, it is an attempt to replace the tangle of tools a hotel runs on with one platform where AI agents handle the operational work and staff sign off on it. The founders are not shy about the ambition. They are shy about one specific thing, which we will get to, because it says more about the company than the product deck does.

01An alphabet soup nobody asked for

Walk into the back office of a mid-sized hotel and you will meet an acronym for every task. The PMS handles rooms. The CRS handles the central reservations. The RMS sets prices. MICE tools handle meetings and events. The POS rings up the bar and the spa. Each was sold separately, each was bought at a different time, and each speaks a slightly different dialect. The result is a workday spent translating between systems that were never designed to talk.

Zaplar's team - which describes itself, unusually for hotel software, as "mathematicians, engineers and AI researchers" - argues that this fragmentation is not a bug the incumbents forgot to fix. It is the shape of an industry that grew its software one bolt-on at a time. Their reading is that you cannot patch your way out of it. You have to start over.

02AI-native, not AI-flavored

Every legacy vendor now has an "AI feature." Zaplar's distinction is architectural: it was built for agents from the first line of code, rather than sprinkling a chatbot on top of a decades-old database. The company frames the difference as the gap between software that was designed around AI and software that had AI layered on after the fact. In practice that means the agents are not a helper in the corner. They are meant to be the thing doing the work.

What work? Voice bookings. Email handling. Requests for proposals from event planners. Guest portals. Keeping the PMS in sync. The follow-up messages nobody remembers to send. Zaplar says its agents learn how each individual venue actually operates - not a generic hotel, but this hotel, with its quirks - and then act the way that property would.

That line is the company's stated vision, and it doubles as a product spec. If the goal is presence, then every minute of admin the software absorbs is a minute returned to the floor. The trick, of course, is trust. Hoteliers are not going to hand a bot the keys to their revenue and their guest relationships on faith.

03Autonomy, with a human holding the pen

Here is where Zaplar makes its most interesting call. The agents plan and suggest; they do not act alone. Every proposed action - a price change, a reply to a guest, an event confirmation - runs through a review-and-approve workflow. Staff see what the system wants to do and sign off, or don't. The company lists "autonomy with oversight" and "full transparency" among its core values, alongside enterprise-grade security.

It is a subtle but important position in a year when much of the AI conversation is about removing humans entirely. Zaplar's wager is the opposite: in a business where a single bad guest interaction travels fast, the winning product is the one that does the busywork while leaving judgment where it belongs. Automation for the boring parts, humans for the parts that matter.

04What's actually in the box

Zaplar breaks its platform into a handful of pieces that map onto the systems it wants to replace:

• MessagingOne inbox for every channel, from email to the OTAs, where agents can draft or handle guest conversations.
• Booking EngineCustomizable direct-booking experiences built to pull business away from the online travel agencies and their commissions.
• Point of SaleUnified payments across rooms, events, spa and food-and-beverage, with configurable options.
• Guest IntelligenceGuest profiles that personalize the stay and surface upsells - the add-on nobody thought to offer.
• Events & VenuesA modern portal for event and venue bookings, with the back-office admin and RFP handling automated behind it.

One detail is worth pausing on, because it tells you how these founders think. On the company blog sits a post titled "The surprising mathematics behind housekeeping scheduling." Deciding who cleans which room in what order looks like a chore. Zaplar treats it as constrained optimization - the same class of math airlines use to schedule crews. It is a small window into a bigger idea: much of hotel operations is an operations-research problem hiding under a hospitality costume.

05The founders who won't cold email you

Now, the shy part. Zaplar has a public rule, written up on its own blog: we don't cold email hotels. Instead the founders go in person - into the lobby, at the desk, face to face with the people who would actually use the thing. In an era of automated outbound and AI-written sales sequences, a startup building AI software chose to sell the least scalable way possible.

It reads as the whole point. You cannot build software that learns how a hotel really runs if you have never stood behind the desk during a Friday check-in rush. The hundreds of owner interviews the team cites were not market research theater; they were the raw material. The refusal to cold email is the same instinct pointed at distribution.

06Model, market and the competition

The business model is recognizably B2B SaaS: hotels and venues pay a recurring fee for a platform that replaces several paid tools, with payment processing running through the built-in POS. Because Zaplar collapses five subscriptions into one, the pitch to a finance-minded operator is not only "better software" but "fewer invoices."

The market is crowded at both ends. On one side sit the incumbents - Oracle's OPERA, and modern cloud players like Mews and Cloudbeds - who own the property-management layer and are racing to add their own AI. On the other, newer YC-backed entrants such as Lance are chasing the same "AI agents for hotels" territory. Zaplar's differentiation is the two-word bet it keeps returning to: AI-native and consolidated. Not a smarter PMS. A replacement for the idea of a PMS as a separate thing at all.

07Four roles, one table

Zaplar's founding team splits cleanly. Axel Andersson Lingbert is CEO. Jon Milles is CTO. Douglas Solberg leads product as CPO. Oscar Karlströmer runs finance as CFO and is the author of the no-cold-email manifesto. The company describes a small crew of engineers and researchers, with backgrounds the site lists as ranging from machine learning to spectral geometry - which is a long way from the usual hotel-software resume, and rather the point.

Funding so far is early and angel-scale: roughly SEK 3 million from Nyman Holding and Michael Ingelög, plus the standard Y Combinator deal. The company has also joined the California Hotel & Lodging Association, a small signal that the Stockholm team is already looking across the Atlantic for its first American operators.

08The quiet frontier

The loud AI stories of the moment are about code and images. The quieter one is about the back offices of ordinary businesses - the scheduling, the invoicing, the follow-ups - where agents can absorb hours of unglamorous work without anyone tweeting about it. Hotels, with their sprawling operational surface and their acknowledged software mess, are close to an ideal test case.

Whether Zaplar becomes the operating system for that or one of several trying is an open question; the company is months, not years, into the work. But the framing is worth borrowing even if you never run a hotel. Find an industry drowning in disconnected tools, listen to the people using them until you understand the job better than they can explain it, and build the thing that does the work while leaving them the decisions. That is a template, and Zaplar is running it in a lobby near you - because, remember, they will not be emailing.