The most interesting thing AI changes about vertical software is that it can now do the work itself rather than help someone else do it, and that moves the budget it competes for from the software line to the payroll. For twenty years, vertical software sold systems of record and the products around them. They streamlined operations, but each one still needed a person to operate it. That kept the spend capped at whatever an operator would budget for software, which in most of these industries is a rounding error next to what they spend on people. We have been looking across vertical AI for the segments where that gap is widest and where the dependency on human labour most constrains growth. Three things convinced us hospitality is where it plays out first.

The labour crisis is structural, and operators have already exhausted their levers: Accommodation and food services has averaged 76% annual turnover since 2001, against 43% for the economy as a whole. Surveys have consistently found that 70%+ of hotels carry openings they cannot fill despite active searching, averaging 6-7 vacancies per property, with housekeeping and front desk hit the hardest. Meanwhile hotels are paying 22% more than in 2019 for 7% fewer hours worked, compounding the problem. Estimates also project that hospitality will be short 8.6 million workers by 2035, around 18% below the staffing it needs. The scheduling lever, the cross-training lever and the non-daily-housekeeping lever have all been pulled. The next increment of efficiency cannot come from managing people harder.

The incremental dollar in a hotel falls almost entirely to margin: A hotel is a fixed-cost asset and the room at 11pm exists whether or not anyone books it, so almost every incremental dollar of revenue drops to the bottom line. That makes every inbound request valuable and every missed one expensive.

The system of record is fragmented but digitised: The top five PMS platforms cover about half of US rooms and the top twenty about 70%. There is no single gatekeeper who can lock a new entrant out, but near-universal PMS penetration means there is a real data substrate to build on.

Enter Dextr: The System of Action

Dextr has built an agent layer that sits on top of a hotel's existing stack and does the work by itself. The guest calls and the agent checks live availability, describes the room, quotes the rate, captures payment and writes the reservation into the PMS. In most cases this happens with no human intervention. We verified this ourselves by calling customer properties before we invested, and the consistent feedback across reference calls was that guests do not realise they are speaking to AI.

A few design choices convinced us this was built by a team which understands hotels and knows what hospitality needs.

It never asks the hotel to replace anything: Dextr sits above the PMS and the vertical tools already in place. Switching costs at the system-of-record layer are brutal and rip-and-replace is a multi-month project that most owners will simply refuse. By declining to compete with the platforms, Dextr turns them from gatekeepers into partners worth leveraging.

It works on top of any underlying system: Rather than waiting in line for API access and inheriting someone else's roadmap, Dextr's agents operate the PMS the way a front-desk clerk does.

Every agent is pointed at a P&L line: The voice agent captures after-hours and overflow calls that were previously lost revenue. The guest agent delivers a high NPS experience for guests and monetises early check-in and late checkout that most properties were giving away free. The staff agent sequences housekeeping and gives operators real control over their largest variable cost. Every agent maps to a line that shows up in the owner's P&L.

It is built for the operator, not the software buyer: Managers run the whole system in natural language from the messaging apps they already live in, on their phone, without logging into a dashboard. The platform translates automatically between the guest's language, the owner's and the housekeeping team's. For an industry with no IT staff and high turnover, zero training is not a nice-to-have.

The system configures itself: Hotels set up and change their own agents by describing what they want in plain English, which is steadily moving onboarding from a human process to an AI-driven one.

The Founders

Sajid Shariff and Scott Arnold bring a combination we rarely see in vertical AI: an operator who has already run the industry's hardest go-to-market, and an engineer who has built production AI systems at scale.

Sajid built OYO's business in the US. That job meant persuading independently owned properties, usually family-run for decades, to change how they operate their single largest asset. He understands the hospitality business, the nuances of the operations, stakeholder incentives and has seen first-hand the challenges plaguing the industry.

Scott spent years at Meta working on infrastructure and systems at scale, with a decade of shipping production-grade AI behind him. At Dextr he has focused on the agent harness platform rather than just the agents themselves. The biggest challenge in deploying vertical AI is onboarding cost and effort, so he built an entire self-onboarding platform that lets hotels configure and modify agents themselves with no involvement from the Dextr team, which is what allows the company to scale rapidly.

Here’s what sets Sajid and Scott apart. The first is customer obsession where they shape the product around what actually fits an operator's workflow rather than what is easiest to build. The second is an unusual ability to hear what customers cannot articulate, where several of the sharpest features in the product came from needs the hotel owners and operators never put into words. The third is range in being able to move from small independent hotels to larger multi-property groups with equal ease, and landing both. The fourth is the ability to build cutting-edge systems that plug into live hotel operations and start showing value quickly.

The results of these are evident in the customer outcomes where feedback on the Dextr product and team is one of the strongest we have seen in the vertical AI space. Dextr has quickly become an indispensable part of the core operations of these hotels and is now contributing to a substantial share of revenue and profits.

Looking Ahead

Beyond hotels, the same labour economics and the same missing action layer exist across the wider experience economy. Campgrounds, RV parks, short-term rental managers, golf courses, spas and casinos all run on lean teams with fragmented software and no intelligence connecting them. Dextr is already live across several of these, and the platform was built so that a new vertical is a workflow problem rather than a rebuild.

We are also cognizant of what will not be the moat. Voice quality is very good today and it will commoditise, and we are not underwriting a model advantage. What compounds is the operating context each property accumulates inside the system, the integration surface across a long tail of platforms nobody else wants to do the unglamorous work of connecting, and a channel where the incumbents have a reason to sell for Dextr rather than against it. A meaningful share of US rooms sits inside brand-managed systems that are hard to enter, and there is a substantial company to build below that line before it needs to be crossed.

We are excited to lead Dextr's seed round and to partner with Sajid, Scott and the team as they build the AI workforce for hospitality.