Serviced Apartment News recently hosted a panel titled “The Tech-Enabled Operator: Hype vs. Reality,” with our Managing Director – Europe, Giles Horwitch-Smith, among the panelists.
The goal was straightforward: cut through the AI noise and talk about what’s actually working and not working for today’s operators. What stood out was how consistent the answer was across the panel.
Two operators running very different businesses, Sam Ghosh of STAY and Bala Balasubramaniam of Viridian Apartments, sat alongside Jay Humphries, Chairperson of HoCoSo, whose consulting work has taken him inside many hospitality and long-stay operations over the years. Here’s what we took from the conversation.
The problem isn’t too little technology. It’s too much of it, disconnected.
“We probably are overinvested as a company, but it’s just that we’ve got too many systems which don’t talk to each other.” Sam Ghosh, VP Operations - STAY at LABS Collective.
It’s a story we hear frequently from operators, and it’s rarely framed that honestly. Most assume their problem is a capability gap. They think they need one more tool, one more feature, but it’s almost never that. It’s that the systems they already have don’t share a common source of truth, so every new tool just adds another silo to reconcile at month-end.
Giles’s answer on the panel was the one he gives in every conversation like this: stop asking whether a system has a feature. Ask where your data is going to live, and which processes you’re willing to route through it and what business outcome you’re seeking. That’s a harder question than most operators want to answer before a demo, but it’s the only one that matters. Don’t ask it and you risk being back in market in eighteen months.
Where AI is earning its keep, and where it isn’t yet.
Giles was asked directly where he’s seeing measurable ROI today, and he gave a straight answer, because this space needs more of those and fewer roadmap slides.
It’s real in guest inquiry handling and post-booking questions. The better tools are genuinely resolving the majority of guest queries before anything needs escalating, and that’s a number you can actually measure. It’s real in maintenance triage: getting the right issue to the right contractor faster, even though predictive maintenance itself isn’t there yet. Revenue management is one of the earliest AI wins in our category, particularly on shorter stays. And document and invoice processing, checking what’s coming in against what a person would otherwise key in by hand, is an area seeing real gains too.
Where it’s not there yet: chatbot ROI in short-and-medium-stay booking, and anything predictive on maintenance. Giles said it on the panel and it bears repeating: a lot of AI investment right now is adding cost without adding return, and the only way out of that is to pick one process, measure your baseline before you touch it, apply the automation, then measure again. Anyone selling you a platform-wide overhaul without that discipline is selling you the hype half of the panel title.
Some things aren’t meant to be automated.
The part of the conversation we keep coming back to wasn’t about technology at all. Both operators on the panel landed on the same boundary, independently, without prompting each other. Escalation, distress, service recovery: that’s where the human has to be. Bala Balasubramaniam put it plainly: a theft, a broken AC unit, a guest in genuine difficulty. Automate around that at your own risk, or you’ll find out about it on a review site. Jay Humphries, whose consulting work has taken him across a much wider set of operators than any one operator sees day to day, framed the same boundary from a wider angle: solve for loneliness on the guest side, and fear on the employee side, and you’ll understand where technology helps and where it gets in the way.
That’s the line we try to hold in everything we build. We’re not in the business of automating the relationship between an operator and their guest. We’re in the business of making sure the back office is clean and current enough that the humans on the ground have the time, and the data, to be good at the parts of the job that actually require a person.
The real risk isn’t falling behind on AI. It’s building on a broken foundation.
Giles’s closing point to the audience was this: if your data is fragmented across spreadsheets and disconnected systems, AI doesn’t fix that. It just gives you the wrong answer faster. That’s not a caution against adopting AI. It’s a caution against adopting it on top of chaos and expecting clarity to follow.
So the answer to the hype-versus-reality question isn’t about which AI feature to buy next. It’s about whether the operational foundation underneath the AI can be trusted. Get the data foundation and workflows right, and the intelligence layer on top of it (human or artificial) will have something solid to work from.
The full recorded conversation is worth the watch if you’re making technology decisions this year: