Is the PMS dead? Inside agentic AI for short-term rentals, with BOOM's Shahar Goldboim
- Aug 3
- 8 min read
Is your property management software already holding you back? That's the provocation behind Episode 1 of AI-Powered Hosting, the new six-part series from Host Planet and BOOM. Our guest is Shahar Goldboim, Co-Founder and CEO of BOOM, an AI-powered platform building what many believe is the future of short-term rental operations.
Shahar's argument is blunt: most operators are running "2012 software in a 2026 market," and the traditional PMS was never built to run a modern portfolio. In this conversation he makes the case for agentic AI – software that doesn't just report and automate, but thinks, decides, and acts across your entire business – and explains what the shift means in practice, whether you manage five properties or 50+.
Key takeaways about agentic AI in short-term rentals
The core problem is fragmentation. Most STR businesses run in silos across disconnected tools, with people manually shuttling information between them.
A PMS is really a "BMS." It's booking management software – a system of record – not something built to run your whole business.
Automation follows rules; agentic AI applies judgment. "When X happens, send Y" is automation. Looking at the numbers and deciding what to optimise is agentic execution.
AI is only as good as your policies and guardrails. The work shifts to designing better policies, not doing repetitive tasks.
Trust is earned in layers – start in co-pilot mode, graduate to autopilot once the system proves itself, and keep humans in the loop where it matters.
The PMS era isn't fully over, but an intelligence layer on top is becoming non-negotiable. Shahar frames AI as infrastructure, "like electricity or the internet."
What's actually broken with the traditional PMS?
Shahar's diagnosis is fragmentation. A typical operation runs the accountant on QuickBooks, marketing on HubSpot, the handyman on Breezeway or Monday – a stack of tools that don't talk to each other. The glue holding it together is people, manually transferring information by text, email, and phone calls.
"Nothing is joined up," as James put it. The whole business runs in silos. And that fragmentation is expensive: it forces humans to do the connecting work that software should handle.
The alternative Shahar describes is "business as software" – a system that doesn't just store and display data, but understands what to do with it. His example: a guest asks for an early check-in. To answer well, the system needs to know whether the previous cleaning is done, whether there's a maintenance issue (a shower being repaired), and what the property's policy is. Without all the information in one place, no one – human or AI – can give a good answer.
Why "PMS" is the wrong name
"A PMS should have been called a BMS," Shahar argues, "because it's booking management software, not property management software." The traditional platform was built as a system of record – to store and display information. But running a rental business is far more than managing bookings. It's marketing, customer support, finance, owner relationships, and the whole ecosystem around the property.
The shift he describes is from a PMS to an operating system – one where many of the actions are done for you, and repeatable work is handled automatically, so the operator can sit above the day-to-day and focus on making the operation better. The one thing that has no replacement, he's careful to note, is the human touch: the ability to connect, talk and create something genuinely different.
Automation vs. agentic execution: what's the difference?
Many operators believe they've already automated their business. Shahar draws a sharp line between automation and agentic execution:
Automation is rule-based. "When this action happens, send this message." Before check-in, fire the check-in instructions. Useful, but rigid.
Agentic execution applies judgment. The system looks at the numbers, spots what needs improving, and optimises – the way a skilled team member would.
His hosting example: instead of manually reviewing every listing to see what's missing from your OTA descriptions, an agentic system surfaces exactly what each property needs improved. It's the same relationship a revenue manager has with dynamic pricing tools like PriceLabs or Beyond – the tool sets a baseline, but a person applies judgment on whether to adjust. Agentic AI brings that judgment into the software itself, including optimising the very policies your AI agents run on. As Shahar stresses, an AI chatbot "works as good as your policies are, and as good as your guardrails" – so the real work becomes improving the policies, not doing the tedious tasks.
Inside an AI operating system for hosting
Asked what BOOM looks like in practice, Shahar contrasts it with the fragmented stack most operators run – separate task management, owner portals and guest portals that require constant logging in and out. Instead:
Everything under one platform, with AI agents for customer support and sales, and a holistic CRM covering owners, contractors, and guests – so communication across email, WhatsApp, and everything else lives in one view instead of scattered inboxes.
An AI assistant you can ask anything – Shahar's example: "What are the best mattresses in my portfolio?" The system reads your reviews and tells you what to buy, so you're making decisions from data, not guesswork.
Build-your-own reports and screens. Rather than paying extra for a fixed analytics view, you can create any report on any data, and design custom working pages and task views for your team.
Playbooks – repeatable tasks that run on a daily, weekly, or monthly schedule. A daily digest of major guest-affecting issues sent to your WhatsApp; a weekly check for guests with upcoming birthdays that automatically triggers a "send a cake" service call and logs the cost to your financials. You build a workflow for anything you need, saving your team hours.
A new optimisation product that improves every segment of the business – revenue, listing descriptions, policies, and communication.
The result, as James summarised, is that a property manager becomes "an architect of hospitality rather than a doer of hospitality." That matters in an industry where managers work 24/7, 365 days a year, where a single bad review dents your ADR, and where rising labour costs and competition are squeezing margins.
Doesn't every PMS have AI now?
Almost every platform has bolted an "AI" badge on somewhere. Does that solve the problem? Shahar's answer is nuanced: yes, any optimisation helps – but bolting features onto legacy software isn't the same as building from the ground up with a vision for the whole ecosystem.
His metaphor: "Adding salt is great, but does it make the dish really good? It'll make it better, but it won't make it perfect." What matters are the underlying ingredients and the complete system you're building. The real question for operators, he says, is simple: which system makes you the most money, saves you the most money, and saves you the most time?
From task management to business orchestration
The deeper shift is from tracking jobs to orchestrating the business – closing the loop. It's not about whether a single task got done, but about looking at the effect each action has on the whole business and creating a loop of improvement: I did the task – how did the guest feel, what review did it earn, what can I improve next time?
That opens the door to genuine personalisation and A/B testing. Shahar's favourite examples are gloriously concrete:
A handyman visiting a family home brings popsicles to leave in the freezer for the kids. Does that $10 gesture lift reviews? Test it, measure the review impact a month later, and keep it or kill it.
Test different welcome touches – a note, a brand of champagne, one hamper versus another – and track the ROI on each, segmented by guest type (couples vs. families, young kids vs. older) and property tier (a $1,500-ADR luxury home vs. a $100 apartment).
As James noted, better data is the whole game: "If you provide AI with more complete data, the better it performs." The system becomes a living experiment log – and can even be told to optimise automatically: keep what lifts reviews, shut off what doesn't, and propose a new plan.
How do you trust software to think and act?
Handing execution to a system that acts on its own makes operators nervous – losing control, mistakes at scale, a guest getting the wrong answer. Shahar's trust framework has three layers:
Co-pilot. The system suggests; you approve. You watch the changes and track how often you agree with its suggestions.
Autopilot. Once your agreement rate crosses a comfortable threshold, you let it act on its own.
Human-in-the-loop policies. You define exactly where a human must step in – refunds, discounts, major problems – while the AI handles the endless easy questions ("where's the remote?").
He illustrates trust with a Tesla story: his wife flatly refused self-driving for over a year, then tried it once – and now refuses to drive any other way. That's the arc of adoption. And unlike human staff, an AI agent gives 24/7 service, never takes vacation, and sidesteps the hire-fire-onboard churn that plagues seasonal businesses.
Is the PMS era really over? What operators should do now
Interestingly, Shahar doesn't declare the PMS entirely dead. It remains the system of record – you still need OTA connections and the core booking functions, and that foundation is fundamental (even as direct bookings grow). What's changing is that an operating system layer sits on top of it, and operators who don't adopt one risk being replaced by competitors who can "produce a better product at a cheaper price."
His framing for the shift: AI is infrastructure, "like electricity, like the internet." Just as businesses eventually had to have computers, then websites, connecting your operation to intelligence is becoming table stakes – and the payoff is a better guest experience and better revenue for everyone.
Frequently asked questions
Is the traditional PMS dead? Not entirely. It remains the system of record for bookings and OTA connections. But according to Shahar Goldboim, it can't run a whole modern business on its own – an agentic AI operating system layer on top is becoming essential.
What is agentic AI in short-term rentals? Agentic AI goes beyond rule-based automation ("when X, send Y"). It applies judgment – analysing data, deciding what to optimise, and acting across the business, from listing descriptions and pricing to guest communication and policies.
What's the difference between automation and agentic execution? Automation triggers a fixed response to a fixed event. Agentic execution looks at outcomes, makes decisions, and improves the process over time – closing the loop rather than firing isolated actions.
How do you keep control when AI acts on its own? Through a layered trust model: start in co-pilot mode (AI suggests, you approve), move to autopilot once you trust its accuracy, and set policies defining where a human must stay in the loop (refunds, discounts, major issues).
What is BOOM? BOOM is an AI-powered operating system for short-term rental operations that unifies task management, CRM, guest and owner communication, custom reporting, and automated "playbooks" – designed to think, decide, and act across the whole business rather than just record bookings.
Watch the full episode
This article is based on Episode 1 of AI-Powered Hosting, featuring Shahar Goldboim, Co-Founder and CEO of BOOM, brought to you by Host Planet.
Watch or listen to the full conversation for Shahar's complete case – including the Tesla analogy and a live tour of what an AI operating system looks like – and subscribe for the rest of this six-part series. Catch the full episode on YouTube, Spotify, or Apple.
Comments