Stop Advising. Start Doing.

Last week Julien Bek at Sequoia published an essay “Services: The New Software” every hospitality operator in my opinion should read. The thesis: for every dollar spent on software, six are spent on people operating that software. The next great companies won't sell tools. They'll sell the work.

Not an accounting program — closed books. Not a CRM — a closed deal.

We read it and recognized our industry in every paragraph.

The Copilot Trap

A typical hotel runs eight to twelve systems. PMS. Channel manager. Revenue management. Booking engine. CRM. Guest messaging. Reputation monitoring. Analytics. An STR management company with 50–200 properties runs even more.

How many of them deliver end-to-end execution?

Mews shows a “smart tip” about the guest. SiteMinder recommends a new rate. Sabre suggests an upsell. Lighthouse flags a competitor’s price drop. Cloudbeds’ Signals and Pricing Intelligence Engine light up a dashboard of pricing and demand “signals”. Apaleo’s open, API-first PMS and Agent Hub expose AI agents and workflows you can plug into your stack.

All useful. All just notifications for your staff to act on. Manually. At 3 AM. If they remember.

Bek calls this the split between intelligence and judgement. Intelligence is knowing the right answer. Judgement is deciding what to do with experience and context. Most hotel operations — recalculate a rate, confirm a booking, dispatch housekeeping, answer a routine question — are pure intelligence. The rules are complex, but they are rules.

AI can already handle these autonomously. The problem isn't that AI isn't smart enough. The problem is that it has no hands.

Two Sides, One Gap

Look at what's happening in travel through Bek's framework.

On the demand side: copilots everywhere. Booking.com built an AI Trip Planner. KAYAK launched a chat-first search. Tripadvisor integrated with Perplexity. Smart recommendations for the traveler.

On the supply side: Airbnb declared an AI-first strategy with agents that will book, modify, and cancel trips autonomously. Expedia shipped an end-to-end AI agent in Hotels.com. Microsoft Copilot Actions can process bookings. OpenAI Operator fills out forms by itself.

These agents are ready to act. But most hotels have no system capable of receiving and executing their commands. It's a highway leading to a city with no parking.

Bolting AI onto Legacy Doesn't Work

Every major PMS vendor is rushing to slap "AI-powered" on the box. Oracle adds AI upsell to OPERA. Cloudbeds launches an Intelligence layer. Mews's smart tips hit 5 million views per week.

Views aren't actions. A recommendation that requires manual execution is just a fancier notification. Add an LLM to a closed legacy core and you get cosmetics, not transformation. No open write APIs, no idempotency, no audit trail for external agents — real automation is impossible.

The Autopilot Playbook

Bek's playbook for autopilots: start where outsourcing already exists. The budget is there, the scope is clear, the ROI is immediate.

In hospitality and STR, this maps perfectly.

What's already outsourced or pushed to the night shift? Phone calls. Routine guest requests. Housekeeping coordination. Night audit. Booking confirmations. Rate distribution. All intelligence-heavy, rule-based, high-volume. Exactly where autopilots win first.

This is what Polydom builds. Una is not a chatbot that answers breakfast questions. It's a digital operations employee that actually does the work:

Takes the call and resolves the issue — from question to PMS task creation, under 500ms response latency.

Executes what your systems recommend. Revenue management says raise the rate — the rate changes. Not a notification. An action.

Acts as a gateway for external AI agents. When Booking, Airbnb, or Copilot Actions send a command, Una validates it against hotel policies and executes. With payment, audit, and SLA.

Coordinates operations. Dispatches housekeeping, manages tasks, processes check-ins and check-outs.

One dollar per hour. Not because we're undercutting — because an autopilot competes with labor, not software. A front desk agent costs $20–25/hour. A missed call at 2 AM costs a booking.

Why STR Goes First

Bek says the best entry point for autopilots is high outsourcing, high intelligence ratio. Short-term rental management fits this almost exactly.

A professional STR company with 50–500 properties faces a scale paradox. Each property is a separate mini-operation with unique rules, rates, and cleaning thresholds. Hiring a dedicated admin per property is impossible. Building a call center across three time zones is expensive and fragile. STR operators already live in the outsourcing logic. They don't need to be convinced to delegate work externally. They need someone to delegate to.

And 80–90% of STR guest requests are routine. Door code. Wi-Fi. Parking. Early check-in. Late check-out. Trash. Pure intelligence by Bek's definition. The autopilot handles these completely. Humans step in only where judgement matters: an unusual conflict, a VIP guest, a non-obvious call.

What Comes Next

Bek ends with a sharp observation: in 2025, the fastest-growing AI companies were copilots. In 2026, many will try to become autopilots. But they face the innovator's dilemma — selling the work means cutting their own customers out of doing it. That's the opening for those building autopilots from day one.

PMS vendors will keep adding recommendation layers. Channel managers will become "commercial platforms" with pricing advice. CRS and booking engines will add AI concierges to hotel websites. All copilots. All making professionals more productive.