In 2023 we wrote about how AI was "going to revolutionise" the sector. It already has — and it is not ChatGPT answering emails. It is agents that close bookings on their own, pricing that adjusts minute by minute, and end-to-end automation that cuts a flat's operating cost by 25% to 45%. This is what we actually do at MyRentalHost in 2026, without the hype.
From the chatbot of 2023 to the agent of 2026
Three years ago, AI in this sector was a chatbot answering "what time is check-in?" with a template. Today an agent:
- Reads the entire booking (dates, guests, language, platform)
- Cross-references your systems (cleaning calendar, lock, guest data under RD 933/2021)
- Decides what to reply and what actions to trigger — it does not just answer, it acts
- Converses in the guest's language, with the booking context, keeping the memory of the whole conversation
- Escalates to a human only when it detects ambiguity or risk
The models behind it: Claude 4.6, GPT-5, Gemini 2 Pro. All available via API at costs that would have looked like science fiction in 2023 (€0.003 per typical interaction).
Where AI is actually moving the needle today
1. Revenue management with ML
The price of your night is not a hunch. Dynamic pricing models process:
- Five years of occupancy and price history for your flat
- Comparables in your area (RevPAR of similar flats)
- Events: conferences, festivals, sport, public holidays
- Lead time: people booking 90 days out pay differently from those booking 5 days out
- The competition's curve: if the flats on your street drop their price, you adjust; if they raise it, you adjust
Typical result (MRH data, 2025): +15-35% revenue versus static pricing. The calculator shows you the figure for your case.
2. Guest support agent 24/7
Before: a human available, or a robotic template. Now:
- The guest asks "is there parking nearby?" → the agent knows the flat is in the Eixample, knows the 3 closest car parks, knows their current rates, and recommends them with opening hours.
- The guest asks "can I have a late check-out?" → it checks the cleaning calendar and the next booking, works out whether it is possible at no cost, and proposes a specific time.
- The guest reports a problem → it categorises it (electrical, plumbing, neighbour noise), prioritises it, opens a job with the right technician, and alerts a human if it is serious.
3. Predicting cancellations and fraudulent bookings
A model trained on thousands of historical bookings picks up subtle signals:
- Anomalous lead time (booked in the last 3 hours of the day with a new card)
- The guest's response pattern (replies / does not reply / replies with a template)
- Match between name, country and platform
- The account's history on other platforms (where there is a cross-match)
Application: run extra verification on high-risk bookings, not on all of them. It saves friction for legitimate guests and cuts problem check-ins by 60-70%.
4. Listing optimisation with visual AI
The model assesses every photo in the listing: lighting, angle, aspect ratio, what it shows. It suggests which photo to put first (the one that generates the most clicks in your segment), what is missing, what is surplus. It also assesses the copy of your title and description and proposes A/B alternatives.
5. End-to-end automation between systems
The most invisible and the most valuable part. When a new booking comes in:
- The agent verifies the guest's details (ID, identity) — automatic
- Generates the lock code for the exact dates — automatic
- Schedules the cleaning with the local cleaner for the day before check-in — automatic
- Sends entry instructions in the guest's language 48 hours ahead — automatic
- Transmits the data to the authorities under RD 933/2021 — automatic
- Switches on the water heater and the heating 4 hours before arrival — automatic
A human only steps in when something falls outside the happy path. This used to be 45 minutes of work per booking. Now it is zero.
What is NOT solved in 2026
Decisions where AI is still weak
Resolving an emotional complaint from a guest, negotiating with an angry neighbour, deciding whether to accept a young group in a sensitive building, stepping into a delicate health situation. A human does that — and should. AI frees up the hours so the human can focus on it.
The mistakes owners make when they bring in AI
- "I buy a chatbot and forget about it." Without integration into your PMS, without training on your specific flat, a generic chatbot is worse than nothing — it frustrates the guest and leaves the key points unanswered.
- "I put AI everywhere." Some points in the process need a human, for brand reasons, legal reasons or context. Knowing where to apply AI and where not to is half the value.
- "Flat platform fee." You pay €99/month for a platform "with AI" that is really just automated mailing. Check WHAT the system does, not just how it is sold.
What MyRentalHost uses today (as of 14 May 2026)
- PriceLabs for dynamic pricing (ML layer)
- Anthropic Claude for the guest conversational agent and replies in the local language
- Our own system for cancellation detection and identity verification
- Make + n8n for automation between systems (orchestration)
- Nuki locks with codes generated by API for every booking
A stack an individual owner would never put together on their own — but which, for us, managing more than 50 flats, makes economic sense to maintain.