SleekAI doesn't replace your PMS - it sits in front of it as the guest-facing layer. For booking actions it links out to your booking engine; for guest-specific context, you can pass logged-in user data or reservation lookups into the prompt. Most properties keep the PMS as the single source of truth and use the bot purely for question-answering, soft upsell, and routing into existing systems.

Whatever language the guest writes in. SleekAI replies in the same language even if your hotel content is only in English - it translates property detail on the fly. For markets where you have localised content (Japanese site copy, Spanish promo pages), it picks the localised source automatically. We've seen properties in Berlin and Vienna get conversation logs in seven languages within the first week.

Yes - multibot lets each property have its own assistant scoped to that hotel's rooms, amenities, restaurant, spa hours, and policies. A shared knowledge layer can cover brand-wide rules - loyalty program, master cancellation policy - while property-specific content stays per-bot. This avoids the cross-contamination risk of a single bot mistakenly quoting Lisbon hours to a Berlin guest.

OpenAI Files vector store handles big content - tour books, restaurant menus, policy PDFs, multilingual brochures - so accuracy stays high even on large or chain sites. Retrieval pulls only the relevant passages into context per turn, keeping token use predictable and answers grounded. For a 200-page property handbook, that's the difference between a reliable answer and a hallucinated one.

It links to your existing booking engine rather than acting as a parallel booking system. We recommend hotels keep the booking flow on their own site - direct bookings save 10-15% in commission versus OTA - and use the chatbot as the conversion catalyst. The split keeps a single source of truth on availability and rates, and avoids reconciliation headaches.

Conversations live in your WordPress database under your control. Using your own OpenAI API key with the standard API terms means guest data is not used for model training by default. For GDPR-conscious EU properties, document the bot in your privacy notice, set retention to match your booking-data policy, and confirm OpenAI's DPA. Encrypt the WordPress database at rest as part of your standard hardening.

By default the system prompt prioritises on-property restaurants and partners, but for guest trust the bot shouldn't pretend the neighbourhood doesn't exist. A balanced approach - own restaurant first, with one or two trusted local partners after - typically wins more loyalty than aggressive own-property steering. You can tune the recommendation order in the system prompt to match your brand standards.

For routine group questions - block availability, F&B minimums, AV setup - the bot can answer from your events page. For active negotiations or contracted business it should hand off to your sales-and-events team with the captured context, since group rates and force-majeure language belong to humans. Configure the system prompt to recognise group-booking signals and route accordingly.