We used to answer a Russian buyer in thirty minutes when it landed during business hours; now the concierge replies in thirty seconds at two in the morning and the human agent only steps in when the lead is worth it. The difference isn't productivity, it's premium experience. The buyer feels the agency is awake when they need it.
+38% qualified leads / month
Trust
What we measure. What we show. What we keep quiet.
We don't publish client names until we have three paying with explicit permission. We don't fabricate metrics. And when a number isn't confirmed, we say so. This is what we measure in production.
Real concierge captures.
Voices from the sector
Three archetypes, three ways to use it.
We manage twenty properties for five families. Each family wants real-time transparency on what's happening with their wealth. The LuxeAI dashboard lets us show them exactly what's happening with their Moraleja property without the human agent having to send a weekly email. The operational load drops 70% without losing closeness.
70% reduction in operational emails
Three offices in Marbella, Estepona and Mijas. Four languages in constant use. Each office used to respond with its own criteria; now there's a single coherent multilingual source of truth. That's what we were missing; I understood in the first demo that LuxeAI isn't a chatbot, it's operations.
Single source of truth across offices
Indicators
What we already measure.
+38%
Qualified leads / month
Provisional, vs human-only qualification
<5 min
Average speed-to-lead
Median 24 / 7
70%
Operational emails avoided
Family office archetype
4
Languages in use
ES · EN · RU · DE
24 / 7
Concierge availability
No drop-off due to human downtime
<48 h
Time to proposal
After initial call
What's already working, not what we promise.
Real LuxeAI platform metrics verifiable right now from this same website. No inflation, no client figures we cannot publish yet.
- Claude Sonnet 4.6AI model in productionVerify on /producto and the home chatbot (response model header).
- 4 languagesNative multilingual supportES · EN · RU · DE. Write to the chatbot in any of the four and check the response.
- < 5 secFirst-message response timeMeasurable by opening the home chatbot and sending a question.
- Frankfurt (EU)Data residencyPostgres Supabase region eu-central-1. No international transfer.
- RLS · DBMulti-tenant isolationPostgres Row Level Security by tenant_id. Auditable on request pre-contract.
- 0 cross-tenant dataBetween clientsAdversarial E2E smoke included in pipeline. Reportable with the client.
Built on industry-standard components, not on smoke.
All infrastructure is industry standard. No black boxes, no unverifiable "branded" technology. Auditable under NDA before any commitment.
Generative AI
- Anthropic Claude Sonnet 4.6 (primary model, canonical alias)
- OpenAI GPT-4 class (evaluated fallback)
- RAG over Postgres pgvector (where applicable)
- Hard daily cost cap · no billing surprises
Web platform
- Next.js 15.1 (App Router · React Server Components)
- Auth.js v5 passwordless magic-link
- Tailwind v4 + custom LuxeAI theme
- Strict TypeScript (noUncheckedIndexedAccess)
Data and infra
- Supabase Postgres Cloud Pro · Frankfurt EU region
- Multi-tenant Row Level Security by tenant_id
- Encryption at rest and in transit
- Daily encrypted off-site backups
Integrations
- WhatsApp Business API (Meta Cloud)
- Instagram Direct via Meta Graph API
- HubSpot · Salesforce · Apimo · Inmovilla (bidirectional sync)
- Resend EU (transactional email)
Operations and security
- Sentry EU Frankfurt · error traceability
- Cloudflare CDN + DDoS protection
- Automatic wildcard SSL renewal
- GitHub Actions pipeline · lint + typecheck + build + tests
What's live, what's in pilot, what's on the roadmap.
Full transparency: we tell you which modules you can use today, which are in pilot with a closed client and which arrive in the coming sprints. No promises of dates without commitment.
Live in production
- Conversational AI Concierge · home chatbot accessible at luxeai.es
- Claude Sonnet 4.6 model with guardrails, PII redactor and daily cost cap
- Server-side contact form (anti-spam + honeypot + rate-limit)
- Multilingual marketing ES + EN indexable · RU + DE noindex pending native reviewer signature
- Full structured schemas (Organization, LocalBusiness, RealEstateAgent, Service × 4, SoftwareApplication, OfferCatalog, FAQPage, BreadcrumbList)
Pilot (restricted access)
- Owner Dashboard · secure-link access · complete UI in private testing with first agencies
- AI multilingual listings studio · pipeline in editorial validation
- HubSpot CRM sync · functional webhook, luxeai_* custom properties
- n8n workflows (qualification + nurturing + agent alert)
Coming sprints roadmap
- WhatsApp and Instagram chatbot in client production (depends on Meta Cloud registration)
- Automatic distribution to Idealista + Fotocasa + Kyero + JamesEdition (depends on API credentials per agency)
- Full Auth.js magic-link (depends on Resend domain verification, pending human action)
- Calendly or Cal.com for automated demo booking (decision pending)
- Off-site encrypted backups to private GitHub (provider in decision)