PLATFORM

The AI orchestration layer for hospitality.

D3x sits above your fragmented hotel stack. Agents resolve guest requests, execute against PMS and operations systems, and run on every channel — with codified hospitality logic underneath.

Book demo See customer proof

+150K

Messages per month

+10K

Hours of voice handled

120x

Faster median response (15m → 7s)

agent.d3x.ai/orchestrationLive

Channels

Guest reaches your property on any channel

WhatsApp
OTA inbox
Email
Voice
Web chat
Social DMs
SMS

Inbound

Guest message received

WhatsApp · room 512

“Hi, I need extra towels in room 512 please.”

D3X Agent Runtime

Agent Core

Reasoning → Routing → Execution → Guardrails

Context & Memory

Conversation history
Guest profile
Reservation data
Property knowledge
Brand voice

Policies & Guardrails

Confidence thresholds
Escalation rules
Human-in-the-loop
Brand voice rules
Data privacy rules

Observability

Tracing · Metrics · Logs · Alerts

Fetch

Retrieve context from systems

PMS · Mews

reservation.find({ room: 512 })
→ guest: John Smith
→ arrival: 2025-05-20

Context retrieved

Integration Fabric

API / Webhook — PMS, ops, messaging, and helpdesk

PMS

MewsOpera CloudCloudbedsApaleo

Housekeeping & Operations

FlexkeepingOptiiSnapfixSweeplyRoomChecking

Messaging & Channels

WhatsAppBooking.comGmailOutlookInstagramMessengerWebchat

CRM & Guest Experience

BookboostLike MagicP3Freshworks

Helpdesk

FreshdeskZendesk

Table Management & Other

AlkimiiFaunditPortalResDiary

Action

Execute actions across systems

Housekeeping · Optii

housekeeping.ticket.create({
  room: 512, type: 'extra_towels'
}) → HSK-88421

Action executed

Respond on guest channel

“Hi John, we've sent extra towels to room 512. Anything else? — Your stay team”

Reply sent

Full Context & Decision Logs

Every step logged with reasoning, inputs, outputs and outcomes.

Self-Learning Knowledge Base

Property knowledge self-learns and self-heals from conversations, corrections, and live stack data.

Feature Learning Loop

Resolved interactions feed back into skills, policies, and routing — the platform improves with every cycle.

Human-in-the-Loop

Auto or review depending on confidence, channel and policy.

Audit & Compliance

Immutable logs, data retention policies and access controls.

Confidence Scoring

Every action carries a confidence score; sub-threshold paths escalate or queue for human review.

THE PROBLEM

Hotel tech stacks were never built to orchestrate.

PMS holds inventory. Housekeeping holds tasks. CRM holds guest data. Channels hold conversations. Each system works in isolation — guest intent gets lost between them.

PMS · inventory

Housekeeping · tasks

CRM · guest data

Channels · conversations

GapNo shared orchestration

Generic AI can reply. It cannot route work across this patchwork. D3x sits as the orchestration layer above it — one runtime, one audit trail, live integrations.

D3x sits as the orchestration layer above this patchwork.

HOW IT WORKS

Four layers. One platform.

LAYER 01

Skills Engine

Codified hospitality domain logic — reservation matching, VIP escalation, brand voice, and hundreds of operational edge cases.

Booking.com _1 suffix matched · housekeeping ticket created · audit logged

LAYER 02

Agent Runtime

The orchestration core — reasoning, routing, policy enforcement, and execution with full audit trails on every guest action.

Intent classified · rules applied · integration called · outcome logged

LAYER 03

Channel Agents

Messaging, voice, and email — each configured with guardrails, escalation policy, and human-in-the-loop where it matters.

WhatsApp intent parsed · PMS fetch · action executed · guest reply sent

LAYER 04

Integration Fabric

25+ live integrations across PMS, housekeeping, CRM, OTA, and messaging. REST and webhook-driven with per-integration observability.

reservation.find() · ticket.create() · CRM update · channel reply

Inside the Skills Engine How agents run See the agents Browse integrations

LEARNING LOOP

A self-healing knowledge base for hotel operations.

Every unresolved question, low-confidence answer, escalation, and staff correction becomes a signal. D3x does not blindly retrain on it — it diagnoses the gap, proposes the smallest safe update, and routes it through review before it changes production behavior.

01 CAPTURE

Catch drift in real conversations

The loop watches missed intents, stale policies, failed retrievals, escalations, and repeated staff edits across messaging, phone, and email.

02 DIAGNOSE

Find the root cause

Was the answer missing from the knowledge base, buried in the wrong document, contradicted by a PMS rule, or too risky to automate?

03 REPAIR

Draft a safe knowledge update

D3x proposes policy snippets, source ranking changes, new examples, and guardrail edits instead of turning every bad answer into training noise.

04 APPROVE

Human review before production

Operators approve the diff, then the change is versioned, tested against similar questions, and logged for audit.

Signals in

feedback.d3x.ai/knowledge

Live loop

Escalations

When a human takes over, the reason becomes structured feedback.

Corrections

Staff edits reveal missing policy, tone, or source context.

Retrieval gaps

Low-confidence answers expose stale or unreachable knowledge.

Knowledge that heals where operations actually break.

Static help centers go stale. Hotel operations change every week: access codes, refund windows, invoice rules, property exceptions, housekeeping vendors, and channel quirks. The D3x loop keeps the operational knowledge layer current without giving the AI permission to improvise.

+ add missing policy snippet

~ rerank source: PMS policy before FAQ

! requires operator approval

Publish with control

Diff-style approval for every proposed policy or knowledge change.
Regression checks against similar guest questions before release.
Versioned source history so teams know why the agent changed.

SKILLS ENGINE

The codified hospitality logic that lets agents act, not just talk.

When a Booking.com reservation enters your PMS through the channel manager, the reservation ID gets an _1 appended. A generic AI workflow fails to match the booking. D3x knows — and hundreds of quirks like this, codified across every PMS and channel manager.

Skill definition

> reservation_id: BK-48291_1> pms_match: SUCCESS> action: housekeeping.ticket.create

Generic AI

Stops at the answer.

Output

Replies to guest

Cannot match reservation
Cannot create ticket
Cannot escalate VIP

D3x agents

Answers, then executes.

Live action

Output

Replies to guest

Matches reservation
Creates housekeeping ticket
Escalates VIP per policy
Logs to audit trail

Hundreds of rules. Years to replicate. Compounds.

Inside the Skills Engine

AGENT RUNTIME

What an agent is at D3x.

A configured workflow tied to a channel, knowledge sources, integrations, guardrails, and an escalation policy — not a chatbot wrapper.

input

Input

Channel message or call

rules

Rules

Skills Engine + guardrails

integrations

Integrations

PMS · housekeeping · CRM

output

Output

Guest response + system action

audit

Audit

Decision rationale logged

INTEGRATION FABRIC

Live in your stack.

25+ integrations across PMS, housekeeping, channel managers, OTA, table management, CRM, and messaging. Average build from API access: 3 weeks.

PMS

Mews

Opera

Cloudbeds

Apaleo

HOUSEKEEPING

Optii

Flexkeeping

Sweeply

RoomChecking

Snapfix

MESSAGING & CHANNELS

Booking.com

Website Chat

SMS

Twilio

Outlook

Gmail

CRM & GUEST EXPERIENCE

Bookboost

Like Magic

HELPDESK

Freshdesk

Zendesk

OPERATIONS

Alkimii

Faundit

Portal

See full integration list

CONTROLS

Autonomous where it should be. Human-led where it matters.

Late checkout until 2pm for €35?

ApproveEdit

Awaiting staff review

CONTROL 01

Human-in-the-loop

High-stakes channels operate as co-pilot. Staff review before send. Per-channel toggle.

Riverside Hotel

FormalWarmConcise

Response warmth

CONTROL 02

Brand voice configuration

Per-property tone, vocabulary, and response style.

WhatsAppLive
VoicePhase 2
EmailPhase 2

CONTROL 03

Phased deployment

Channel-by-channel rollout. Property-type guardrails.

12:04:01guest.authenticate
12:04:02reservation.match
12:05:02folio.charge.add

CONTROL 04

Audit logs

Full conversation history with decision rationale, queryable for review.

Agent controls dashboard

Brand voice configuration

Audit log view

GDPR Compliant (today)·SOC 2 Type II (in progress, Q3 2026)·EU-hosted infrastructure·Data not used to train third-party models

Trust & Security

IN PRODUCTION

Live

Measured in production, not projected on a roadmap.

Throughput & efficiency

+150K

Messages per month

+10K

Hours of voice handled

120x

Faster median response (15m → 7s)

60%

Containment · zero human touch

+95%

Volume growth · no headcount increase

+20

Hours saved per team member per month

Teams & reach

2,000+

team members using D3x

60+

hotel groups

countries D3x is used in

See proof

TALK TO US

30 minutes with the founder.

A working conversation about what AI orchestration looks like at your group's scale — not a generic product tour.

Book demo Read customer proof