An adaptive reasoning layer for the complete river-cruise journey.

A traveler may describe a journey in natural language rather than exact database fields:

“We're a couple in our 60s. We don't want anything too strenuous. We've already been to Amsterdam. We like wine and historical towns. Around $5,000 each. We have 10 days in October.”

This is a constraint-and-preference problem, not merely keyword search.

QUERY
   ↓
POLOXI

Hard constraints
├── 2 travelers
├── October
├── ≤ 10 days
└── ~$5K/person

Preferences
├── Wine and history
├── Easy/moderate activity
└── Avoid a repeat-heavy experience

          ↓
Candidate itineraries
Rhine · Danube · Rhône · Douro · Seine
          ↓
Evidence gathering
Dates · Price · Excursions · Activity levels
Availability · Promotions · Guest history
          ↓
Score → eliminate → refine → loop
          ↓
Best 3 journeys + explanation

An advisor asks:

“My clients sailed the Danube before. Anniversary couple, active but don't want strenuous excursions, love wine, about $12K total, traveling September. What should I propose?”

Client profile + Past cruises + Current inventory
+ Itineraries + Excursions + Promotions
+ Loyalty + Pricing
                ↓
             POLOXI
                ↓
Recommended cruise + Alternative + Why
+ Suggested excursions + Applicable benefits

POLOXI becomes an AI sales-intelligence layer, not merely a chatbot.

Cruise promotions can include eligibility rules, loyalty status, referral benefits, sailing-specific offers, availability constraints, category requirements, and incompatible combinations.

“What is the strongest valid offer this customer qualifies for?”

H1 → Standard promotion
H2 → Loyalty benefit
H3 → Referral savings
H4 → Solo offer
H5 → Early booking
H6 → Compatible benefit combination

Evaluate eligibility → test incompatibilities
→ compare customer value and commercial impact
→ explain the strongest valid offer

This is safer and more useful than allowing a language model to guess which promotion applies.

“I'm 68. I can walk but don't like steep hills. My wife loves museums. We both enjoy wine. What excursions should we take tomorrow?”

Guest A: moderate mobility · wine · history
Guest B: museums · history
             +
Tomorrow's port and available excursions
             ↓
Hard constraints + preferences + schedule conflicts
+ activity level + distance + availability
             ↓
Recommended itinerary

That is decision intelligence, not semantic search.

“We're arriving late tonight. We have the bike excursion tomorrow morning, but my husband hurt his knee. We also have dinner reservations. What should we change?”

Rather than returning an FAQ, POLOXI constructs an appropriate, policy-grounded plan across multiple connected domains.

“I want something like our Rhine cruise, but warmer, somewhere we haven't visited, preferably with wineries, and my wife doesn't want more than a seven-hour flight.”

An ordinary search system struggles because these are not exact database fields. POLOXI progressively narrows what “like Rhine” means.

H1 → European river experience
H2 → Wine-oriented
H3 → Historic towns
H4 → Similar duration
H5 → Similar activity level
          ↓
Evidence + customer history
          ↓
Candidate destinations
Douro · Rhône · Bordeaux · Danube · others
          ↓
Clarify → compare → narrow → recommend

POLOXI can continuously evaluate the most relevant next offer for a particular reservation.

Reservation + Guest history + Preferences
+ Inventory + Trip characteristics + Current offers
+ Loyalty + Previous purchases + Likelihood
                    ↓
                  POLOXI
                    ↓
              Best Next Action

The recommendation is contextual, evidence-backed, and constrained by inventory and business rules—not a generic “if customer X, offer Y” rule.

“Why are cancellations for this itinerary increasing?”

                    QUESTION
                       ↓
                    POLOXI
                       ↓
          ┌────────────┼────────────┐
          ↓            ↓            ↓
        Price       Experience   Operations
          ↓            ↓            ↓
 Promotions       Reviews       Disruptions
 Competitors      Excursions    Flights
 Value changes    Ship/route    Ports
          └────────────┼────────────┘
                       ↓
                 Gather evidence
                       ↓
               Reweight candidates ↺
                       ↓
                    Converge
                       ↓
              Root-cause analysis

The same architecture used to investigate claims denials or underwriting failures can investigate cruise cancellations, complaints, conversion changes, and operational deterioration. POLOXI remains domain-independent.