Move millions of people, gracefully.

A demand model that reads the signals arriving by the hour: flights and ferries booked into the area, the pace of bookings, rates and availability across the area, road traffic, events and what people are saying online. It forecasts demand night by night, recommends rates and capacity to the revenue manager, who approves them, and shows which signals moved each recommendation.

• Live signals read together: flights, ferries, booking pace, area rates and availability, traffic, events, online mentions
• Demand forecast night by night, with its uncertainty shown
• Rate and capacity recommendations, each with the signals behind it
• Nothing changes until the revenue manager approves it

An illustration. On a Tuesday in June a music festival is announced for the coming Saturday. Within hours the model sees ferry and flight bookings to the island climb, online mentions of the festival spike, traffic toward the port rise, and the island’s other hotels fill faster than usual while their average rate starts to move. By the afternoon it forecasts that Saturday will sell out, days before last year’s calendar would have shown it. On Wednesday morning it recommends raising the 80-room hotel’s Saturday rate from €180 to €230, and shows the signals behind the recommendation. The revenue manager approves it at 11:20. Saturday still sells out, at the new rate: in this illustration, 28 % more room revenue that night, €18,400 instead of €14,400.

A capture service that measures the streets and sites a visitor actually uses — steps, kerbs, gradients, surfaces — and computes step-free routes from those measurements rather than assuming them. The routes feed the itinerary agent and the entry desk, so an accessible plan is a verified one.

• Steps, kerbs, gradients and surfaces captured to measurements
• Step-free routes computed from the measurements, not assumed
• Routes shared with the itinerary agent and the entry desk
• Re-captured when a street changes

A wheelchair user wants to reach the castle from the harbour. The shortest line climbs three flights of steps. The captured model knows where every step, kerb and steep lane is, and returns a route that is longer — four hundred and twenty metres — but never steeper than five per cent and never interrupted by a step. It reaches the itinerary as a route that has been measured, not a promise.

An agent that holds a traveller’s plans as intent rather than a fixed schedule. It takes every request, books what it can against live availability and the forecast, asks the right person for what it cannot grant, and keeps one itinerary current in the traveller’s own language.

• Every request taken, however many arrive at once
• Booked against live availability and the forecast
• What it cannot grant asked of the right person
• One itinerary kept current, in the traveller’s own language

On Sunday at 18:00 a traveller with a party of eight sends six requests in a row: a private boat tomorrow, a table at the island’s busiest seafood taverna, a gluten-free birthday cake, a photographer at sunset, their Friday ferry moved to the afternoon, and a visit to the museum after hours. The agent reads the forecast first: wind force 7 after 13:00, so the boat is booked for the morning, with a licensed skipper. The taverna is full at 21:00, so it takes 21:30 on the terrace and has the cake delivered there. The photographer is booked at the castle for sunset, and the ferry moved to 16:00 with the transfer moved to match. The museum visit is not the agent’s to grant, so it asks the museum’s events office, which offers a private visit at 18:00 on Thursday, and the traveller accepts. Six requests, all resolved, in one thread and in their own language.

A guard on every traveller’s chain of legs — transfer, ferry, transfer, hotel, flight. When a sailing is banned or runs late, it re-plans the chain end to end, moving the transfer, holding the room and protecting the flight, and tells the traveller in the thread they already have.

• Every leg of a journey watched as one chain
• A banned or late sailing re-plans everything behind it
• Transfers moved and rooms held before the traveller asks
• The traveller told in the thread they already have

At seven in the morning the port authority bans small fast ferries for the day: the wind is at force 8. The 10:30 catamaran will not sail; a larger ferry at 14:00 will. The traveller’s morning transfer is moved to meet it, the next hotel is told to expect a late check-in, and the evening flight is checked and still made. The traveller wakes to one message with the new plan, not to a closed ticket office.

An orchestration engine for a port day. It takes the berth schedule and each ship’s shore-excursion plan, and staggers tenders, coaches and guides so arrivals at the busiest sites spread across the hours instead of landing at once. It re-solves when a ship runs late or tendering is suspended.

• Berth schedule and excursion plans solved as one day
• Tenders, coaches and guides staggered against site capacity
• Re-solved when a ship runs late or the sea closes tendering
• The destination sees the whole day before it starts

Three ships are due. Left alone, most of their guests would reach the old town between half past eight and half past nine, and the square would be over capacity before ten. The plan staggers the tenders and the coach departures, sends one ship’s excursions to the far side of the island first, and spreads arrivals across the morning. When the third ship reports forty minutes late, the afternoon re-solves around it rather than stacking it on top of the lunch peak.

Movement analytics for a resort’s public spaces. The cameras already in place are read on site and turned into anonymous counts per zone, never faces or identities, so the team sees where guests gather, when and for how long, and can change the space to match.

• Existing cameras read on site; only anonymous counts leave them
• Occupancy and time spent per zone, hour by hour
• Patterns across days turned into specific changes
• No face recognised, no guest identified, no individual track kept

A resort wants to know how guests actually use its public spaces. The cameras already in place are read on site and turned into anonymous counts per zone (lobby, breakfast room, pool and terrace) hour by hour; no face is recognised, no guest identified and no individual track kept. After two weeks the patterns are plain: the breakfast queue peaks at 08:30, the shaded loungers are full by 10:00 while the sunny ones stand empty, and the terrace overflows after 21:00 while the lobby bar is quiet. The operations manager chooses three changes: a second breakfast counter from 08:15, shade over twelve more loungers, and the evening music moved to the terrace. The following weeks are counted the same way, and each change shows in the counts.

A capacity desk for timed entry. It holds each slot’s capacity against live conditions — the queue at the gate, the heat, the ships in port — and when a closure is declared it moves the affected visitors into slots that are still open and tells them before they set out.

• Slot capacity held against the gate, the heat and the ships in port
• A declared closure moves affected visitors into open slots
• Every moved visitor told before they set out
• No slot sold past what the site can hold

A heat advisory closes the site from one until five in the afternoon. About three hundred visitors hold tickets for those four hours. Instead of a closed gate and a crowd outside it, each of them is offered the nearest open slot that morning or evening, in their own language, and the move is written to their ticket the moment they accept. The morning slots fill to their limit and no further.

An on-demand transport engine for an island. Requests from visitors and residents are pooled into shared rides with virtual stops, and the shuttle’s route is re-formed continuously as new requests arrive — the same allocation engine that runs transfers, applied to a whole island.

• Requests pooled into shared rides
• Virtual stops where people actually are
• Routes re-formed continuously as requests arrive
• Residents and visitors on the same service

At sunset everyone wants to reach the west coast at once. Instead of one timetabled bus that is full at the first stop, requests are pooled: a couple at the harbour, three guests at a hillside hotel, a family on the beach road, each met at a virtual stop a short walk away. The shuttle’s route re-forms with every new request, and in the quiet morning hours no vehicle drives an empty loop.

A coordination layer for the day everything has to move at once. It holds one live roster of every guest, hotel and vehicle, messages every affected traveller in their own language, allocates vehicles to the hotels inside the zone and routes them to muster points. The incident team releases every order, and every decision is recorded as it is made.

• One live roster of guests, hotels and vehicles
• Every affected traveller messaged in their own language
• Vehicles allocated to hotels in the zone and routed to muster points
• The incident team releases every order; every decision is recorded

A hazard zone is declared across the west of the island. Five hotels sit inside it. Within minutes the incident team sees how many guests each one holds, which vehicles are nearest, and a proposed allocation to the three muster points outside the zone. They release it. Every guest receives the instruction in their own language, drivers receive their pickups, and the board shows who has been collected and who is still waiting — for the people running the response, not reconstructed afterwards.