What problem does it solve?
A large airport asks passengers to find their way through unfamiliar buildings, in a hurry and often in a second language. The questions repeat all day: where is my gate, how far is it, where is baggage claim, is my flight on time, where can I eat before boarding, where is the lounge. Signage answers some of them, information desks and uniformed ambassadors answer the rest, and both are limited by staff numbers, shift patterns and the languages the staff speak. Queues at the desks build up when several departures leave at once, which is when passengers are least patient.
Airports have started to put AI in the places where these questions are asked. Aena, which operates the Spanish airports, has a virtual assistant that, according to a 2025 report on its startup program, reduces the load on information points and speeds up service to the passenger. LaGuardia Gateway Partners describes its hologram as a resource during peak periods, when terminal staff are busy with other guests.
Soul Machines says its Digital Iris avatar at Dallas Fort Worth was designed for travelers who fly less often and are less comfortable with online resources. Lyon-Saint Exupéry (VINCI Airports) tested a 3D avatar on two kiosks in 2025, and Boston Logan (Massport) began a pilot of a hologram concierge in 2026. These are statements of intent: no airport on this page has published how many desk contacts the assistant replaced.
The Dallas Fort Worth source is a 2022 pilot update and no later status was found, so it shows that the design exists, not that the kiosks are still running.
How does it work?
- Put the assistant where the questions start. A kiosk, hologram or screen sits at check in, in the concourse and at baggage claim, and the same assistant runs on the airport's website, app and WhatsApp so a traveler can ask before arriving. Aena's kiosks show a QR code that moves the conversation to its chatbot Oli on the app or WhatsApp, so the traveler keeps the answer while walking.
- Hear the question in the traveler's language. Passengers speak or type, and the system recognizes the language. A kiosk camera can let the assistant detect that someone is standing in front of it, as the Dallas Fort Worth avatar does according to its vendor. Miami's release says its assistant gives the same answer in any language across its web, WhatsApp, kiosk and hologram interfaces.
- Answer from live airport data. The assistant reads flight and gate information, the directory of shops and restaurants, security and service information and the airport's FAQ. Miami's data supplier normalizes real time operational data for it; Soul Machines says the Dallas Fort Worth avatar connects by API to flight and weather data.
- Show the way. For a gate, lounge or baggage belt the assistant draws the route on an indoor map. Aena links its Barcelona kiosks to Aena Maps so the passenger sees the route and the estimated walking time from the information point.
- Hand over to a person. When the assistant cannot answer, or the traveler prefers a human, it starts a video call with information staff (Madrid-Barajas runs these around the clock) or sends the traveler to an ambassador.
- Learn from what was asked. The airport reviews the questions the assistant could not answer and fixes the content, the map or the staffing at the desks.
- Audience
- Customer facing
- Autonomy
- Supervised agent
- Adoption
- Early adopters
- Channels
- Kiosk and branch, Digital human, Web chat, Mobile app, WhatsApp, Phone and voice
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
| KPI | Median | Reported range | Data points | Claimed by |
|---|---|---|---|---|
| Users served | Not pooled | at least 800,000 | 1 | 1 organization |
Value drivers: Customer experience, Lower cost to serve, Inclusion and access, Employee productivity.
Indicative value
An airport with 20 million passengers a year
USD 160,000 to USD 2.5 million
Information desk handling cost avoided per year
How this is calculated
Formula: infoContacts * assistantShare * costPerContact. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Information questions asked at desks, ambassadors and the contact center per year infoContacts, contacts per year | 400,000 | 1,000,000 | Editorial assumption, replace with your own desk tallies and contact center logs. |
| Share of those questions the assistant answers without a person assistantShare, fraction of contacts | 0.2 | 0.5 | Editorial assumption. No airport on this page discloses a containment or deflection rate. |
| Cost of a staff handled information contact costPerContact, USD per contact | 2 | 5 | Editorial assumption, replace with your own fully loaded desk cost per contact. |
What it leaves out: Gross handling cost only. It leaves out the kiosk or hologram hardware, content and map maintenance, the video call staff that remain, and the value to passengers of finding a gate faster or being answered in their own language.
Who already uses it?
7 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
Aena
Spain · Travel and hospitality · 2025
Aena, the Spanish airport operator, puts virtual assistant kiosks at its information points and links them to Oli, its multilingual passenger chatbot.
Madrid-Barajas installed nine assistants in terminals 1, 2, 4 and 4S in May 2023, with live video calls to information staff around the clock, and in May 2024 Barcelona-El Prat began a network of about 50 devices that answer questions about services, flight status, shops and the location of check in desks and gates, tied to the airport's indoor guidance app Aena Maps.
A QR code on the kiosk hands the passenger over to Oli on the Aena app or WhatsApp. The releases about the kiosks do not say they use AI: the Madrid screens offer video calls and FAQs, and the Barcelona kiosks answer from configured frequently asked questions.
The AI claim rests on Oli: a 2025 report on Aena's startup program says the chatbot, built by the startup Airport.ai and enriched with generative AI, runs in 8 languages across web, app, WhatsApp, the airport assistants, Facebook and WeChat at 42 airports of the network. Aena said in a June 2024 release about the Madrid-Barajas customer experience award that Oli had served more than 800,000 users since its launch.
- Users served: at least 800,000, Aena's Oli chatbot, since its launch, as said in a June 2024 release about the Madrid-Barajas award
"el chatbot de Aena que habla 8 idiomas y está presente a través de diferentes soportes en las terminales del Aeropuerto, proporcionando información en tiempo real y que desde su implantación ha atendido a más de 800.000 usuarios"
Claimed by: organization
Aéroports de Lyon (VINCI Airports)
France · Travel and hospitality · 2025
In April 2025 Lyon-Saint Exupéry Airport started a trial of a 3D conversational avatar on two interactive kiosks in Terminal 1, one in the departures area at Place des Lumières and one in the baggage reclaim area. The assistant answers spoken questions, shows an interactive airport map, gives live flight and gate information and suggests things to do in the Lyon region. VINCI Airports presents it as the first trial of its kind at a European airport and as a complement to signage, interactive terminals, the mobile app and staff. The initial version works in French and English, and the VINCI Airports release says the agent features around 30 languages.
No outcome disclosed.
LaGuardia Gateway Partners
United States · Travel and hospitality · 2026
LaGuardia Gateway Partners, the manager and developer of LaGuardia Terminal B in New York, put a life size AI hologram named Bridget near the terminal's food hall in May 2026. Travelers ask it in conversation where gates, shops, premium lounges and baggage claim are and receive live mapping and step by step directions. It speaks English and Spanish, has captions, high contrast displays and an interface positioned for wheelchair users, and is meant to work next to the terminal's human guest experience ambassadors, particularly at peak times. Additional units are planned for the two concourses. No usage figures have been published.
No outcome disclosed.
Massachusetts Port Authority (Massport)
United States · Travel and hospitality · 2026
In September 2026 Massport started a pilot of Amelia at Boston Logan International Airport, a life size AI avatar on a hologram display with a touch screen. Amelia gives directions to gates, restaurants, shops, lounges and baggage claim, shows live maps and TSA wait times, and can pass a traveler to live customer service staff. Her units sit in every terminal, before and after security, and work in English, Spanish and French at launch. The airport describes her as an addition to its customer service team, and expects about 100,000 uses in the first year, which is a forecast and not a result.
No outcome disclosed.
Miami International Airport
United States · Travel and hospitality · 2026
In April 2026 Miami International Airport and the indoor mapping company Mappedin launched the MIA Virtual Assistant, a set of conversational agents that give passengers the same location aware answers on the web, WhatsApp, mobile, kiosks and holographic avatars. Satisfi Labs supplies the web and WhatsApp agents and Hypervsn the holograms, Radiant Technology normalizes the real time operational data, and the system runs inside the airport's own Microsoft Azure environment using Azure OpenAI. Miami-Dade County described it as the first large scale deployment of geo aware, agentic AI at an airport. No usage, accuracy or passenger satisfaction figures have been published, and local coverage questioned its priority against the airport's broken elevators and escalators.
No outcome disclosed.
Egyptian Holding Company for Airports and Air Navigation
Egypt · Travel and hospitality · 2025
In October 2025 the Egyptian Holding Company for Airports and Air Navigation launched Ask Mariam (اسأل مريم) at Cairo International Airport, working with Orange Egypt, and described it as the first interactive virtual assistant powered by AI in Egyptian airports. Travelers and visitors use interactive devices in Arabic or English to get answers on flights, travel and arrival procedures, the location of services in the airport, transport and nearby tourist sites. The holding company said it plans to bring the experience to Sphinx International Airport when the Grand Egyptian Museum opens. No usage figures are disclosed.
No outcome disclosed.
Dallas Fort Worth International Airport
United States · Travel and hospitality · 2022
In June 2022 Dallas Fort Worth International Airport (DFW) began a pilot of Digital Iris, a digital person on information kiosks in Terminal B. Soul Machines created the avatar's appearance and animation, and IBM Watson Assistant provides the speech recognition, language understanding and answers, connected by API to flight and weather data. Digital Iris answers by voice, without touch, about flights and gates, restaurants, the airport map and airport services, and was designed for infrequent travelers who are less comfortable with online resources. The October 2022 update said it had handled thousands of conversations and that DFW planned to extend it through the airport. No later report on that expansion was found.
No outcome disclosed.
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- Live flight, gate and baggage belt data from the airport operational database
- A floor plan with points of interest that can be routed
- A current directory of shops, restaurants, lounges and services with opening hours
- Approved answers on security, customs, transport links and passenger rights
Systems to integrate
- Flight information or airport operational database
- Indoor mapping and routing service
- Retail and restaurant directory
- Kiosk or hologram hardware and its content management
- Staff video call or ticketing tool for handover
- Web, app and WhatsApp channels for the same assistant
Complexity: Medium
The build needs a live feed from the airport operational database for flights and gates, a maintained indoor map, a directory of shops and restaurants with opening hours, kiosk hardware with a microphone, speaker and camera that meets accessibility rules, and a way to reach staff by video. The assistant only reads from these systems, so there is no transaction risk, but each source needs an owner who keeps it current.
- 1
List the questions and where they are asked
Count the questions at the desks, the contact center and the ambassadors for four weeks. Rank them by volume and by location, and place the first kiosks where the same few questions queue the desk.
- 2
Connect the live data before the model
Hook up flights, gates, belts and the directory first, and test that the assistant quotes them correctly and says when a feed is down. A fluent answer built on stale data is worse than no assistant.
- 3
Build the route and the handover
Add indoor routing with walking times and a video or ambassador handover. Test with passengers who have reduced mobility and with the languages your top ten origin markets use.
- 4
Pilot at one terminal with desk staff in the loop
Start with one hall, compare the desk queue before and after, and sit staff next to the kiosk for the first weeks to catch wrong answers.
- 5
Widen to more terminals and to the phone
Add locations, then the website, app and WhatsApp, and review unanswered questions weekly.
Guardrails
- Answers on flights, gates and times only from the live operational feed, with a visible timestamp
- A refusal and a pointer to a desk for anything about visas, customs rules or rebooking that the content does not cover
- Disclosure on the screen that the passenger is talking to AI, and no storage of voice or camera images beyond the session
- Accessible design with captions, high contrast, a reachable height and an audio option
KPIs to instrument
- Questions answered without handover, by location and language
- Desk queue length and wait time before and after the kiosk opened
- Share of answers that were wrong or stale in a weekly sample
- Handover requests and refused questions, by topic
- Passenger satisfaction at the kiosk, by language
Human in the loop
Ambassadors and information staff stay at the desks and take every handover, every question the assistant refuses, and every passenger who prefers a person. A named owner reviews a sample of conversations each week and approves content changes.
Common failure modes
- Stale flight or gate data
- A late gate change is not in the feed and the assistant sends passengers to the wrong place. Show the data timestamp and fall back to a desk when the feed is older than a few minutes.
- A kiosk nobody uses
- Passengers walk past a unit that is off their path or does not say what it can do. Place units where the desk queue forms, show example questions on the screen and measure use per unit.
- Weak recognition in a noisy hall
- Speech recognition fails near announcements and crowds. Offer touch input next to voice and test in the real hall at peak time.
- Language coverage that is announced but not tested
- A long language list hides poor quality in a few of them. Test each language with native speakers and show only the ones that pass.
What are the risks and rules?
EU AI Act
Depends on design
A kiosk that only answers questions is limited risk: Article 50(1) requires that passengers are informed they are interacting with an AI system, unless that is obvious from the context. The camera function decides the tier.
Annex III point 1 lists AI systems intended for emotion recognition (point 1(c)) and remote biometric identification (point 1(a)) as high risk, so an assistant that reads emotion from the camera, which Soul Machines describes for the Dallas Fort Worth avatar as "emotional detection and response", or that identifies travelers by their face, is a high risk system with the provider and deployer duties that follow. Keep both out of the design.
Article 50(3), the duty to tell people that an emotion recognition system is used on them, is only an additional disclosure and does not lower the tier.
Rules that apply
Guidance
- Article 50, transparency obligations for providers and deployers of certain AI systems (European Union, Europe). People must be told when they interact with an AI system, and when an emotion recognition system is used on them.
- Annex III, high risk AI systems referred to in Article 6(2) (European Union, Europe). Point 1 lists remote biometric identification and emotion recognition systems as high risk, which a kiosk camera must not be used for.
- Directive (EU) 2019/882, European Accessibility Act (European Union, Europe). Sets accessibility requirements for self service terminals, including interactive terminals that provide information, which covers airport information kiosks in the EU.
Controls to put in place
- AI disclosure on the screen before the first answer
- Inventory entry for the assistant with an accountable owner and a list of data sources
- No retention of voice or camera data beyond the session, and a documented legal basis if any is kept
- Accessibility test with users who have reduced mobility, low vision or hearing loss before each new unit goes live
- Weekly sample review with a record of corrections
Frequently asked questions
- What do airport passenger assistants answer?
Flights and gates, directions to lounges, shops, restaurants and baggage claim, and general airport services. Aena's Barcelona kiosks answer questions on services, flight status, shops and the location of check in desks, gates, lounges and baggage belts, and show the route with an estimated time. The sources describe information and wayfinding only, not rebooking or claims.
- How many passengers use them?
Aena said in a June 2024 release about the Madrid-Barajas award for customer experience that its chatbot Oli had served more than 800,000 users since its launch, on all its channels and not only on kiosks. Miami and LaGuardia Terminal B have not published usage numbers for their kiosks and holograms. Soul Machines, the vendor of the Dallas Fort Worth avatar, said in October 2022 that Digital Iris had handled thousands of conversations since its June pilot launch.
- Do the assistants replace information desk staff?
The airports describe them as additions. LaGuardia Gateway Partners says its hologram is meant to complement the human ambassadors, and Madrid-Barajas runs its kiosks with a live video call to information staff around the clock. No airport has published a reduction in desk staff or contacts.
- Is an airport kiosk assistant high risk under the EU AI Act?
Not when it only answers questions. It then falls under the transparency duty of Article 50: passengers must know they are talking to AI. It is high risk if the kiosk recognizes emotion from the camera or identifies people by their face, because Annex III point 1 lists both. A design like the Dallas Fort Worth avatar, which its vendor says includes "emotional detection and response", would fall there. Decide on the camera function early and keep both out.
How to cite this page
Blits.ai AI Use Case Library, "AI assistant for airport passengers on kiosks, holograms and chat", last verified 14 October 2026, https://www.blits.ai/ai-use-cases/airport-passenger-assistant. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 14 October 2026: First published