There are three types of AI, and roughly a dozen names for each. AI-powered, agentic, predictive, conversational, hyper-personalized. Some mean something. Some are marketing adjectives wrapped around a rules engine. The terms blur together, but underneath the labels, there are only three. Here is each one, and where it fits.
The three types of AI
1. Predictive AI (machine learning)
A model trained on historical data to score, forecast, or categorize. It creates nothing new; it makes a confident guess.
- Travel example: fare-drop prediction or trip-disruption modeling.
- Events example: attendance forecasting, or matchmaking that pairs a delegate with the right session or exhibitor.
2. Generative AI (LLM-based)
A large language model that produces new content from a prompt: text, images, sometimes audio. Fluent, often convincing, occasionally wrong.
- Travel example: summarizing a support call, or a natural-language query against reporting data.
- Events example: drafting speaker bios or a venue RFP response.
3. Agentic AI
A generative model given tools, memory, and a goal. It works in multiple steps to finish a task without approval at each one. The newest, and the loudest right now.
- Travel example: booking that takes “book me Tuesday to Friday in Berlin” from prompt to confirmed itinerary.
- Events example: sourcing that takes “find a venue for 150 people in Berlin in October” to shortlist, approvals, and booking.
What a TMC means when it says AI
The vocabulary you hear in corporate travel, and where it lives.
- Predictive AI (machine learning)
- AI matchmaking or recommendation: pairs a user with an option, a fare, hotel, or route, from their history and preferences.
- Predictive intervention: flags a probable issue, a policy breach, compliance drift, or disruption risk, before it becomes one.
- Generative AI (LLM-based)
- Natural-language query: you ask a plain-English question (“what did we spend on New York flights in Q1?”) and get an answer with cited data.
- Virtual assistant or chatbot: generative AI wrapping a knowledge base and a fixed set of tools, answering common questions and resolving simple requests.
- Real-time coaching: listens to a support call and prompts the human agent with scripts, policy, or traveler history. It makes the agent faster, not traveler-facing.
- Agentic AI
- Autonomous rebooking: rebooks a disrupted itinerary without waiting for the traveler to ask.
How the vocabulary shifts for meetings and events
Same three types, different names. If you run events alongside your travel program, these are some of the terms you’ll hear.
- Predictive AI (machine learning)
- Attendee matchmaking: the event version of the recommendation layer, suggesting connections between attendees, exhibitors, and sessions, with automated hosted-buyer scheduling.
- Generative AI (LLM-based)
- Immersive or AI-generated environments: generative AI producing 3D event spaces, mapped video, or interactive brand experiences.
- Agentic AI
- Agentic operating system: a platform reframing its whole architecture around agentic AI.
None of this tells you which AI to buy. That is deliberate. The three names are a filter, not a shopping list. Once you can place a capability in one of them, the marketing loses its grip: you stop buying “AI-powered” and start buying a model that predicts, generates, or acts. The vocabulary will keep multiplying. Learn the three, and you can place whatever the industry names next.