A flight can be cheap and still be completely wrong for the trip. It may arrive too late for a meeting, leave at an impossible hour, or stop being good value once baggage, a long overnight connection, or an airport change is part of the journey.
Apricot is built for that decision. It combines AI agents, live flight-market data, a trade-off engine, and preference learning to turn a travel brief into a small shortlist of flights you can actually choose between. Each recommendation comes with a reason — not just a rank.
It starts by understanding the trip
You describe a trip as you would describe it to another person: “I need to be in Berlin by Thursday morning for a meeting, I can return Friday evening, and I am taking a bag.”
The first job of Apricot’s AI is to turn that into a useful brief. It identifies the route, dates and practical requirements, but it also reads the intent behind them. A meeting makes a reliable arrival more important than saving a small amount. Returning on Friday evening may protect the weekend. Bringing a bag changes which fare is genuinely good value.
Those details become part of the decision from the beginning, rather than filters you have to discover and apply one by one after the results arrive. When an essential detail is missing, Apricot asks for it instead of quietly building the wrong trip around an assumption.
It evaluates the journey, not only the ticket
Once the trip is clear, Apricot searches current flight inventory and compares the details that make an itinerary work or fail: cost, departure and arrival times, total duration, stops, connection length, overnight layovers, airport changes, baggage inclusion, fare flexibility, airlines, and the preferences you have stated.
When you give the dates some flexibility, it also examines the viable alternatives around them. A different departure day may save money, avoid an awkward connection, or give you more time at the destination. The point is not to produce more combinations. It is to find the combinations worth considering.
This is where the product’s judgment lives. A $60 saving can be worthwhile on a relaxed trip; it may be meaningless if the cheaper itinerary arrives on the morning of an important meeting. A direct flight may deserve the premium for a two-day visit but not for a long holiday. Apricot does not pretend those trade-offs have one answer for everyone. It makes them visible and applies them to the trip you described.
A modern recommendation engine, not a static sort
Underneath the shortlist is a hybrid decision system. AI agents interpret the language of the trip and carry the conversation forward; live market data supplies the available options; and a structured recommendation layer compares each itinerary against the brief, your stated priorities, and the preference signals Apricot has learned over time.
It does not treat every fact as separate. The engine looks at how they change one another. A lower fare means something different when it removes an evening from a two-day trip. A short connection means something different when the next flight is the only way to arrive before an important event. A bag included in the fare can change the real comparison entirely. Rather than asking one opaque model to declare a winner, Apricot brings those relationships together and explains the decision in language you can challenge.
A shortlist should contain real choices
The system does more than sort every flight from best to worst. It also considers the shortlist as a whole, so that the options are useful beside one another.
Each recommendation earns its place by resolving the trip in a meaningfully different way: perhaps protecting an important arrival, preserving time at the destination, avoiding an unreasonable connection, or making a genuine saving without quietly creating a worse journey. It does not mean presenting three nearly identical flights at slightly different prices and calling that choice.
The recommendation can therefore be different from the cheapest itinerary and different from the mathematically shortest one. What matters is whether the option earns its place in the shortlist: it solves a meaningful version of the traveller’s problem, and Apricot explains why.
It learns your patterns without overruling you
Every trip is different, but people do develop habits. Over time, Apricot uses patterns from a signed-in traveller’s own searches — such as repeatedly favouring morning departures or avoiding long connections — as preference signals for future results.
That is not a profile built from other people’s behaviour, and it is not a substitute for what you say. Your current brief always takes priority. If you normally prefer direct flights but today say that price matters most, Apricot responds to today, not to an old assumption.
None of these parts is useful alone. Together, they make the recommendation more personal without making it mysterious.
The conversation continues after the shortlist
Three recommendations are a starting point, not a final verdict. If the first set does not feel right, you can say so in the same language you used for the trip: “None of these work — I need to arrive before nine,” or “show me a direct option with a bag.” Apricot distinguishes a request to adjust the search from a question about an option already on screen, and continues from the context rather than making you begin again.
You can also ask practical questions: whether a selected fare includes checked baggage, why one option was preferred, or how two itineraries compare. If you ask about a live fact such as availability or the latest price, Apricot treats that differently from a question about the earlier snapshot and refreshes the search when needed.
The goal is not to make flight search sound more intelligent. It is to give you a recommendation you can understand, question, and improve — until it feels right for the trip in front of you.
If a recommendation does not hold up for your trip, we would like to know: ask@flyapricot.com