AI Agent for a Travel Agency: Quote Less, Close More

A small travel agency that gets 80 enquiries a month builds around 45 quotes and closes about a dozen. Replace those numbers with your own, but the shape repeats everywhere: two to three hours per quote, roughly 90 hours a month poured into quotes that never closed. This is not a sales failure. It is what happens when the most expensive work in the business, building a price, gets done before anyone checks whether the enquiry has dates, a budget and a passenger count.

A travel quote is the most expensive product you give away for free

A family package to a nearby destination can be an ₪18,000 deal, yet the agency's gross profit on it is a small slice of that: say ₪1,600, a figure to swap for your real one. Against it sits a cost nobody writes down: three hours of an experienced agent who, at full employer cost, runs about ₪120 an hour. Every quote is ₪360 of labour, so 33 unclosed quotes are roughly ₪12,000 a month that never became revenue.

Worse, the quote is almost always built twice. An enquiry reading "looking for a package to Rhodes in August, how much?" is missing everything that sets the price: exact dates, children's ages, hotel level, board type, budget. The agent quotes on assumptions, the client replies "actually only the first week, and only full board", and the work restarts. That was not pricing but data gathering, done the most expensive way possible.

Nine questions to close before anyone touches the price

A travel brief is a fixed list, which is exactly why it can leave the human agent's head:

• Dates and flexibility. Not "August" but a range: two or three days of movement is often hundreds of shekels per person.
• Passenger count and children's ages. A child's age drives the price and the room type.
• Departure airport and flight type. Direct or with a stopover, and what gap the client will absorb.
• Budget per person. The question agents fear asking, and the one that saves the wasted work.
• Hotel level and board type. Rating, full board or breakfast, distance from the beach.
• Kosher and Shabbat requirements. A kosher kitchen, or a Saturday flight being out.
• Purpose of the trip. A wedding, a family holiday, a sports event.
• Documents. Passport validity, visa needs, anyone on a foreign passport.
• Whether a competing quote is on the table. And if so, what is in it.

An AI agent collects that list in a conversation rather than a form: one question at a time, in Hebrew, Russian or English, on WhatsApp or in the website chat, at any hour. A lead card then appears in the WhaleBiz CRM with an AI summary of the brief, so the human agent opens a brief rather than a chat. That is the job of an AI sales and lead-qualification agent: not to price for you, but to bring only the enquiries worth pricing.

Whoever answers first sets the frame for the comparison

Someone shopping for a family package usually writes to several agencies the same evening. The first real answer does not only win attention, it sets the language of the comparison: if you were first to ask about the children's ages and to explain why a direct flight costs more, every other quote is read against your frame. Whoever replies third explains themselves inside somebody else's definitions.

Be precise about what a first answer is. "Got it, we will get back to you" is not one. An answer is a real reply from approved material, a continuation of the brief, and a concrete call slot from the open calendar. We covered the same logic across channels in the article on AI agents for hotels and guesthouses, and it also sits behind the tourism and hospitality solutions page.

Four ways to handle a travel enquiry, and what each one costs

The row worth staring at is peak season: the human assistant breaks first, exactly when every enquiry is worth more. The AI agent is the only channel here whose response time does not change between January and August.

What the agent closes on its own, and what it never decides

Here is the uncomfortable part: in travel an overpromise breaks on day one. The agent is not connected to supplier booking systems. It does not see seats, available rooms or live prices, so it never confirms availability and never commits to a price. Issuing tickets and building the quote stay with a person.

What it closes alone is most of the traffic, the part eating the day:

• Repeat questions from approved material: what the package includes, baggage allowance, change and cancellation policy, when the balance is due.
• Completing the brief and screening out enquiries outside your range, so nobody burns a call on a dead end.
• Booking a scoping call in the right agent's calendar, with a reminder beforehand.
• Collecting documents: passport scans, names in Latin letters, expiry dates. It takes images, files and voice messages, says what is missing and attaches everything to the card, in the language the client opened in.

What it deliberately does not do: give binding advice on visas or document validity, pointing instead to the official source or a person at the agency; approve refunds or unusual discounts; give medical advice on vaccinations or traveller insurance; promise upgrades. In each case the conversation moves to a person with the full context.

The follow-up that turns "let me think about it" into a booking

In travel the decision window is long: a client gets a quote, checks whether the leave is approved at work, and disappears for two weeks. In most agencies nothing happens then, and a quote that cost three hours quietly goes cold.

A follow-up that works here leans on what the client said: the date they asked for is approaching, the question left open, the hotel they liked. Three rules set at setup make it an asset instead of a nuisance: every message refers to a detail from the conversation, there is a maximum frequency and a stop point after which the lead moves to a cold list, and any reply hands the conversation to a person. Follow-ups are unlimited and there are no trigger meters, so a well-built sequence does not raise your monthly cost. We took the same mechanism apart on an old database in the piece on the cold lead revival bot.

Peak season and disruptions: the day the inbox explodes

One overload is predictable: the months before the holidays and the summer. The other is not: a cancelled flight, a strike, a change in the security situation. On such a day more messages arrive in an hour than in several normal days, all urgent, all from clients who already paid.

That is when separating the traffic is worth money. The agent replies immediately to everyone, delivers the approved update the agency drafted, flags who is genuinely urgent by the signals you defined, for example someone waiting at an airport, and routes them first to the person on duty. Everyone else gets a real answer instead of silence. On volume, a reply to a voice message, an image or a file counts as 2 messages, so pick a plan against your busiest month.

Which plan this lands on

For an agency whose enquiries arrive on WhatsApp and that needs calendar booking, the entry point is Standard, ₪490 per month plus a one-time ₪1,990 setup: WhatsApp through the official Meta API, calendar booking, media up to 100 files, 2,000 AI messages a month, a knowledge base of up to 15 documents, and the WhaleBiz CRM with Excel export, within 14 business days. An agency whose traffic all arrives at a website widget or Telegram can start on Start, ₪190 per month plus ₪990 setup, as long as it is clear that it has no WhatsApp and no calendar booking. Several agents with separate calendars, Instagram and Facebook, or a deposit taken inside the chat require Business, ₪990 per month plus ₪2,990 setup, up to 21 business days.

Prices exclude VAT, the setup fee is paid in one payment before work starts, and the subscription begins on the day the agent goes live. Setup business days are counted from the moment we receive all your materials, so anyone who wants an agent working through the summer rush should start six weeks earlier. Ongoing changes after go-live, such as rewording an answer, are included in your plan's monthly change allowance. Full details on the pricing page.

Frequently asked questions

We have no fixed price list. How can an AI agent answer "how much does it cost"? It does not price. It asks for dates and flexibility, passengers and children's ages, departure airport, hotel level and board type, kosher requirements and budget per person. If you pre-approved ranges it gives them; if not, it says pricing is individual, books a call and hands over the brief.

Can the agent check real flight or hotel availability and book it? No. It is not connected to supplier booking systems and does not see seats or rooms in real time, so it never confirms availability or guarantees a price. It answers from approved material, collects the brief, books a call and gathers documents.

Clients send passport photos and voice messages. Does that count against the quota? What is counted is agent replies, and a reply to a voice message, an image or a file counts as 2 messages. Standard includes 2,000 messages a month with ₪0.30 overage; Business includes 5,000 with ₪0.25.

How does a weeks-long follow-up avoid becoming spam? Every message leans on a detail the client said, there is a maximum frequency and a stop point after which the lead moves to a cold list, and any reply hands the conversation to a person. Follow-ups are unlimited and there are no trigger meters.

Which plan do we start on and how long is setup? WhatsApp with calendar booking starts on Standard, ₪490 per month plus a one-time ₪1,990 setup, up to 14 business days. A website widget or Telegram only can start on Start, ₪190 plus ₪990 setup. Instagram or a deposit in chat requires Business, ₪990 plus ₪2,990 setup.