The audit

One day, four sessions, everyone who touches a booking.

4

sessions

Director, direct sales, product and CRM, marketing

12

people

Hichem, Darja, Ana, Ivana, Marija, Nenad, Romina, Jadranka, Lea, Kika, Nikolina, Djordje

3

countries

Ljubljana in the room, Zagreb and Belgrade online

1

answer

Different desks, different jobs, the same obstacle

You chose the three to five interview package. We covered twelve people in four sessions, which is the largest one. We have not charged the difference and we are not going to.
20 August 2026
What everyone said

“If we were able to very quickly get any information that is currently in BlueVendo.”

Kika, marketing, Ljubljana, asked what the single biggest thing would be

Everything you need is in BlueVendo. Nobody can get it out except by clicking.

The same sentence, from seven desks

Ivana, Zagreb

Opens BlueVendo for every inquiry to find which hotel is free and which one fits.

Jadranka, Belgrade

Opens it for every invoice and clicks each passenger to read four prices.

Nenad, Belgrade

Exports and reshapes passenger, roaming and visa lists for every flight, two airline formats.

Kika, Ljubljana

Retypes prices into PDFs and HTML tables. Again every time prices move.

Nikolina, Zagreb

An hour a week finding which hotel has a reduction for which booking window.

Jadranka, again

Built a whole season of departure against last year tables by hand.

Marija, Belgrade

Reads the system live during a call to tell an agency whether seats are left.

Hichem

Occupancy sheets filled in by hand every two to three days. Sometimes daily.

This is one bottleneck wearing eight faces, not eight problems. It is also why one solution pays three times here, once in each company.
Pain 1 · the guest decides in ten minutes
24 to 48 h
your answer in season
5 to 15 min
what it needs to be
  • The same guest sends the same request to four or five agencies.
  • Inbound converts at 40 to 50 %. The winner is whoever answers first.
  • A booking from your website is not a booking. It lands in Outlook and waits for a free desk.

“He is there that five minutes, that ten minutes. If you are one hour later, maybe he looked somewhere else. If you answer him immediately, he is for you that moment.”

Hichem M. Allouche, 20 August 2026
Not just our opinion. MIT and InsideSales: a lead answered within 5 minutes is 21 times more likely to qualify than one answered after 30. Harvard Business Review: within an hour, 60 times more likely than after 24 hours.
Pain 1 of 5
Pain 2 · you cannot see how a charter is filling

By the time a flight is visibly empty, the only tool left is a discount.

  • Occupancy sheets are filled by hand every two to three days, sometimes daily, in three offices.
  • There is no pace against the same window last year. Jadranka built one season of it manually.
  • A departure filling faster than last year is a price you could have raised and did not.
  • A departure that froze is found on Friday and ends as a last minute at a loss.

“We are blocked with Excel sheets and an old system. We lose two days to change the prices. And then the plane is almost full and I cannot increase the prices immediately.”

Hichem M. Allouche, 20 August 2026
This is the single most expensive gap in the business. The charter is already paid for. Every seat priced wrong, in either direction, comes straight off the margin.
Pain 2 of 5
Pain 3 · half your business waits in the same queue
What a sub-agent is. You are the tour operator: you rent the aircraft and hold the hotel contracts. A sub-agent is a retail travel agency (Palma, Relax, Intelekt in Slovenia; Big Blue, Travelland, Jolly, Rapsody in Serbia; Uniline in Croatia) that sells your package to its own customer and takes a commission. They are business customers, not travellers.
  • More than half of Slovenian volume and about 40 % of Serbian comes through them.
  • The network is 100 to 150 agencies.
  • They email a request and need an availability answer now, or the seats drift.
  • Stop sale goes out as a translated email written by hand to every one of them.
193

reservations in one week

Belgrade, read live during the session

140

of them from the network

Only 53 were direct guests

“They sell until we send them a stop sale. We receive the email from the hotel, we translate it, prepare an email, send it to the agency. It is a lot of writing.”

Hichem M. Allouche, 20 August 2026
Pain 3 of 5
Pain 4 · complaints, answered in the evening

“It is not even about the time. It is about energy. Because complaints are a little bit toxic.”

Nenad Popović, Belgrade, 20 August 2026
  • Rising every year, in all three markets.
  • Clients now write them with AI, so they arrive flawlessly argued.
  • Every reply must hold under consumer and tourism law, so nobody junior can write one.
  • Marija and colleagues answer them after work, on the phone together, because the day is full.
What this becomes

A draft in ten minutes instead of two hours in the evening.

A private assistant loaded with the relevant law, your general conditions and every reply your team has written before. It reads the complaint, checks what is actually owed, and drafts the answer in your tone and your language.

It never sends. Marija reads it, changes a word, presses send. The hard ones still go to your lawyer.

This sits inside Element 1, the AI Foundation, not in the CRM. It is a project inside the assistant the people who answer complaints already have open.
Pain 4 of 5
Pain 5 · where the week goes

Sixty five hours a week, across fourteen people, spent looking things up.

Invoicing, BelgradeJadranka, measured: "one day a week"
8 h
Invoicing, Zagrebmeasured: half a day to a whole one
6 h
Checking availability per inquiryestimate: 8 sales people, 40 min a day
26,7 h
Occupancy sheets by handestimate: 3 offices, every 2 to 3 days
6 h
Complaintsestimate: 2 people, evenings
6 h
Operational lists per departureestimate: 7 flights, two airline formats
4,7 h
Supplier price controlestimate: Jadranka, when there is time
3 h
Price tables and ad research1 h measured, Nikolina
2,5 h
Sending lead offers by handmeasured, Ljubljana
1,5 h
64,4 h

every week, group wide

16,5 hours of that is measured in their own words. The rest is our estimate, built from what they described, and marked as such.

1,5

full time people

2.576 hours a season

11 %

of your total capacity

of fourteen people

Pain 5 of 5
What that actually costs you

The wages are the small number.

Direct cost
31.000 EUR

a year in wages

2.576 hours at a blended fully loaded 12 EUR. Built from your own published salary data in all three countries.

Opportunity
1,5 people

of capacity, released

Enough to carry roughly 11 % more volume without hiring anyone. That is the goal you set, not one we invented.

The real one
the answer
that arrives late

and the seat priced wrong

Both are revenue, not cost. Both are on the next page.

Group figures from the public registers: about 6,7 mio EUR revenue and about 500.000 EUR profit across the three companies, on fourteen people. Oasis Turs alone grew 48 % last year on 3,81 employees.
Bizi.si, Fininfo.hr, APR
Element 1 · for your team

AI Foundation

  • One business organisation, you as administrator, all three companies under it, every domain added.
  • Two to three hours one to one with each person, in the room. Not a group webinar.
  • A system prompt written for their job, in their language, with your tone and your rules in it.
  • Integrations to mail, M365 and OneDrive, and to BlueVendo through the browser.
  • Projects for their real tasks: complaint replies under law, supplier offer emails, price tables, passenger lists, the annual plan.
  • Two months of email support and a monthly call for questions after handover.

What you also get, which is not software

  • A written internal AI usage policy: what may go into a model, what may not, which account, who approves.
  • A training certificate for every person, which is what the EU AI Act asks you to be able to show.
Licences are yours, not ours. You buy them directly from Anthropic, in your own name. We do not resell them and we do not mark them up, so nothing here is hidden margin on someone else's software.
Today twelve of your people work on personal accounts with company data in them. That, not the tool, is the exposure.
Element 1 of 3
Element 1 · why this is not optional

Two articles of the EU AI Act already apply to you.

Article 4 · in force since February 2025

AI literacy

Any company whose staff use AI at work must ensure they have sufficient AI literacy, and must be able to demonstrate it. Twelve personal accounts with no training and no policy is the opposite of that.

Article 50 · in force since 2 August 2026

Transparency

A system that talks to a customer must disclose that it is AI. This applies directly to the booking assistant in Element 2. We build it that way from the first line.

The penalty ceiling for these obligations is 15 million EUR or 3 % of worldwide annual turnover, whichever is higher. For a group your size that is around 200.000 EUR. In Slovenia the supervising authority is AKOS.
What we deliver against it: documented training per person, a signed usage policy, a record of which systems are used for what, and the disclosure built into the assistant. Not a promise, a folder.
Element 1 of 3
Element 2 · for your guest

AI Booking Assistant

Live demo
  • Ten clicks, or one sentence spoken. You asked for the microphone yourself: people do not want to type.
  • Two options at the end, never three. Your rule, your words.
  • Live availability and price read from the booking system, not a brochure.
  • One personalised link, valid 24 hours, sent as the automatic reply on email, Instagram and Messenger. One build, every channel, which is what you chose over ten separate builds.
  • Anything unusual goes to a person, with the offer already drafted. A child price, an upper floor, a birthday. Your sales team raised this and it is answered in the design.
  • Three markets, three languages, on all three websites.
Also the answer to the sub-agent queue. The same engine answers agency emails with live availability and a net price, in their language, in seconds. The reservation itself stays a human click.
Element 2 of 3
Element 2 · the important distinction

This does not replace your booking engine. It stands in front of it.

What you have today

The BlueVendo search on your website

  • The guest must already know the destination, the dates and the party.
  • It answers one question: is this available and what does it cost.
  • It cannot handle "somewhere warm in September with a water park for the kids under a thousand each".
  • Anyone who does not fit the form writes an email instead, and waits 24 to 48 hours.
What we add on top

An assistant that asks, then recommends

  • It reads the same BlueVendo data, so prices and seats are identical. Nothing is duplicated.
  • It handles the guest who does not know what they want yet, which is most of them.
  • It picks two and explains why, instead of returning a list of twenty.
  • It captures the email and hands the office a finished lead, not a blank inquiry.
Your booking engine keeps doing its job. Every reservation still lives in BlueVendo, and moves to the new system with it in 2028.
Element 2 of 3
Element 2 · why we think this moves the number

You have already run this experiment once.

Your own proof. You replaced the website a year ago. Marija: since then 60 to 70 % of inquiries arrive through the site instead of by plain email. Nikolina: around 90 % of website inquiries are finalised. Djordje: the new site shows real time availability and the situation improved a lot.

Better front end, more direct bookings. That is the same lever, one step further.
Where you are today: 15 % of bookings online. Where you said you want to be: 70 %. Every point of that is a booking that costs nobody an hour.
Published benchmarksEffect
Chatbot assisted package booking against self service web booking30 % vs 10 %
Expedia and Booking.com, reported lift in booking conversion with AI agentsover 30 %
Lead answered in 5 min against 30 min, likelihood to qualify (MIT and InsideSales)21×
Lead answered within 1 h against 24 h, likelihood to qualify (HBR)60×

We do not use these numbers in the return calculation. The calculation later uses two percentage points, which is far below every one of them, on purpose.

Element 2 of 3
Module 1 · for Hichem

The two numbers you look at before anything else, every morning.

  • Every departure, filled against the same window last year. Not how full it is. How fast it is filling.
  • Frozen departures surface by themselves. No new booking for three or four days, with the reason next to it.
  • Fast departures too, which is the money you currently leave on the table. Raise 30 on the remaining seats.
  • A price recommendation with the reasoning, and one button to apply it.
  • Payment risk by sub-agent: who used to pay ten days before departure and now pays two.
  • The Monday digest, one email at seven: the two departures that need a decision this week.

“That flight is not selling for one week or for three days. Nobody booked a new reservation. I need to have an alert. This is like frozen, something is wrong.”

Hichem M. Allouche, 20 August 2026
You also asked for the opposite case, and it is the one nobody ever builds: “if the AI can read that a flight is filling up faster than the others, it needs to detect it and say, you need to increase the prices.”
Module 1 of 4
Module 2 · for Darja, Ana, Ivana and Marija

Inquiry, offer, reply, payment. In one place, for three offices.

What it shows

  • Every inquiry and where it came from
  • Which offers went out and which are still silent
  • Who confirmed and has not paid, with the hours left on the option

What it does by itself

  • Reminder at 24 hours, another at 72
  • Then a call list, with a name and a number
  • Matches your daily bank export to reservations, so paid disappears from the list

Why they asked for it

Marija: “That is the statistics that we need to see directly, so that we know where to correct things.”

Ivana: the follow up happens today too. It happens by phone, in the evening, from memory.

72 hour options are already how you work. This module does not change the rule. It just stops the rule from living in somebody's head.
Module 2 of 4
Module 3 · for Nenad, Marija and the network

Half your volume, answered at the speed of the website.

  • Every agency request is read, checked against the charter and the hotel allotment, and answered in the agent's own language with a net price and one alternative.
  • The reservation stays a human click. The button is disabled until passenger data is attached, and it says so on the screen.
  • Stop sale to the whole network in one action. Hotel email in, translated message out to 140 addresses, allotment set to zero, seats you want to hold back kept back.
  • Run from wherever you want it run. You said your people in Tunisia already receive the hotel stop sales and forward them to each company by hand. One person can now close a date for all three markets at once, and nobody in Ljubljana, Zagreb or Belgrade has to repeat the work.
26 h

median reply in peak

last August

14 s

with the desk

drafted, then sent by a person

Djordje described the cost of the current speed exactly: a website booking sits in Outlook for an hour while the desk is busy, and by the time somebody opens it the seats are gone. The guest saw them free a moment earlier and takes it out on your team.
Module 3 of 4
Before the number, the reason for it

Six goals. All of them yours.

Better information to decide with

Pace against last year, on every departure, every morning, instead of a sheet filled in by hand every third day.

A higher booking rate

An answer in minutes instead of a day and a half, on the website, by email and to the agency network.

Prices that move on Monday

Not on the Friday before departure, and upward as often as downward.

Growth without hiring

1,5 people of capacity back, which is roughly 11 % more volume with the team you have.

A new market every year

Competitors, airports and programme gaps watched continuously, so opening Czech Republic or Romania is a decision, not a project.

Compliant, and able to prove it

Twelve personal accounts become one governed organisation, with training records and a policy.

Goals
The return, element by element

Each one pays for itself on its own. That is deliberate.

Element 1

AI Foundation

  • 20 to 40 % more productive per person, on their own job and their own tasks, not on a generic tool.
  • Less overtime. Complaints drafted in ten minutes instead of two hours in the evening, in Marija's own tone and under the right law.
  • Growth without hiring at the same rate. More revenue stops meaning proportionally more people.
  • Revenue and profit per employee up 20 to 40 %. The same team carries a bigger year.
4 people

of extra capacity at twenty people and a 20 % gain, with no extra salaries. One payment, no monthly fee.

Element 2

AI Booking Assistant

  • About 1 mio EUR books online today, which is your 15 %. Going to 30 % is another 1 mio EUR a year, roughly 80.000 EUR a month.
  • Nine bookings in ten complete themselves. The office is left with the ones that actually need a person.
  • The answer arrives before the competitor's. Same guest, four agencies, and yours replies at 21:47 on a Tuesday.
  • You already have the proof: the better website moved online bookings from nothing to 15 %.
+1 mio EUR

a year at 30 % online, against a one off 15.000 EUR. Your 70 % target sits well above this.

Element 3

Custom CRM

  • The right data, and the decision while it still matters. Price up on a fast departure, price down or advertise on a slow one, on the day and not two weeks later.
  • Campaigns and prices adjusted on evidence, per departure and per market.
  • New markets opened on signals, not on a Sunday spent reading five websites.
  • Higher sales conversion. Automatic reminders for your team and for the client, so no offer sits and no payment is chased late.
  • One clear pipeline. Who is waiting for an offer, who has to pay, who is going to call today.
25.000 EUR

is what you put the right information at in year one. It costs 24.000 once, and every year after that is clean.

The return
What it should return, on your numbers

Move the sliders. The defaults are deliberately low.

Estimated from your booking volume and conversion. This is the one number nobody could give us on the day.
170
In percentage points, on a base of 40 to 50 %. Every published benchmark is many times this.
2,0 pt
Ljubljana said 1.000 to 1.900 EUR. Belgrade sells lower. Blended.
1.200
About 17.000 seats from Belgrade, plus Ljubljana and Zagreb.
148
Raising 30 EUR on the last seats of a fast one, or catching a frozen one two weeks earlier. Your own numbers were 30 EUR up and 90 EUR down.
450
The aircraft is already paid for, so an extra seat carries hotel and transfer cost only.
40 %
Extra bookings a year
0

Extra revenue a year
0

Extra profit a year, both levers
0

Against a current group profit of about 500.000 EUR
0

Investment recovered after
0

Your inputs multiplied out, not a promise. The hours saved in section 5 are not counted here at all.

The return
The investment

Three elements. Take one, or take the system.

Elements selected0
Audit fee already paid− 1.500 EUR
All three together, −10 %0
Paid in
Total on signature, excluding VAT0
1.180 EUR / monthBooking assistant 490, CRM 690. Foundation has no monthly fee.

Not included, and not ours. Claude licences, about 20 EUR per person per month, bought directly by you in your own name. Scraping costs for the market module only if they pass 100 EUR a month, which on ten markets we do not expect.
Outside this scope. Anything agreed later and not listed above is billed at 119 EUR an hour, quoted and approved by you before we start it.
Prices exclude VAT. Offer holds until 30 September 2026.

Investment
How it runs

Something measurable is live in four weeks.

4 weeks

AI Foundation. Organisation, accounts, one to one sessions in all three offices, projects, policy, certificates. Then 8 weeks of support: email plus a monthly call.

8 weeks

Booking assistant. Catalogue loaded, three languages, tested against real past inquiries before it touches a live site.

12 weeks

Custom CRM. Four modules, three markets. Counted from the day we have everything on the next page, not from signature.

They run in parallel, not one after the other. The foundation lands first on purpose, because it is what makes the other two usable on the day they go live.
Timeline
What we need from you

Six things, none of them difficult.

Inquiries per week, per office

In season and out. A rough number is enough.

Darja or Ana, Ivana, Marija

Website booking statistics

How many reservations actually come through the site. It turns our estimate into your fact.

Djordje offered to ask the developer

Three files

The departure and booking window table, the shared seat allocation sheet, the special offer control sheet. Screenshots are enough.

Jadranka and Nenad

A push on BlueVendo

Max and Karol answered you on 22 August but not us. Their answer decides how deep two modules can go.

You, directly

One contact per office, and an owner for the CRM

Romina in Zagreb, plus a name in Ljubljana and Belgrade, so Sara can coordinate directly. And with Kika on leave, whoever owns the CRM day to day once it is live. We run it for the first three months either way.

You, this week

The list of people

Names and roles for the foundation, so we can plan the one to one sessions across the three offices.

You and Nenad

Next steps
Later, when the dependencies clear

Four things we deliberately left out of this offer.

Each one is worth doing and each one waits on something that is not ours to decide. No prices here, because pricing them today would be guessing with your money.

Waits on Synesis

Invoicing preparation

Filter the day's returns, produce one file with every price line and the advance invoices matched. Jadranka loses a day a week to this. We need to know what Synesis can export first.

Waits on a legal decision

Passport collection

Upload by link with a retention rule and a checklist per charter. Nenad said the current way is not compliant, so this needs your legal position before it is built.

Waits on the market

Voice on the phone

Djordje was right that Serbian callers still want a person. In two seasons that will have changed. The assistant already has the voice input; the phone line is a separate decision.

Waits on Axon Leap

Writing back and dynamic pricing

Placing a reservation automatically, and real price ladders per departure. Both belong to the new system in 2028. We build so that day is a connector change.

Later phases
1 / 30