AI automation · Ukraine → worldwide

I automate the processes
that actually matter

Every business has work clients pay for. And work that just eats the day: requests scattered across three chats, hunting for documents, calling back for details you should have asked once. I find the second kind — and remove it.

See the cases → Free process review →
Artur Pysarskyi·I build every system myself·first review free, no pressure
Where I start

Not every process is worth automating

Automating a bad process gives you the same mess, only faster. So I start by breaking the work into steps and asking one question about each: would the client notice if this didn't happen?

They'd notice — that's green, it's what they pay for. They wouldn't, but the business would stall — yellow, the internal machinery. They wouldn't and nothing would happen — red.

Chaos: every move on its ownSystem: everything moves one way →
Green: what the client sees
The therapist gives the massage. Someone answered the question. The product arrived.

This is the part of the work people actually pay you for.

You don't remove it — you scale it. A client wrote at 11:40pm and got a proper answer? Value was created, even while you slept.
Yellow: the internal machinery
Issue the invoice, reconcile the payment, file the report, approve a document, hand over the shift.

The client never sees it and never pays for it. But skip it and the business stalls.

It isn't a mistake, it's machinery. Nobody ever designed it — it took shape on its own. You don't throw this out, you get it in order: five minutes instead of an hour.
Red: what shouldn't exist at all
This is where your day disappears:
requests in three chats and no single list · "one second, I'll find it" — ten minutes in folders · calling the client twice because nobody asked the first time · a spreadsheet kept alongside the CRM
This is what we remove. Not because automation is trendy, but because these steps never should have existed.
Green — straight to the client
Yellow — arrives, after extra stops
Red — spins in place
80%
In process management, up to 80% of steps in a typical workflow create no value for the client. Some are necessary — reporting, safety, compliance. The rest simply eats hours. Below are the eight places they hide most often.
01Moving data around+
02Overproduction+
03Excess inventory+
04Waiting+
05Errors and rework+
06Over-processing+
07Unnecessary motion+
08Unused potential+

Recognise any of them? Then we already have something to talk about.

What I build

One process — one system

Chatbot development, CRM systems and business process automation on Make.com and n8n with AI. These systems are already built and running — but none gets installed "as is": logic, copy and integrations change for your business.

💬
Customer support chatbot
A client writes at 11:40pm — your Telegram, Instagram or WhatsApp chatbot answers in twenty seconds, with your prices and terms. You wake up to a booking, not a missed message.
Case: real-estate bot →
🗓️
Appointment booking bot
Online booking without an admin: the bot books an open slot and reminds the day before. Fewer no-shows, fewer "when am I booked again?" calls.
Case: Instagram manager →
📄
Document processing (OCR)
Photo of a contract → text recognised, data extracted, file tagged in a database. Search takes seconds, not half a day.
Case: document archive →
👤
Hiring automation
A CV arrives — AI reads and scores it. You move the candidate with buttons in Telegram, never opening a spreadsheet.
Case: HR system →
⚙️
Something of your own
An unusual process that fits no list? That's the interesting part. We'll break it down on a call and I'll design for it.
Message me →
Cases

Systems that actually work

Not mockups and not slide decks. I built each of these end to end, and each one works. Two you can watch in action, one you can test yourself. These are a selection from dozens of systems built; bigger ones are in progress right now. If your case isn't here — write to me: chances are I've built something close.

01
HR / Recruiting WORKING SYSTEM
The whole hiring cycle, run with buttons in Telegram
+
Before

CVs in the inbox, data in a spreadsheet, emails written by hand. Every candidate meant 20–30 minutes of routine — and the more applications came in, the more people fell through the cracks.

After

An application arrives — AI reads the CV, scores it, checks for duplicates and files it into the CRM. The recruiter gets a card in Telegram and sends a test task or rejection with one tap. Every decision updates the database and emails the candidate automatically.

AirtableMakeGemini AITelegramFilloutSembly
02
Real estate WORKING SYSTEM
An AI consultant that speaks four languages
+
Before

Clients write at all hours in different languages. The questions repeat — residency, taxes, prices. Agents burn hours on conversations that lead nowhere while hot leads get lost in the noise.

After

The bot answers in four languages and holds the client's language throughout. Facts come only from the agency's knowledge base — zero invented numbers. It qualifies in conversation and, when the client is ready, hands the agent a structured lead and logs it to the CRM.

Claude AIMakeAirtableTelegram
03
Legal / Finance WORKING SYSTEM
Photo of a document → structured database in 15 seconds
+
Before

Contracts and invoices scattered across email, cloud and paper. Finding one takes 20–30 minutes. A quarterly reconciliation means days of manual digging.

After

Send a photo to Telegram — the text is recognised, AI identifies type, parties, dates and amounts, and the document lands in a structured database. The original goes to the archive; search takes seconds.

TelegramCloud VisionOpenAIMakeNotion
04
Beauty / Local services WORKING SYSTEM
Instagram manager: booking, reminders, client base
+
Before

DMs arrive around the clock, the admin replies during business hours. Bookings by hand, no reminders, client history living in someone's memory.

After

An AI manager answers questions, books time slots via Cal.com, handles reschedules and cancellations, sends reminders the day before, and builds a client database from the first message.

InstagramCal.comOpenAISendPulseMake
05
Auto service TRY IT YOURSELF
AI service advisor for a garage: by voice, 24/7
+
The problem

An auto service runs on phone calls. A client calls in the evening — nobody picks up. The mechanic is under a car and can't reply in chats. The owner never sees how many requests got lost in between.

What the system does

The client describes the problem by voice or text — AI transcribes it, classifies the category (suspension, engine, electrics, body, service) and matches the right mechanic by specialisation. The mechanic replies by voice too: free slots and a price — AI turns that into concrete date buttons for the client. Then booking, confirmation, status updates. Everything goes through the bot: client, mechanic and owner each see their part.

TelegramAirtableGemini AIMakevoice input

This is my own system — the largest I've built (around 90 modules). It runs, and you can test it yourself: message me and I'll give you access to the bot so you can walk the whole client journey in two minutes.

How I work

Three steps, no surprises

You always know what comes next and what it costs. No "well, this part needs extra budget".

01
Process review
Thirty minutes on a call. You show how the work happens now — I break it into steps and mark where value is and where waste is. Free. If there's nothing worth automating, I'll say so.
02
Prototype in days
Not a deck, not a proposal. A working piece of the system with your data that you can talk to yourself. Then you decide whether we continue.
03
Launch and handover
Everything is built in your accounts. Documentation, training, support for the first two weeks. The system belongs to you — I don't keep you on a hook.
What stays with you
The system, in your own accounts. Make or n8n, Airtable, the bot. The logins are yours. My access lasts as long as it's needed.
A knowledge base you edit yourself. Change a price in the table and the bot answers the new way. No need to ask me.
Documentation in plain language. What goes where, how to stop it, what to do if it breaks.
An hour of handover. We go through the system together until it makes sense.
Two weeks of support. Something always surfaces in the first days. That's my job, not yours.
The right to walk away. There's no subscription to me. If it's not for you, take the system and run it yourself.
About

I've worked with people for years.
Only the tools changed.

Artur Pysarskyi

Before automation I spent years selling clothing on eBay: hundreds of buyers across dozens of countries — from the US to Australia. Sourcing, messaging, returns, shipping parcels halfway around the world.

It sounds far from AI, but it gave me the thing that matters: I know what a business looks like from the inside when everything rests on you. I know how much time repeat questions eat, and how easily one unsent reply loses a client. I automate pain I've lived through myself.

Then came hundreds of hours of practice: automation and AI courses, n8n certifications, dozens of scenarios built. Every system on this page was built by me end to end.

100s
buyers worldwide over years of trading
3
countries lived and worked in: Ukraine, Cyprus, Sweden
prototype
a working prototype built fast for your specific request — so you can see it, not imagine it
EN / UA
remote-first, no time-zone barriers
n8n · certifiedMake.comClaude AIAirtableTelegram BotsOpenAIOCRCal.com
Questions

What people ask most

How much does it cost?
+
There are three parts, and I name all of them upfront.

1. Build — one-time, depends on the process. As a guide: a done-for-you bot from $400, a bot with a CRM system from $800, document processing from $1,500. Exact figure after a free review.

2. Tools — roughly $30–80/mo depending on volume (Make, Airtable, AI usage). You pay these directly from your own accounts: you see real costs and never pay me a markup on top.

3. Support — optional, from $150/mo: changes, updates, monitoring so the system doesn't go stale. First two weeks after launch are free.
How fast will I see a result?
+
The first working prototype is ready within a few days of our call. A full launch with testing usually takes one to two weeks.
Do I need a developer?
+
No. Everything runs on no-code platforms with AI. You get documentation and a handover session — no programming needed to operate it.
Will the AI make things up to my clients?
+
No. The bot answers only from your approved knowledge base. If the answer isn't there, it says so honestly and hands the conversation to a human. No invented prices, timelines or promises — that's a hard rule in every system I build.
Who owns the system after launch?
+
You do. Everything is built in your accounts, with documentation and a full handover. You can run it yourself or keep me on support — your choice, not a condition.
Do you build chatbots for Telegram and Instagram?
+
Yes. A chatbot for Telegram, Instagram Direct, WhatsApp or Viber — wherever your clients actually write. It answers questions, quotes your prices, books appointments, captures contacts and keeps a client database. I take projects from Ukraine and from English-speaking markets — the US, the UK, Scandinavia, Australia, New Zealand.
What can actually be automated in a small business?
+
Most often: replying to clients in messengers, online booking and reminders, keeping a client database (CRM), recognising and sorting documents, screening CVs when hiring, and owner reports. If an action repeats daily along the same script, it can almost always be automated.
What if I don't actually need this?
+
I'll tell you straight. If you get three requests a day and handle them comfortably, automation won't pay off — and an honest answer saves you money. I'd rather build a system that works than sell one that doesn't.
Tell me what hurts

Thirty minutes and you'll know whether there's anything worth automating. No decks, no pressure — we break down your process and the decision stays yours.

Telegram · fastest
@apcholo→ Instagram
@apcholo→ Email
workwithartur@gmail.com→ Phone · WhatsApp
+38 093 268 36 86→
Where I work from. I live in Lutsk, Ukraine. Everything is built online — in your accounts, in English or Ukrainian.
What I usually build: a chatbot in Telegram, Instagram or WhatsApp; online booking and reminders; a small-business CRM; leads and follow-ups; document processing; hiring automation.