i.
You tried it and “something works”
ChatGPT wrote you a good description. The next one was mediocre, the third invented a feature that does not exist. Without rules and checks, AI is a talented colleague who remembers nothing from yesterday.
AI consulting · AI consultant · Athens & Thessaloniki, Greece
Everyone talks about AI. Where does it pay off in your business?
AI consulting means finding where artificial intelligence is worth introducing in your work, where it is not, and in what order — before you buy tools or pay someone to set them up. An AI consultant does not sell you software; they tell you what to ask of it.
I am John Kiskipelis, an AI and e-commerce consultant based in Athens and Thessaloniki, Greece. I do not talk about artificial intelligence from a distance: I built Epoptes, the first Greek platform that measures whether AI models recommend a business, I run daily automations with Claude inside my own companies, and I teach “AI for E-commerce” at ELHub. I work with any business that has repetitive work involving text, data or customers — and UpCommerce, the agency I founded, takes on implementation when needed, as a separate engagement.
Creator of Epoptes (GEO monitoring)
Speaker, Balkan eCommerce Summit 2026
Instructor, “AI for E-commerce”, ELHub
15+ years in Greek e-commerce
When you need an AI consultant
The tool exists. The use of it is missing.
The people who find me do not ask “what is AI”. They have tried it. Their problem is that it does not produce the same result twice, or that they do not know where to start with it in the business.
ii.
Someone proposed an “AI solution”
Software with a monthly subscription, a chatbot, an agency that “does everything with AI”. You have nobody on your side to tell you whether you need it, what it will really cost and what will happen to your data.
iii.
Your team does work that repeats
The same emails, the same reports, stitching data from three systems every Monday. You suspect hours a month could disappear — but you do not know where to start without breaking something.
What I help with
Six areas — not a methodology with a name.
Every business gets stuck somewhere different. We start there, and the scope adapts — it can be one hour on one question or months alongside the team.
Where AI pays off — and where it does not
We map where time goes: content, customer service, data, repetitive tasks. We pick the two or three points with the biggest benefit and the smallest risk. The rest we deliberately leave alone.
Rules instead of prompts
A prompt that “asks” the model is no guarantee. Everything you know about your work — lengths, forbidden words, what is never invented, when it stops and asks — is written as a rule and checked with code, not with hope.
Tool and vendor selection
Claude, ChatGPT or Gemini? An off-the-shelf automation tool or custom? Is the software you were offered worth it? I answer without selling any of it — and tell you what to ask the vendor.
Data, risk, GDPR
What goes into a model and what must not. Where the keys are kept. At which steps a human approves before anything reaches the customer. Mistakes here do not show in the demo — they show six months later.
Training your team
From half a day so everyone uses a model properly to a programme per department. The goal: that you do not need me for every new use. I teach the same subject to e-shop managers at ELHub.
How AI sees you (GEO)
Your customers now ask ChatGPT “who should I choose”. We measure with Epoptes whether you appear, who appears instead of you, and what changes on the site and off it so that it recommends you. The GEO service →
Advice or implementation?
The “what” and the “how” are two different jobs.
A
The consultant
Says what is worth doing, why, in what order and with which rules. Works in hours or monthly, and the advice does not depend on whether a project follows. That is what I do, on this page.
B
The implementation
Builds the automations, writes the rules into code, connects the systems and hands them over with documentation. The UpCommerce team does this as a separate AI implementation service. When the change concerns the whole business rather than one process, there is the AI transformation programme.
What I have built
I advise on things I run myself.
Three systems working today inside my own companies — with the mistakes they made along the way, because that is where one learns.
Product
Epoptes
3 models
ChatGPT, Claude and Gemini, plus Google AI Overviews — every month, for 90+ businesses in 7 languages
A GEO monitoring platform that writes the questions customers ask, runs them on the models, and delivers a report with recommendations. The lesson: whatever can be checked with code is not left to the model — otherwise it recommends “create an FAQ page” to a site that has one.
Content
Production rules for e-shops
50 / 50
product descriptions under the minimum length in the first test without rules — and 0 invented features
The model was not lying; it said little, said it the same way, in everyone’s words. The fix was not a better prompt but a rules file and a script that checks lengths, repetition, forbidden words and numbers before anything goes out. I teach it in the “Claude for eShops” masterclass.
Operations
Morning brief in Slack
6 sources
appointments, client sales, articles from three blogs, Epoptes measurements, weather and name days — in one message at 9:00
A scheduled task that runs by itself every morning. Nothing technically impressive — and that is exactly the point: most hours lost in a business are this kind of data stitching, not “hard” problems.
Who I am
I learned AI by building with it, not by reading about it.
Fifteen-plus years in Greek e-commerce, founder of UpCommerce and co-founder of ELHub. In recent years my daily work has been building systems with AI models that do real work: measurements, content, reports, checks — and teaching e-shop owners how to do the same.
As a consultant I do not sell tools, I take no commission from vendors, and I do not recommend AI where a good spreadsheet is enough. If the right answer is “do not do it yet”, you will hear it.
EpoptesCreator, GEO monitoring SaaS
Balkan eCommerce Summit 2026Speaker: “Be the one AI recommends”
ELHubInstructor “AI for E-commerce”, masterclass “Claude for eShops”
MBA InstructorRené Descartes College (CNAM)
UpCommerceFounder, 1st Skroutz Partner agency
GrowthMentorMentor to founders & brands
Tell me how your business works — I tell you where AI pays off.
Thirty minutes, no commitment. You describe where your team’s time goes and you leave with two or three points worth looking at — whether we continue or not.
Book a call →30 minutes · online · no cost
Frequently Asked Questions
FAQ about AI consulting
Finds where artificial intelligence is worth introducing in a business, where it is not, and in what order. In practice: maps where the team’s time goes, picks the two or three points with the biggest benefit, defines rules and checks so the output is stable, recommends tools without selling them, and trains the people who will use them. They do not build the systems themselves — that is implementation work.
When you have tried ChatGPT or Claude, saw that something works, but cannot make it work the same way every time. When someone has proposed software or an “AI solution” and you do not know whether you need it. Or when your team spends hours on repetitive work and you suspect it should not. If, on the other hand, you know exactly what you want built, you do not need a consultant — you need implementation.
The first 30-minute conversation is free. After that it depends on the format: hours on one specific question, a two-to-three-week mapping with a written plan, or a steady monthly engagement while the business changes. You get a proposal with a specific scope after the first conversation. I do not sell software and I take no commission on tools I recommend.
Any business with repetitive work involving text, data or customer communication: e-commerce, services, B2B, tourism, healthcare, education, retail. My base is e-commerce, where AI has the most immediate results — product descriptions, customer service, sales analysis — but the same principles apply in every sector. Size: from the sole professional to teams of dozens.
The ones that work for the specific job, and they are almost always fewer than you think. For most cases: one of the three major models (Claude, ChatGPT, Gemini) with properly written rules, an automation tool to connect it to your systems, and checks written in code for whatever must not be left to the model’s judgement. Custom development only when off-the-shelf is not enough — and then I tell you.
In the businesses I see, no. It removes the part of the work nobody wanted to do — the fifty similar descriptions, the same twenty emails, stitching data from three systems — and leaves time for the part that needs judgement. The human stays in the loop where a mistake costs: before something reaches the customer, before a price changes, before anything goes out publicly.
With 30 minutes in which you describe how the business works today and where time is lost. From there I tell you whether I see a benefit and where, or whether I do not. If we go ahead, the first deliverable is always written: what changes, in what order, what is needed from you, what we will measure.
Related: Digital marketing consultant → · For e-shops: Ecommerce consulting → · Implementation by the team: AI implementation →
