AI transformation · Strategy, implementation, training, measurement
AI transformation: not a chatbot in the corner — a change in how the business works
AI transformation is the programme through which a business moves from isolated experiments with ChatGPT to the daily use of artificial intelligence in processes that are measured. The UpCommerce group delivers it with four arms under one roof: strategy from a consultant who builds with AI himself, implementation from a team that sets up automations every day, training from ELHub, and monthly measurement from Epoptes. For businesses in Greece with a team, in any sector.
What it is and is not
The difference between “we have AI” and “we work with AI”
Most businesses in Greece already have someone using ChatGPT. Few have changed anything in how they work. The gap between the two is AI transformation — and it does not close by buying software.
Which steps of the work the model takes on, which the human does, and where the approval point is. Written down, agreed, measured.
What AI may do on its own, from which sources, with which checks before anything goes out. What never goes into a model.
No layoffs, no replacement of judgement. It removes the repetitive and leaves time for the part that needs a human — and for the customer.
Where it starts
Three areas where the benefit shows in the first month
We do not start from the most impressive. We start where the most hours are lost with the least risk.
Content and marketing
Product descriptions, category copy, newsletters, ad variations, replies to reviews. Here AI produces volume — and here it needs the strictest rules, because an invented feature goes straight to the customer. UpCommerce’s digital marketing team already works this way for its clients.
Customer service
The same twenty questions, every day, by email, chat and social. An assistant that answers from your own documents and policies — and hands over to a human when unsure — changes response time and frees the team for the cases that need judgement.
Data and operations
Reports stitched by hand from three systems, stock checks, invoice reconciliation, the morning “what happened yesterday”. The least visible hours of a business — and the easiest to automate safely.
And one that is often forgotten: how AI sees you
While the business learns to use the models, its customers are already asking them “who should I choose”. Whether you appear in those answers is part of the transformation — see GEO / SEO for LLMs.
The programme, in four phases decided one at a time
Sessions with management and with the people who do the work. Where time goes, what repeats, what it costs when it goes wrong. Deliverable: a written plan with three starting points, an order, and what will be measured.
One process, start to finish. The “before” is measured, it is built with rules and a human approval point, the “after” is measured. If it does not perform, you learn it in a month and not in a year.
What the pilot taught becomes a rule: what AI does alone, what gets approved, what never goes in. The team is trained per role, on its own processes — not on generic examples.
The next processes are built by the implementation team the same way. Every month a review: hours returned, mistakes caught, what opens next. And the business’s visibility in AI, measured by Epoptes.
The four arms
Why one group and not four vendors
Strategy — John Kiskipelis
Mapping and decisions with management are done by a consultant who has built an AI product himself and runs automations daily in his own companies. In scheduled sessions per phase — the day-to-day work of the programme belongs to the team.
Implementation — the UpCommerce team
Automations, rules written in code, connections to e-shop, CRM, email and Slack, checks before anything goes out. Delivered with documentation so you do not depend on us.
Training — ELHub
The E-Commerce Learning Hub trains e-shop managers in “AI for E-commerce” and runs the “Claude for eShops” masterclass. The same material is adapted to your own processes, so the team carries on by itself.
Measurement — Epoptes
The group’s GEO monitoring platform measures every month whether ChatGPT, Claude, Gemini and Google AI Overviews recommend the business — so the transformation is visible from the outside, not only from within.
Examples from the group itself
Whatever we propose, we have first put through our own companies. Three small, real ones:
In the first test on fifty products without rules, the model invented nothing — but wrote all the descriptions too short and with the same four openings. A rules file and a check script solved both. That is transformation in practice: boring and measurable.
Every morning at 9:00 a message in the team’s Slack gathers appointments, client sales, articles from three blogs and the Epoptes measurements. Nobody stitches it by hand any more — and nobody starts the day without knowing what happened yesterday.
Epoptes writes questions, runs them on three models, compares with competitors, produces PDFs and recommendations, bills and delivers — automatically, every month, for dozens of businesses. With rules enforced in code, not requests to the model.
The 37 pages of upcommerce.gr were built with a structure the models read and are measured every month: in October 2026 the brand appears in 75% of checks across the three models. See how it is done in website development.
AI Implementation
If you want one process built now, without a programme: automations, content with rules, AI customer service, reporting.
AI implementation →AI Consulting
Strategy only, from John Kiskipelis, as a separate engagement — without implementation.
AI consulting →GEO / SEO for LLMs
Get recommended by the models when the customer asks. Measurement and work on the site and off it.
GEO →Ecommerce Consulting
For e-shops: profitability per channel, marketplaces, cross-border — by John.
Ecommerce consulting →Digital Marketing Agency
Marketing execution by the team of six: Google, Meta, Skroutz, social, email.
The agency →Athens & Thessaloniki
Start with a conversation about where your time goes
In the 30-minute intro call we will ask how the work is done today — not what AI you want. From there we tell you whether we see a programme, a single implementation, or nothing yet.
Frequently asked questions about AI transformation
What is AI transformation?
Changing the way a business works so that artificial intelligence takes over the repetitive part of the work — with rules, checks and people who approve where a mistake costs. It differs from “adding a chatbot”: it concerns processes, roles and data, not a tool. It starts from one process and scales only once the first one has been measured.
Which businesses is the programme for?
Businesses with a team — from five people upwards — where there is enough repetitive work involving text, data or customers for the change to be worth it: e-commerce, service companies, B2B, tourism, healthcare, retail. For smaller businesses or for a single process, AI implementation as a standalone service is usually the right start.
How long does an AI transformation programme take?
Mapping two to three weeks, the first pilot thirty days, and scaling to the remaining processes three to six months depending on the size of the team. You do not sign up for the whole journey on day one: each phase delivers something measurable and the next one is decided on that basis.
How much does it cost?
Per phase, not as one package. Mapping has a fixed cost given after the intro call; the pilot and the scaling are priced according to the processes and systems involved. The cost of the AI tools themselves (subscriptions, API calls) is separate, yours and transparent — usually far smaller than most people expect.
Who does what within the programme?
Four arms of the same group. Strategy and mapping are done by John Kiskipelis, in scheduled sessions with management. Implementation — automations, rules in code, system connections — by the UpCommerce team. Training by ELHub, the group’s education arm. Measurement of AI visibility by Epoptes. You have one project lead for everything.
What about our data and GDPR?
Before a single line of automation is written, we define what may be sent to a model and what may not: customers’ personal data, financial figures, access keys. We use the providers’ business accounts, where data is not used for training, and every step that touches a customer has human approval until it is proven stable.
What will you need from us?
One person in the business who knows how the work is done today and has two to four hours a week for the pilot. Access to the systems involved. And the willingness to measure the “before” honestly — without it, the “after” cannot be proven.
