AI & Custom Solutions

You're not missing AI. You just haven't seen what it would do for you.

AI for business, starting from your own work.

Drag through one working day, 08:00 to 16:00. Every marked block is time that leaks without being written down anywhere.

Whose day shall we X-ray?

One working day, 08:00–16:00

Arrow keys: 5 minutes at a time. Shift + arrows: 1 minute. Home and End: start and end of the day.

Nobody

A customer who always pays nine days late after a holiday; nobody has seen the pattern.
Lost quotes: nobody records why.
Callers who hang up at 15:01 and never call back.
Pallets that wait two days for one missing document.

What nobody does never shows up on anyone's timesheet.

Maria · accounts, building-materials wholesaler, Volos

3 h 15 min today on work nobody sees

48 min handed back today

Over a month of 21 working days: ≈ 17 hours

The rest stays with the person: checking, and whatever isn't worth automating yet.

Where the time goes: She calls the five customers with the largest open balances.

The formulaHanded back = Σ (minutes the system finishes without a person, block by block) − minutes of checking.

We count conservatively, as in the first month of a pilot. These are typical timings; we time yours with a stopwatch, in the X-Ray.

Measure your own day Book an X-Ray

Or keep scrolling: why nobody measures this time.

Maria · accounts, building-materials wholesaler, Volos · The same day as a table: today on the left, with AI on the right.
As it is today	With AI
08:10–08:55 · 45 min · Reads Types 23 supplier invoices into the ERP.	The invoices are read and posted as drafts. Maria checks the totals and approves the 3 exceptions. (10 min handed back today)
09:30–09:50 · 20 min · Reads Hunts for the delivery note behind an invoice that doesn't match.	Invoice and delivery note are matched automatically. The difference arrives flagged, with both documents side by side. (6 min handed back today)
10:40–11:00 · 20 min · Creates Writes reminders to three customers with overdue balances.	The emails are drafted with the right amounts. Maria reads them and sends them. (5 min handed back today)
11:30–12:05 · 35 min · Decides Classifies expenses for myDATA, one at a time.	Each classification arrives suggested, with a confidence score. Maria checks the uncertain ones and a sample of the rest. (8 min handed back today)
13:20–13:45 · 25 min · Reads Reconciles courier cash-on-delivery payouts against orders.	The reconciliation runs every morning. Two open items are left for Maria. (6 min handed back today)
14:30–14:50 · 20 min · Hears & speaks Answers sales: ‘Has the customer paid?’	Sales ask the assistant first, and it answers from the ERP. Only the awkward cases reach Maria. (4 min handed back today)
15:20–15:50 · 30 min · Creates Builds the weekly balances report in Excel.	The report comes out of the ERP as a draft. Maria checks the figures before she sends it. (9 min handed back today)
Nobody A customer who always pays nine days late after a holiday; nobody has seen the pattern.	—

Giorgos · sales, plastic-packaging plant, Almyros industrial area · The same day as a table: today on the left, with AI on the right.
As it is today	With AI
08:05–08:35 · 30 min · Reads Retypes orders that came in by Viber and email into the ERP.	Orders land as drafts. Giorgos checks them and posts them. (8 min handed back today)
09:00–09:55 · 55 min · Creates Builds a quote from scratch: drawing, material, cycle time, price.	The RFQ is read, 4 similar past quotes are found and a draft is built with ERP prices. Giorgos checks material, cycle time and price before it goes out. (12 min handed back today)
10:30–10:50 · 20 min · Hears & speaks Phones production: ‘When does the customer's order ship?’	The assistant answers order-status questions from the ERP, for the customer too, on Viber. (4 min handed back today)
11:40–12:10 · 30 min · Reads Searches his email for the latest raw-material price.	The assistant finds the price and shows the email it came from. (5 min handed back today)
14:00–14:25 · 25 min · Creates Writes follow-ups on 6 open quotes.	Six drafts, one per quote, with whatever was said last time. Giorgos edits them and sends them. (4 min handed back today)
15:15–15:45 · 30 min · Creates Pulls together the week's sales for the owner.	The week's figures pull themselves together from the ERP. Giorgos still writes the commentary for the owner. (9 min handed back today)
Nobody Lost quotes: nobody records why.	—

Eleni · front desk, dental practice, Heraklion · The same day as a table: today on the left, with AI on the right.
As it is today	With AI
08:00–08:40 · 40 min · Hears & speaks Plays back 14 overnight voicemails and calls everyone back.	The overnight calls were answered by the digital assistant, which said it was AI in its first sentence. 9 bookings are waiting to be confirmed. (10 min handed back today)
09:15–09:45 · 30 min · Hears & speaks Phones to confirm tomorrow's appointments.	Reminders go out on Viber. Eleni sees only those who asked to change. (8 min handed back today)
10:20–10:40 · 20 min · Hears & speaks Answers ‘what time do you close?’ and ‘is there parking?’ yet again.	The assistant answers from the practice's own information. (5 min handed back today)
11:30–12:00 · 30 min · Reads Types details from insurance-fund forms.	The forms are read and the fields filled. Eleni checks before anything is saved. (6 min handed back today)
13:00–13:25 · 25 min · Hears & speaks Calls back everyone who hung up during the lunchtime rush.	At peak time, the assistant picks up whatever nobody else can: it books, or takes a message for a person. (5 min handed back today)
15:00–15:30 · 30 min · Optimises Rebuilds tomorrow's schedule after three cancellations.	The system suggests who on the waiting list to offer the gaps to. Eleni approves, and a Viber message goes out. (4 min handed back today)
Nobody Callers who hang up at 15:01 and never call back.	—

Stavros · logistics warehouse, port of Volos · The same day as a table: today on the left, with AI on the right.
As it is today	With AI
08:00–08:30 · 30 min · Reads Types CMRs and delivery notes into the WMS.	The documents are read on the dock, from a phone photo. Stavros confirms quantities and codes. (8 min handed back today)
09:10–09:30 · 20 min · Optimises Three lorries arrived at once. He hunts for a free dock.	Arrivals are predicted from the hauliers' GPS and every lorry already has a dock and a slot. (4 min handed back today)
10:15–10:45 · 30 min · Sees Photographs a damaged pallet, writes an email, and the dispute begins.	The gate camera keeps a photo of every pallet on arrival. The damage report comes out as a draft, and Stavros checks it before it goes. (6 min handed back today)
11:45–12:05 · 20 min · Hears & speaks Answers customers: ‘Has my pallet left?’	The assistant answers on Viber, from the WMS. Anything it can't find there, it passes to Stavros. (5 min handed back today)
13:30–14:00 · 30 min · Predicts Plans tomorrow's shifts by eye, from yesterday's numbers.	A volume forecast suggests how many people tomorrow needs, shift by shift. Stavros decides. (4 min handed back today)
15:00–15:25 · 25 min · Decides Writes up stock-count discrepancies in Excel.	Discrepancies come out of the data, with the most likely cause beside each one. Stavros checks the ones that involve real money. (4 min handed back today)
Nobody Pallets that wait two days for one missing document.	—

The evidence

The need is there. The picture isn't.

The barrier is imagination and know-how. The figures say so plainly.

8.93%

of Greek firms with 10+ staff use AI. EU average: 19.95%.

In 2024 it was 9.81%, against 13.48% in the EU. Europe climbed; Greece slipped.

Eurostat, isoc_eb_ai, 2025, firms with 10+ staff · official statistics Source Eurostat, isoc_eb_ai, 2025, firms with 10+ staff · official statistics (opens in a new tab)
≈79%

of Greek firms with 10+ staff have never even considered using AI. EU: ≈68%.

Derivation: non-users 89.81% minus those who considered it 11.05%.

Eurostat, isoc_eb_ai, 2025, firms with 10+ staff · our analysis Source Eurostat, isoc_eb_ai, 2025, firms with 10+ staff · our analysis (opens in a new tab)
61.8%

of those who considered it and didn't go ahead said they lacked the expertise. 54.3% stalled on cost (EU: 38.2%). Only 22.0% said ‘we don't need it’.

Eurostat, isoc_eb_ai, 2025, firms with 10+ staff · our analysis Source Eurostat, isoc_eb_ai, 2025, firms with 10+ staff · our analysis (opens in a new tab)
92%

of SMEs name not knowing AI's benefits and uses as a main barrier.

AI Readiness Index 2025 (ENTERSOFTONE, AUEB/ELTRUN) · survey co-authored by a vendor, sample size not published Source AI Readiness Index 2025 (ENTERSOFTONE, AUEB/ELTRUN) · survey co-authored by a vendor, sample size not published (opens in a new tab)

When an owner says ‘we don't need it’, they usually mean ‘I've never seen what it would do here’. So this page doesn't explain AI. It shows it to you, on work like yours.

Why it stays invisible

Five reasons you can't see it.

None of them is negligence. All of them are human.

Frame 1 of 5

Sliced thin

Five minutes here, ten there, across 30 people. No line on the balance sheet says ‘retyping’.

Quality pioneer Armand Feigenbaum called it the ‘hidden factory’: the plant inside the plant that produces nothing but rework and waiting.

5 min · 10 min · 7 min · 4 min
Frame 2 of 5

Normalised

‘We've always done it this way.’ The workarounds became job descriptions.

(no figure: this is our own observation from the field)

workaround, since 2014
Frame 3 of 5

The wrong picture

You picture a chat window. AI reads, sees, hears and predicts.

Only 1.88% of Greek firms with 10+ staff use AI in production. EU: 4.14% (Eurostat 2025 · official statistics). Source Eurostat isoc_eb_ai (opens in a new tab)

reads · sees · hears · predicts
Frame 4 of 5

Knowledge lives in people

Your best technician is also your biggest bottleneck.

In a study of 5,179 customer-support agents, the least experienced resolved 34% more issues an hour with an AI assistant: the best people's know-how reached everyone (NBER w31161 · academic study). Source NBER w31161 (opens in a new tab)

only Nikos knows
Frame 5 of 5

DIY fails quietly

The in-house pilot that ‘didn't work’ convinced you it's hype.

AI tools built with an outside partner reached day-to-day use about 67% of the time. Internal builds managed about 33%, half as often (MIT NANDA, 2025 · research, self-reported, mostly large US companies).

pilot: on hold

The six senses

You hear ‘AI’ and picture a chat window.

In practice it does six things. Three lean on logic, three on perception and creation.

Logic · workflows, systems, data

Predicts

Tells you what's coming, from your own history.

ExampleHow much yoghurt the supermarkets will order on Friday, if it hits 34 °C and a promotion is running.

What disappears: The ‘just in case’ batch that ends up as waste.

Try it See the service
Optimises

Finds the best order, route or setting within your constraints.

ExampleIn what order 14 concrete mixers should leave, so that none waits on site.

What disappears: The morning phone call: ‘where's the mixer?’

Try it See the service
Decides

Proposes a decision with a confidence score. Anything uncertain goes to a person.

ExampleWhich myDATA category each expense belongs to. The sure ones go through, the uncertain ones wait for the accountant.

What disappears: Classifying thousands of documents, one at a time.

Try it See the service

Perception & creation · image, speech, text

Sees

Looks at every part, every pallet, every label, often through cameras you already have.

ExampleOne cracked cap among thousands of bottles an hour.

What disappears: The one-in-fifty spot check.

Try it See the service
Reads & hears

Understands documents, emails, messages and speech, in Greek.

Example1,500 delivery notes a month, and the 15:01 call nobody picked up.

What disappears: Retyping. The voicemail nobody plays until the next morning.

Try it See the service
Creates

Drafts from your own data: quotes, replies, reports, descriptions.

ExampleA draft quote from the RFQ and your ERP prices, ready to check before noon.

What disappears: The blank document at 17:30.

Try it See the service

Yes, we know: the real brain doesn't split this neatly. The drawing is a map, not neuroscience.

Use-case atlas

Twelve trades. None of them looks like a chat window.

Lawyers, pharmacies, olive mills, solar parks. The same six senses, entirely different work each time. Find yours, or the nearest thing to it.

Use cases built on how these trades actually work. Wherever there's a number, its source sits beside it; where no reliable source exists, we leave it blank.

Law firm, 5 lawyers · Heraklion We've written this notice ten times. Nobody remembers where the good one is. ReadsCreates Before and afterBack to the scene

With AI
Searches your own documents (notices, contracts, briefs) within the access rights already in place, and drafts with a reference to the source. Pulls deadlines and hearing dates out of documents and into the calendar. If documents must never leave the office, it runs on an open Greek model on your own server.

What disappears
Digging through the shared drive. The deadline that lives only on a sticky note.

Where the person stays
A lawyer writes the final version and signs it. The system gives clients no legal opinion.

What we measure
research hours per matter · first reply time to a new client · missed deadlines (target: zero)

Reference point
Legal and tax professionals expect AI to give back ≈5 hours a week each (survey of 2,275 professionals). Source survey · forecast (opens in a new tab) survey · forecast

Typical range
60–100 hours a month for a firm of 3–8 lawyers

See the service X-ray your own
Accounting office, 200 clients · Larissa Invoices arrive as data now. We still classify them one by one. ReadsDecides Before and afterBack to the scene

With AI
Pulls clients' expenses from myDATA, checks VAT numbers, VAT rates and totals, and suggests a classification with a confidence score. On Viber, it chases clients for missing documents and files them.

What disappears
Classifying 6,000 documents a month by hand. Asking for the same receipt a fifth time.

Where the person stays
Anything low-confidence goes to the accountant. Nothing reaches AADE, the tax authority, without their approval.

What we measure
minutes per document · share of suggestions approved unchanged · clients per accountant

Reference point
The myDATA REST API (v2.0.0) accepts expense classifications from the accountant, for the clients who have authorised them. Source official documentation (opens in a new tab) official documentation

Typical range
70–100 hours a month for 100–300 clients

See the service Try it at the station X-ray your own
Olive mill and exporter · southern Crete Every shipment needs the same papers in three languages. And the buyer in Hamburg wanted an answer yesterday. ReadsCreatesHears & speaks Before and afterBack to the scene

With AI
Reads certificates of analysis and buyer specifications, checks every value against the limits and prepares the shipping documents in Greek, English and German. An email and WhatsApp assistant answers buyers from the specification database, around the clock.

What disappears
Copying figures from a PDF into Word, three times over. ‘We'll get back to you on Monday.’

Where the person stays
The export manager approves every document. Prices and availability come only from a person.

What we measure
hours per shipment · shipments returned for document corrections · reply time to buyers

Reference point
There's no reliable public figure. We'll measure it in your X-Ray.

Typical range
No range: we don't print a number we can't stand behind.

See the service Try it at the station X-ray your own
Tutoring centre, 300 students · Volos Every September we tell the same two hundred parents the same things. Then we write three hundred monthly progress notes. Hears & speaksCreates Before and afterBack to the scene

With AI
A Viber assistant answers parents about timetables, absences and payments, from the centre's own records. For teachers, it produces exercise variants and draft progress notes from their own remarks.

What disappears
The 21:00 phone calls. The Sunday spent on three hundred progress notes.

Where the person stays
Only the teacher grades and assesses. Automated grading is a red line for us.

What we measure
reply time to parents · teacher hours spent on admin · enrolments in August–September

Reference point
Greek households spent ≈€614m on general-education tutoring in 2023. Source study, via press · legal analysis (opens in a new tab) And: AI that assesses learning outcomes is high-risk under the AI Act (Annex III, obligations from 2 Dec 2027). Source study, via press · legal analysis (opens in a new tab) study, via press · legal analysis

Typical range
30–50 hours a month for 150–400 students

See the service Try it at the station X-ray your own
Pharmacy · Heraklion ‘Have you got it? Can you hold it for me?’ Ten times a morning, with the queue out of the door. Hears & speaksReads Before and afterBack to the scene

With AI
On Viber, the assistant says whether a product is in stock, straight from the stock system, and holds it for collection. In the back office, it reconciles supplier invoices against deliveries and returns.

What disappears
The phone ringing over the counter. Reconciling returns at the end of the month.

Where the person stays
Only the pharmacist gives advice. The assistant handles availability and holds only, never health questions, and keeps no health data.

What we measure
counter calls at peak time · holds placed via Viber · hours spent reconciling invoices

Reference point
Greece has ≈101 pharmacies per 100,000 people, against 32 in the EU (2023 data). Source industry body (opens in a new tab) In a market that crowded, whoever answers first wins. industry body

Typical range
15–30 hours a month

See the service Try it at the station X-ray your own
Car dealership and service workshop · Larissa ‘Is my car ready?’ Half the service desk's day is that one question. Hears & speaksSees Before and afterBack to the scene

With AI
A voice assistant gives repair status from the workshop system and books service slots, out of hours too. From photos the customer sends, it makes a first assessment of the car they want to trade in.

What disappears
The call that ends in voicemail. The valuation visit that wasn't needed.

Where the person stays
A person gives the final trade-in price, after seeing the car.

What we measure
missed service calls · bookings made out of hours · reply time to sales enquiries

Reference point
About 36% of calls to dealership service departments were missed opportunities (≈53 million calls, US, 2025). Source vendor data (opens in a new tab) vendor data

Typical range
40–70 hours a month

See the service Try it at the station X-ray your own
50-room hotel · Hersonissos Every May we train seasonal staff from zero. By August they're still asking where the pool towels are. Hears & speaksPredicts Before and afterBack to the scene

With AI
A staff assistant, by voice or text and in five languages, answers from the hotel's own procedures. Guests get their own assistant on WhatsApp, Viber and email. In the kitchen, an occupancy and nationality forecast says how much food the buffet needs.

What disappears
‘Ask Sofia’, ten times a shift. The buffet that comes back half full.

Where the person stays
Complaints and special requests go to reception, with the whole conversation attached.

What we measure
days until a seasonal hire works unaided · guest complaints · kg of food waste per guest-night

Reference point
Only 13.51% of Greek accommodation businesses with 10+ staff use AI, against 21.31% in the EU (2025). Source official statistics (opens in a new tab) official statistics

Typical range
45–70 hours a month in season, for 40–60 rooms

See the service Try it at the station X-ray your own
Taverna, 120 covers · Rethymno At 21:00 the phone rings in German, the courtyard is full and nobody can get to it. Hears & speaksReads Before and afterBack to the scene

With AI
The assistant books tables in Greek, English and German, straight into the reservation book, and answers allergen questions only from the kitchen's official list. Supplier e-invoices turn into a cost per dish.

What disappears
The booking lost at peak time. ‘I don't think it has nuts in it.’

Where the person stays
For a serious allergy, the assistant always hands over to someone from the kitchen.

What we measure
unanswered calls at peak · bookings made out of hours · ingredient cost per dish

Reference point
There's no reliable public figure. We'll measure it in your X-Ray.

Typical range
15–30 hours a month

See the service Try it at the station X-ray your own
Estate agency · Chania The buyer from Sweden messaged on Saturday night. By Monday he'd booked a viewing with another agency. Hears & speaksCreates Before and afterBack to the scene

With AI
Replies in under a minute, in the buyer's language, asks about budget, area and timing, and books a viewing into the agent's calendar. Writes every listing in four languages from a single fact sheet.

What disappears
The lead that goes cold over the weekend. The same listing, typed four times.

Where the person stays
The viewing and the negotiation stay with the agent.

What we measure
first reply time · viewings booked · leads per agent

Reference point
Leads called within 5 minutes were about 100 times more likely to be reached than at 30 minutes. Source an old study, still the most cited (opens in a new tab) an old study, still the most cited

Typical range
30–50 hours a month for 3–10 agents

See the service Try it at the station X-ray your own
Solar parks, 12 MW · Thessaly An inverter dropped out on a Thursday. We spotted it in the monthly settlement. SeesPredicts Before and afterBack to the scene

With AI
Compares each string's output with what today's sun says it should be, and tells dust, shading and faults apart. Forecasts tomorrow's output to shrink imbalance costs.

What disappears
The fault that eats weeks of output before anyone notices. Panel cleaning by the calendar instead of by the data.

Where the person stays
The technician decides whether to drive out to the park. The alert tells them why.

What we measure
performance ratio · hours to fault detection · €/MWh imbalance cost

Reference point
Only 16.50% of Greek energy companies with 10+ staff use AI, against 33.57% in the EU (2025). Source official statistics (opens in a new tab) There's no public Greek savings figure; we measure it in the pilot. official statistics

Typical range
No range: we don't print a number we can't stand behind.

See the service Try it at the station X-ray your own
Building materials and ready-mix concrete · Almyros The contractor sends a bill of quantities as a PDF and wants a price today. We need three days. ReadsOptimisesCreates Before and afterBack to the scene

With AI
Reads the bill of quantities, matches every line to an ERP item code and drafts a quote at list prices. For ready-mix, it sequences the loads so mixers don't queue on site.

What disappears
Three days per quote. The mixer waiting on site with its drum turning.

Where the person stays
Discounts and credit limits are the salesperson's call. Prices come only from the ERP, never from the model.

What we measure
hours per quote · quote win rate · waiting minutes per delivery run

Reference point
In construction, 6.96% of Greek firms with 10+ staff use AI, against 10.79% in the EU (2025). Source official statistics · press (opens in a new tab) The big players have started: the TITAN Group has published closed-loop AI optimisation at one of its US plants. Source official statistics · press (opens in a new tab) official statistics · press

Typical range
No range: we don't print a number we can't stand behind.

See the service Try it at the station X-ray your own
Refrigeration and air-conditioning service company · Heraklion July: forty call-outs a day, eight technicians. The owner builds the schedule in his car. Hears & speaksOptimisesSees Before and afterBack to the scene

With AI
The assistant logs the fault by phone or Viber, asks for a photo of the unit's rating plate and reads it. Technicians are scheduled by area, skill and the parts already in the van. The job report is written from the technician's voice note.

What disappears
The kilometres driven for a part that wasn't in the van. The report written up at night.

Where the person stays
Urgent calls, such as a cold room full of stock, are seen by a person first.

What we measure
first-visit fix rate · kilometres per repair · hours from call to dispatch

Reference point
There's no reliable public figure. We'll measure it in your X-Ray.

Typical range
No range: we don't print a number we can't stand behind.

See the service Try it at the station X-ray your own

source: notice 2019/14 deadline: 23 Nov Law firm, 5 lawyers · Heraklion · use case AI-generated image

confidence 0.98 · approved 0.61 · to the accountant Accounting office, 200 clients · Larissa · use case AI-generated image

acidity 0.28 ≤ 0.30 EL · EN · DE Olive mill and exporter · southern Crete · use case AI-generated image

progress notes: 300 drafts grades stay with the teacher Tutoring centre, 300 students · Volos · use case AI-generated image

in stock · held until 19:00 Viber · 08:12 Pharmacy · Heraklion · use case AI-generated image

status: ready 17:30 scratch, rear bumper Car dealership and service workshop · Larissa · use case AI-generated image

answer from procedure 4.2 forecast: 212 breakfasts tomorrow 50-room hotel · Hersonissos · use case AI-generated image

table for 4 · 21:30 · DE allergens: from the official list Taverna, 120 covers · Rethymno · use case AI-generated image

reply in 58 s viewing: Mon 11:00 Estate agency · Chania · use case AI-generated image

string 7: −22% vs expected cause: dust, not a fault Solar parks, 12 MW · Thessaly · use case AI-generated image

line 14 → item 20-1187 load 3: 10:40 Building materials and ready-mix concrete · Almyros · use case AI-generated image

model and serial: read route: 3 visits, one van Refrigeration and air-conditioning service company · Heraklion · use case AI-generated image

Is your line of work missing from the map?

Describe a day to us

Feed the Machine

Give it some work.

Four stations, four verbs. Pick one, try it and see what changes.

Nothing you do here is stored.

Reads

Invoices, delivery notes, certificates. It reads them the way you do, except it doesn't tire at the hundredth.

Drag a document here, or select one and press Enter.

Κυανή Διαγώνιος Πολυμερή Α.Ε.
Οδός Διαγωνίου 47, 383 33 Βόλος
ΑΦΜ 099474704 · ΔΟΥ Βόλου

Τιμολόγιο πώλησηςΣειρά · Αρ. Α · 004718Ημ/νία 07/10/2026

ΠελάτηςΣυσκευασίες 47 Μ.Ι.Κ.Ε. · ΑΦΜ 800474705

Περιγραφή	Ποσ.	Τιμή	Αξία
Κόκκοι πολυπροπυλενίου PP-H, σάκος 25 kg	40	31,20	1.248,00
Masterbatch λευκό, σάκος 25 kg	4	86,50	346,00
Πρόσθετο UV, σάκος 20 kg	2	112,00	224,00

Καθαρή αξία1.818,00 €ΦΠΑ 24%436,32 €Πληρωτέο2.254,32 €

Τρόπος πληρωμής: Επί πιστώσει, 60 ημέρεςΚΥΑΝΗ ΔΙΑΓΩΝΙΟΣ
ΠΟΛΥΜΕΡΗ Α.Ε.
ΒΟΛΟΣ

Σελίδα 1 από 1Α · 004718

Κυανή Διαγώνιος Πολυμερή Α.Ε.
Οδός Διαγωνίου 47, 383 33 Βόλος
ΑΦΜ 099474705 · ΔΟΥ Βόλου

Τιμολόγιο πώλησηςΣειρά · Αρ. Α · 004718Ημ/νία 07/10/2026

ΠελάτηςΣυσκευασίες 47 Μ.Ι.Κ.Ε. · ΑΦΜ 800474705

Περιγραφή	Ποσ.	Τιμή	Αξία
Κόκκοι πολυπροπυλενίου PP-H, σάκος 25 kg	40	31,20	1.248,00
Masterbatch λευκό, σάκος 25 kg	4	86,50	346,00
Πρόσθετο UV, σάκος 20 kg	2	112,00	224,00

Καθαρή αξία1.818,00 €ΦΠΑ 24%436,32 €Πληρωτέο2.254,32 €

Τρόπος πληρωμής: Επί πιστώσει, 60 ημέρεςΚΥΑΝΗ ΔΙΑΓΩΝΙΟΣ
ΠΟΛΥΜΕΡΗ Α.Ε.
ΒΟΛΟΣ

Σελίδα 1 από 1Α · 004718

Σημείο Μηδέν Δομικά Υλικά Ο.Ε.
ΑΦΜ 801047477

Δελτίο αποστολήςΑρ. ΔΑ · 01183Έναρξη 08/10/2026 08:40

Παραλήπτης: Ορθή Γωνία Κατασκευαστική Ε.Ε. · ΑΦΜ 999474710
Σκοπός: Πώληση
Φόρτωση: Αποθήκη έδρας, Αλμυρός
Παράδοση: Εργοτάξιο, Οδός Τομής 12, Βόλος
Όχημα: Ιδιόκτητο φορτηγό

Είδος	Ποσότητα
Τσιμέντο CEM II/B-M 42,5N, σάκος 25 kg	120 σάκοι
Χάλυβας οπλισμού B500C Ø12, ράβδος 12 m	1.850 kg
Δομικό πλέγμα T131, φύλλο 2,10 × 5,00 m	24 τεμ.

ΠαρέλαβαΠαρελήφθησαν 118 σάκοι. 2 σκισμένοι.

Σελίδα 1 από 1ΔΑ · 01183

Εργαστήριο Ελέγχου Τροφίμων Διχοτόμος Ε.Ε.
Πιστοποιητικό ανάλυσης

Αρ. 26/118702/10/2026 · 06/10/2026

Πελάτης: Ελαιώνες 47 Α.Ε.
Προϊόν: Εξαιρετικό παρθένο ελαιόλαδο
Παρτίδα: L26-1014
Προδιαγραφή: Εισαγωγέας, Αμβούργο

Παράμετρος	Αποτέλεσμα	Όριο ΕΕ	Αγοραστής
Οξύτητα (% ως ελαϊκό οξύ)	0,28	≤ 0,80	≤ 0,30
Υπεροξείδια (meq O₂/kg)	7,9	≤ 20,0	≤ 12,0
K232	1,82	≤ 2,50	≤ 2,20
K268	0,14	≤ 0,22	≤ 0,18
ΔK	0,002	≤ 0,010	—
Ολικές πολυφαινόλες (mg/kg)	412	—	≥ 300

Υπεύθυνος εργαστηρίουΔ. Κ.

Σελίδα 1 από 126/1187

ERP formSupplier invoice

099474704 VAT number valid · check digit 4
800474705 VAT number valid · check digit 5
Totals: they reconcile (net + VAT = total due)

By hand ≈ 4:30AI 0:06

myDATA-ready (JSON)

{
    "issuer": {
        "vatNumber": "099474704",
        "country": "GR",
        "branch": 0
    },
    "counterpart": {
        "vatNumber": "800474705",
        "country": "GR",
        "branch": 0
    },
    "invoiceHeader": {
        "series": "Α",
        "aa": "004718",
        "issueDate": "2026-10-07",
        "invoiceType": "1.1",
        "currency": "EUR"
    },
    "paymentMethods": [
        {
            "type": 5,
            "amount": 2254.32,
            "note": "on credit, 60 days (to check)"
        }
    ],
    "invoiceDetails": [
        {
            "lineNumber": 1,
            "netValue": 1248.0,
            "vatCategory": 1,
            "vatAmount": 299.52,
            "expensesClassification": [
                {
                    "classificationType": "E3_102_001",
                    "classificationCategory": "category2_2",
                    "amount": 1248.0
                }
            ]
        },
        {
            "lineNumber": 2,
            "netValue": 346.0,
            "vatCategory": 1,
            "vatAmount": 83.04,
            "expensesClassification": [
                {
                    "classificationType": "E3_102_001",
                    "classificationCategory": "category2_2",
                    "amount": 346.0
                }
            ]
        },
        {
            "lineNumber": 3,
            "netValue": 224.0,
            "vatCategory": 1,
            "vatAmount": 53.76,
            "expensesClassification": [
                {
                    "classificationType": "E3_102_001",
                    "classificationCategory": "category2_2",
                    "amount": 224.0
                }
            ]
        }
    ],
    "invoiceSummary": {
        "totalNetValue": 1818.0,
        "totalVatAmount": 436.32,
        "totalGrossValue": 2254.32
    },
    "_netex": {
        "confidence_min": 0.72,
        "fields_to_review": [
            "paymentMethods"
        ],
        "vat_checks": "both valid"
    }
}

The structure follows the myDATA schema; in delivery we use the version in force at the time.

Within EU limits for extra virgin olive oil
Tight: 0.02 from the buyer's limit. The export manager is notified.
The values flow into the shipping documents: EL · EN · DE

Parameter	Result	EU limit (EVOO)	Buyer spec	Verdict
Free acidity (% as oleic acid)	0,28	≤ 0,80	≤ 0,30	within, tight (0.02 margin)
Peroxide value (meq O₂/kg)	7,9	≤ 20,0	≤ 12,0	within
K232	1,82	≤ 2,50	≤ 2,20	within
K268	0,14	≤ 0,22	≤ 0,18	within
ΔK	0,002	≤ 0,010	—	within
Total polyphenols (mg/kg)	412	—	≥ 300	within

Supplier invoice

Document type: Τιμολόγιο πώλησης · confidence 0.99
Issuer: Κυανή Διαγώνιος Πολυμερή Α.Ε. · confidence 0.98
Issuer VAT number: 099474704 · confidence 0.99
Tax office: Βόλου · confidence 0.93
Address: Οδός Διαγωνίου 47, 383 33 Βόλος · confidence 0.95
Series and number: Α · 004718 · confidence 0.97
Issue date: 07/10/2026 · confidence 0.99
Buyer: Συσκευασίες 47 Μ.Ι.Κ.Ε. · confidence 0.98
Buyer VAT number: 800474705 · confidence 0.99
Line 1: Κόκκοι πολυπροπυλενίου PP-H, σάκος 25 kg · 40 σάκοι × 31,20 € = 1.248,00 € · confidence 0.96
Line 2: Masterbatch λευκό, σάκος 25 kg · 4 σάκοι × 86,50 € = 346,00 € · confidence 0.95
Line 3: Πρόσθετο UV, σάκος 20 kg · 2 σάκοι × 112,00 € = 224,00 € · confidence 0.94
Net amount: 1.818,00 € · confidence 0.99
VAT 24%: 436,32 € · confidence 0.99
Total due: 2.254,32 € · confidence 0.99
Payment method: Επί πιστώσει, 60 ημέρες · confidence 0.72 · to check
Expense classification (suggested): Αγορές πρώτων υλών και υλικών · confidence 0.91

Delivery note with a handwritten remark

Document type: Δελτίο αποστολής · confidence 0.99
Number: ΔΑ · 01183 · confidence 0.97
Dispatch start: 08/10/2026 08:40 · confidence 0.96
Sender: Σημείο Μηδέν Δομικά Υλικά Ο.Ε. · confidence 0.98
Sender VAT number: 801047477 · confidence 0.99
Recipient: Ορθή Γωνία Κατασκευαστική Ε.Ε. · confidence 0.98
Recipient VAT number: 999474710 · confidence 0.99
Purpose of movement: Πώληση · confidence 0.99
Loading point: Αποθήκη έδρας, Αλμυρός · confidence 0.95
Delivery point: Εργοτάξιο, Οδός Τομής 12, Βόλος · confidence 0.94
Vehicle: Ιδιόκτητο φορτηγό · confidence 0.90
Line 1: Τσιμέντο CEM II/B-M 42,5N, σάκος 25 kg · 120 σάκοι · confidence 0.97
Line 2: Χάλυβας οπλισμού B500C Ø12, ράβδος 12 m · 1.850 kg · confidence 0.95
Line 3: Δομικό πλέγμα T131, φύλλο 2,10 × 5,00 m · 24 τεμ. · confidence 0.94
Handwritten receipt note: Παρελήφθησαν 118 σάκοι. 2 σκισμένοι. · confidence 0.71 · to check

Olive-oil certificate of analysis

Laboratory: Εργαστήριο Ελέγχου Τροφίμων Διχοτόμος Ε.Ε. · confidence 0.97
Certificate: 26/1187 · confidence 0.98
Client: Ελαιώνες 47 Α.Ε. · confidence 0.98
Product: Εξαιρετικό παρθένο ελαιόλαδο · confidence 0.99
Lot: L26-1014 · confidence 0.97
Sampled · issued: 02/10/2026 · 06/10/2026 · confidence 0.98
Free acidity (% as oleic acid): 0,28 · confidence 0.98
Peroxide value (meq O₂/kg): 7,9 · confidence 0.97
K232: 1,82 · confidence 0.96
K268: 0,14 · confidence 0.96
ΔK: 0,002 · confidence 0.95
Total polyphenols (mg/kg): 412 · confidence 0.94

Want this on your own documents?

Book the 45-minute Scan call

Free, 45 minutes: ten questions, and we fill in the Hidden Cost Meter together.

Sees

A bottling line, 600 bottles in this clip. Some of them have a problem. Who's going to spot them?

Checks one in fifty, as well as anyone can in the seventh hour of a shift.

Passed: 600
Inspected: 12
Defects caught: 0
Defects that reach the customer: 9

The inspector saw 12 of 600 and caught 0 of 9.

Rejection threshold: 0.80

See vision quality control

Your line runs all day. The inspector's eyes tire. The camera's don't.

Want this on your own line?

Book the 45-minute Scan call

Free, 45 minutes: ten questions, and we fill in the Hidden Cost Meter together.

Hears & speaks

Listen to two recorded calls with the assistant. It speaks Greek and English, and says it's AI in its first sentence.

The first sentence is always the statement that you're talking to an AI.

Dental practice · appointment

In Greek · 1:21 · Greek and English subtitles

The digital assistant says in its first sentence that it is an artificial intelligence system; straight after, that the call is transcribed and how to reach a person. It books a cleaning for Tuesday 17 November at 17:30, reads the details back, says it will send a confirmation on Viber and sets a reminder for the day before. When the caller asks about a sore tooth, it gives no medical advice and arranges a call back from someone at the practice the same day.

The call as text

Assistant (AI) 0:00Hello, you're speaking with the Demo Dental Practice's digital assistant, an artificial intelligence system.Γεια σας, μιλάτε με τον ψηφιακό βοηθό του Οδοντιατρείου Demo, ένα σύστημα τεχνητής νοημοσύνης.
Assistant (AI) 0:07The call is transcribed only to book your appointment. For a person, say “person” or press 0. How can I help?Η κλήση απομαγνητοφωνείται μόνο για να κλείσουμε το ραντεβού σας. Για άνθρωπο, πείτε «άνθρωπο» ή πατήστε 0. Πώς μπορώ να βοηθήσω;
Caller 0:17Good afternoon. I'd like an appointment for a cleaning, next week if possible.Καλησπέρα. Θα ήθελα ένα ραντεβού για καθαρισμό, αν γίνεται την άλλη εβδομάδα.
Assistant (AI) 0:22Certainly. For a cleaning I have Tuesday the 17th at half past five in the afternoon, or Thursday the 19th at ten in the morning. Which suits you?Βεβαίως. Για καθαρισμό έχω Τρίτη 17 στις πέντε και μισή το απόγευμα ή Πέμπτη 19 στις δέκα το πρωί. Τι σας βολεύει;
Caller 0:31Tuesday, half past five.Την Τρίτη, στις πεντέμισι.
Assistant (AI) 0:34Tuesday 17 November, at 17:30. What name is it under?Τρίτη 17 Νοεμβρίου, στις 17:30. Σε ποιο όνομα;
Caller 0:38Eleni Papadopoulou.Ελένη Παπαδοπούλου.
Assistant (AI) 0:40Eleni Papadopoulou, is that right? I'll send a confirmation on Viber to the mobile you're calling from, ending in 03.Ελένη Παπαδοπούλου, σωστά; Θα σας στείλω επιβεβαίωση στο Viber, στο κινητό από το οποίο καλείτε, που τελειώνει σε 03.
Caller 0:48That's right.Σωστά.
Assistant (AI) 0:50Let me read that back: a cleaning, Tuesday 17 November at 17:30, for Eleni Papadopoulou. Would you like a reminder the day before?Να επαναλάβω: καθαρισμός, Τρίτη 17 Νοεμβρίου στις 17:30, στο όνομα Ελένη Παπαδοπούλου. Θέλετε υπενθύμιση την προηγούμενη μέρα;
Caller 1:00Yes, thank you.Ναι, ευχαριστώ.
Caller 1:02Oh, and one of my teeth hurts a little. Should I be worried?Α, και με πονάει λίγο ένα δόντι. Να ανησυχήσω;
Assistant (AI) 1:07I can't give medical advice. I can ask someone from the practice to call you back today. Would you like that?Δεν μπορώ να δώσω ιατρική συμβουλή. Μπορώ να σημειώσω να σας καλέσει σήμερα κάποιος από το ιατρείο. Θέλετε;
Caller 1:13Yes, that's better.Ναι, καλύτερα.
Assistant (AI) 1:15Noted. Someone from the practice will call you today. Have a good day.Σημειώθηκε. Θα σας καλέσει άνθρωπος από το ιατρείο μέσα στη μέρα. Καλή συνέχεια.

What the assistant understood1:21

“I'm AI” disclosure: yesconfirmed
Transcription notice: yesconfirmed
Intent: Appointmentconfirmed
Service: Dental cleaningconfirmed
Date: 2026-11-17confirmed
Time: 17:30confirmed
Name: Eleni Papadopoulouconfirmed
Phone: caller_id (…03)confirmed
Confirmation via: Viberconfirmed
Reminder: The day beforeconfirmed
Status: Bookedconfirmed
Needs a person: yesconfirmed
Why: Medical questionconfirmed
What happens next: Call back todayconfirmed
Health details stored: noconfirmed

Handed to a person

The assistant records that there was a medical question, not what was said. The least health data possible.

What the assistant captured (JSON)

{
    "intent": "booking",
    "service": "Dental cleaning",
    "date": "2026-11-17",
    "time": "17:30",
    "name": "Eleni Papadopoulou",
    "phone": "caller_id (…03)",
    "confirmation": "viber",
    "reminder": "day_before",
    "status": "booked",
    "handoff": {
        "required": true,
        "reason": "medical_question",
        "action": "callback_today"
    },
    "ai_disclosure_played": true,
    "transcription_notice_played": true,
    "health_details_stored": false,
    "language": "el"
}

This is what reaches your calendar, PMS or ERP. Nothing else.

Hotel · transfer

Greeting in Greek, then English · 1:30 · Greek and English subtitles

The assistant says in Greek, then in English, that it is an artificial intelligence system; the rest of the call is in English. It finds the guest's booking and books a private transfer from Heraklion airport for 3 June at 23:30, with a child seat, for 45 euros paid to the driver. It doesn't decide the late check-out request itself: it passes it to reception, who will reply by email before arrival.

The call as text

Assistant (AI) 0:00[In Greek] Hello, you're speaking with the Demo Hotel's digital assistant, an artificial intelligence system.Γεια σας, μιλάτε με τον ψηφιακό βοηθό του Ξενοδοχείου Demo, ένα σύστημα τεχνητής νοημοσύνης.
Assistant (AI) 0:07Hello, you're speaking with the Demo Hotel's digital assistant, an artificial intelligence system.
Assistant (AI) 0:14This call is transcribed to handle your request. To reach a person, say “reception” or press 0. How can I help?
Guest 0:23Hi. We arrive on the 3rd of June and our flight lands at ten past eleven at night. Can you arrange a transfer from Heraklion airport?
Assistant (AI) 0:31Of course. What name is the booking under?
Guest 0:35Martin. M, A, R, T, I, N.
Assistant (AI) 0:39Martin, M-A-R-T-I-N. I can see a double room with a sea view, 3 to 10 June. Is that you?
Guest 0:48That's us. Two adults, and we'll need a child seat.
Assistant (AI) 0:52A private transfer for two adults with one child seat is 45 euros, paid to the driver. Pickup at arrivals at 23:30, Thursday 3 June. Shall I book it?
Guest 1:05Yes, please.
Assistant (AI) 1:07Booked. I'll send the confirmation by email and WhatsApp, with the driver's number on the day.
Guest 1:13Could we also have a late check-out on the last day?
Assistant (AI) 1:16Late check-out depends on occupancy that day, so I've passed your request to reception. They'll confirm by email before you arrive.
Guest 1:25Perfect, thank you.
Assistant (AI) 1:27Thank you, Mr Martin. See you in June.

What the assistant understood1:30

“I'm AI” disclosure: yesconfirmed
Transcription notice: yesconfirmed
Language: enconfirmed
Intent: Transfer bookingconfirmed
Arrival: 2027-06-03confirmed
Flight lands: 23:10confirmed
Name: Martinconfirmed
Booking: BK-4417confirmed
Guests: 2 adults · 1 child seatconfirmed
Pickup: HER · 23:30confirmed
Price (from the system): 45.00 EUR · PMSconfirmed
Status: Bookedconfirmed
Confirmation via: email · WhatsAppconfirmed
Requests: late check-out · With receptionconfirmed
Needs a person: yesconfirmed
Why: Needs a decision based on occupancyconfirmed
What happens next: Reception replies by emailconfirmed

Handed to a person

What the assistant captured (JSON)

{
    "intent": "transfer_booking",
    "booking_ref": "BK-4417",
    "guest_name": "Martin",
    "arrival": "2027-06-03",
    "flight_eta": "23:10",
    "pickup": {
        "place": "HER arrivals",
        "time": "23:30"
    },
    "pax": {
        "adults": 2,
        "child_seats": 1
    },
    "price": {
        "amount": 45.0,
        "currency": "EUR",
        "source": "PMS",
        "paid_to": "driver"
    },
    "confirmation": [
        "email",
        "whatsapp"
    ],
    "status": "booked",
    "requests": [
        {
            "type": "late_checkout",
            "status": "pending_reception"
        }
    ],
    "handoff": {
        "required": true,
        "reason": "needs_occupancy_decision",
        "action": "reception_email"
    },
    "ai_disclosure_played": true,
    "transcription_notice_played": true,
    "language": "en"
}

This is what reaches your calendar, PMS or ERP. Nothing else.

Synthetic voices (AI) for the assistant and the callers alike.

Greek numbers, names and addresses are the hard part. That's why we run 100 test calls before every handover. See the AI Front Desk

Want one of these on your own phone line?

Book the 45-minute Scan call

Free, 45 minutes: ten questions, and we fill in the Hidden Cost Meter together.

Predicts

One yoghurt, one dairy, one Friday. Change the weather, the season and the promotion, and see what the shelves will ask for.

Product: strained yoghurt, 1 kg

Friday's temperature 22 °C

Season

Shelf promotion

Forecast demand: 1,000 units
Range: 920–1,080
‘Just in case’ production: 1,150 units
Waste by habit: 150 units (13%)
Waste with a forecast: 80 units (7%)
Risk of an empty shelf: low

The forecast learns from your own history: sales, weather, holidays, promotions, tourism. Without at least two years of history, we don't promise accuracy. We say so up front.

See Data & decisions

Want this on your own sales?

Book the 45-minute Scan call

Free, 45 minutes: ten questions, and we fill in the Hidden Cost Meter together.

The meter you can't see

Put a number on the invisible.

Pick a department and move up to four sliders. The formula sits under the result, so you can check it.

Hidden Cost MeterWork sheet

Department Accounts

Documents per month: 600
Minutes per document: 5 min
Documents with an error or exception (%): 8%
Minutes to fix each error: 20 min
Hourly labour cost: €18.20/hour

Eurostat lc_lci_lev, 2025 · official statistics

Hidden cost per year€14,414

Hidden hours per month: 66
What could be automated: €7,207conservative€10,090ambitious
People back on real work: 0.22–0.31 full-time equivalentsWe don't count jobs cut. We count time given back.
Payback against a pilot: 10–21 months Against a typical pilot of €6,000 to €18,000.

What we'd build

Document reading into the ERP and myDATA, with an approval queue
Automatic matching of invoice, delivery note and order
Cash-on-delivery and bank reconciliation

Funding. In the X-Ray we check which programme fits, if any does, against the current call text.

The formula

Hidden hours/month = (documents × minutes + documents × error % × fix minutes) ÷ 60

Cost/year = hidden hours/month × 12 × hourly cost (+ external cost, where there is one). Automatable = labour cost × department coefficient (0.50–0.70) + external cost × a second coefficient. People = hidden hours/month × coefficient ÷ 150 hours. Payback = pilot cost ÷ (automatable ÷ 12).

The coefficients are our own conservative estimate, shown here so you can challenge them. 150 hours is one person's productive time in a month (≈1,800 a year).

Book the full X-Ray

The form will already hold your numbers.

What we build

Six services. Each one is measured.

Method

X-ray first. Scalpel second.

Probe · Workflow X-Ray · 1–2 weeks

We watch the work on site, time it with a stopwatch and read the workflow from your data.
Architect · Opportunity map · 1 week

Every opportunity with value, feasibility, time to value and payback.
Test · Measured pilot · 4–8 weeks

The pilot runs on real work, with a person in control and a weekly KPI board.
Harden · Scale · ongoing

Monitoring, security, training, roll-out, and a quarterly report of the hours handed back.

Pilot targets are signed before any code is written.

The X-Ray is credited 100% against the pilot if you sign within 60 days.

Three sizes of X-Ray
Compact	Up to about 10 people, one workflow, one week. The top 3 opportunities.	€890
Workflow X-Ray	Up to 3 departments, two weeks, on site. The top 5 opportunities and a demo on your own data.	€1,900
Plant X-Ray	For manufacturing plants: 10 working days on site, including energy, downtime, quality and a camera test on one line.	from €4,900

Prices exclude VAT.

See the X-Ray How we work, step by step

Trust

It tells you it's AI. In its first sentence.

It says it's AI in its first sentence, as Article 50 of the AI Act requires.
A person is always one step away: say ‘person’ or press 0.
Your data doesn't train models.
Processing in the EU.
Logs and a data processing agreement (DPA) on every project.

Read the AI Trust Standard

Your ERP has AI now. Who redesigns the work around it?

ERP vendors are building AI agents into their systems. But the work also lives outside the ERP: in email, on Viber, in Excel, in the e-shop. That's where we connect.

80% of SMEs cite ERP and CRM integration as a barrier to AI. AI Readiness Index 2025 · survey co-authored by a vendor Source AI Readiness Index 2025 (opens in a new tab)
In June 2026, ENTERSOFTONE unveiled ERP versions with AI agents. euro2day · press Source euro2day (opens in a new tab)

Systems we speakSoftOne (Soft1 Web Services) · Entersoft · Epsilon (via a partner) · myDATA (REST API) · e-invoicing providers · WooCommerce · Gmail and Microsoft 365 · Viber · WhatsApp

See systems integration

Proof

If it isn't measured, it didn't happen.

Every project we publish has before and after numbers, a method note, what the AI didn't do, and the person who got their hours back.

Our first measured cases will go here. Until then, the twelve use cases are further up. Back to the map

Only 10% of AI's value is the algorithm. 20% is data and technology. 70% is the workflow, the people and the governance. That's where we work.

BCG · consultancy Source BCG (opens in a new tab)
Klarna's AI assistant cut resolution time from 11 minutes to under 2. In 2025, the company started hiring people again, for quality. That's why every assistant we build has a way out to a person.

Klarna · company · Entrepreneur · press Klarna Klarna (opens in a new tab) Entrepreneur Entrepreneur (opens in a new tab)
About 95% of GenAI pilots show no measurable financial return. That's why we sign the targets before we start.

MIT NANDA, 2025 · research, self-reported, mostly large US companies Source Fortune · MIT NANDA, 2025 (opens in a new tab)

See the work

Common questions

The questions we hear first.

What can AI actually do for a small or medium-sized business?

Six things: it reads documents, sees through cameras, hears and answers calls and messages, predicts demand and breakdowns, optimises schedules, and drafts quotes and reports. In practice that means less retyping, fewer unanswered calls and faster quotes. Which of the six matter to you is what the X-Ray shows, on your own data.

How much does it cost?

Diagnosis is fixed-price: Compact €890, Workflow X-Ray €1,900, Plant X-Ray from €4,900, credited 100% against the pilot. A pilot starts from €6,000, and the AI Front Desk from €390/month + €2,500 set-up. Prices exclude VAT. The models are usually the cheapest part; the real cost is connecting to your systems.

How long does it take to pay back?

It depends on volume, which is why we measure before we promise. Worked example: 1,500 documents a month, each 5 minutes faster, counting only the 70–85% that go through without a person. That frees 88–106 hours a month, €19,100–23,200 a year at €18.20 an hour. With a €15–25k build and €400 a month to run, payback takes 10–21 months.

Will it replace staff?

No, and we don't design it that way. AI takes the repetitive work: typing, searching, the same question for the hundredth time. The hours it frees go to work that needs a person: customers, quality, sales. Every system we build has a way out to a person, and every project we publish says what that person did with the time.

What about our data, GDPR and the AI Act?

We act as your processor, under a data processing agreement with a list of sub-processors. Your data doesn't train models, processing happens in the EU and we keep logs. Every assistant says it's AI in its first sentence, as Article 50 of the AI Act has required since 2 August 2026. It's all written down in our AI Trust Standard.

Does it work with our ERP (SoftOne, Entersoft, Epsilon)?

Yes. We use Soft1 Web Services for SoftOne, and the partner tools Entersoft and Epsilon provide. From 1 February 2027, invoices reach myDATA only through a certified provider, not straight from the ERP, and we connect to that as well. If your ERP has its own AI, we connect what sits outside it: email, Viber, the e-shop, Excel.

Is there funding?

There's no open programme for everyone right now. A €150m AI scheme for SMEs has been announced, but its terms aren't published yet. Beneficiaries of the ESPA advanced digital-transformation call are still implementing. Every X-Ray includes a funding annex: which programme fits, if any does, based on the current call text.

Where do we start?

With a free 45-minute Scan call: ten questions on volumes, systems and whatever is eating your hours, and we fill in the Hidden Cost Meter together. If it's worth it, the Workflow X-Ray follows: within one to two weeks you have a map of hours and euros, five opportunities with payback, and a demo on your own data.

Answers last reviewed: 13 October 2026

Show us one working day.

We'll show you which of its hours don't need to be human, and what your people could do with them.

Book an X-Ray Start with the free 45-minute Scan

We reply within one working day. Offices in Heraklion and the Almyros industrial area.

AI for business, starting from your own work.

The need is there. The picture isn't.

Five reasons you can't see it.

Sliced thin

Normalised

The wrong picture

Knowledge lives in people

DIY fails quietly

You hear ‘AI’ and picture a chat window.

Predicts

Optimises

Decides

Sees

Reads & hears

Creates

Twelve trades. None of them looks like a chat window.

Give it some work.

Reads

Supplier invoice

Delivery note with a handwritten remark

Olive-oil certificate of analysis

Sees

Hears & speaks

Dental practice · appointment

Hotel · transfer

Predicts

Put a number on the invisible.

What we'd build

The formula

Six services. Each one is measured.

X-ray first. Scalpel second.

Probe · Workflow X-Ray · 1–2 weeks

Architect · Opportunity map · 1 week

Test · Measured pilot · 4–8 weeks

Harden · Scale · ongoing

It tells you it's AI. In its first sentence.

If it isn't measured, it didn't happen.

The questions we hear first.

Show us one working day.