Skip to content

A B2B shop run by AI — a human only approves

A wholesaler of photovoltaics, energy storage and heat pumps. The system prepares the customer's enquiry, price, transport and reply in their language. A human sends it.

Numbers with a stamp

The shop runs in production, and we build a shop like this in 4–6 weeks. Every number below has a measurement date and a source — you can check it yourself, without asking us.

  • 14,139

    product cards in the shop

  • over 30,000

    offers synchronised automatically, LIVE 24/7

  • 13

    languages in which the shop speaks to the customer

  • 8

    product categories

  • over 50

    AI roles, each with its own scope of work

  • 4–6

    weeks to build a shop like this and take it to production

24 hours in 90 seconds

One real day, 15.10.2026, hour by hour. Each bar is one workstation: it lights up in the hour in which it closed a task — the brighter, the more tasks closed. Look at three in the morning.

Chart of the day 15.10.2026: seven workstations on an axis from 00:00 to 23:59 the human sleeps, the company works 00:00 03:00 06:00 09:00 12:00 15:00 18:00 21:00 23:59 Chart of the day 15.10.2026 in vertical layout: hours from top to bottom, seven workstations 1234567 00:00 03:00 06:00 09:00 12:00 15:00 18:00 21:00
  1. 87

    Sales and enquiry handling

    replies to enquiries, changes to orders

  2. 264

    Offers and transport

    new offers in the catalogue, delivery quotes

  3. 14

    Documents

    proformas and invoices

  4. 26

    Copy and publications

    product descriptions, articles, posts

  5. 486

    Shop building

    new shop and panel features

  6. 470

    Quality control

    tests, fixes, supervision of the work

  7. 33

    A human approves

    team questions resolved by a human

A real day: 15.10.2026. Numbers rounded, no case contents.

Hour-by-hour data
Tasks closed in the hours of 15.10.2026, by workstation
HourSales and enquiry handlingOffers and transportDocumentsCopy and publicationsShop buildingQuality controlA human approves
00:001102030201
01:00080020200
02:00090020200
03:001101220400
04:00070420100
05:001100020200
06:001101020200
07:00270020100
08:0005006100
09:001100020200
10:004201120206
11:005100020300
12:005200020204
13:003200030102
14:009103210200
15:00793020200
16:00993020205
17:009100320205
18:00770230200
19:008100120204
20:008200720205
21:003200420200
22:00270020201
23:00160020200

The other side of the counter

Step into the customer's shoes. On the left you send an enquiry, on the right you watch what the shop's back office does in the meantime — right up to the button that a human clicks.

You as the customer

Language
Product

Dzień dobry, potrzebuję 2 palety Jinko 405W z dostawą. Jaka cena?

Shop back office

06:40 / 10:00

  1. Enquiry understood
  2. Price from the price list
  3. Transport calculated
  4. Reply in the customer's language
  5. Human clicks: send

Dzień dobry,

Jinko 405W: •••,•• EUR za sztukę przy zamówieniu 2 palet.

Dostawa pod wskazany adres: •••,•• EUR.

Pozdrawiamy, Casali Energy

Amounts are illustrative, blurred. Clock movement simulated.

Only a human clicks this button. Only then does the reply go out to the customer.

The shop as a building

Seven departments, each on its own floor, with a human on the roof. Hover over a floor or tap it — you will see what that department does for the customer.

  1. Human: decision and dispatch Reads, approves and sends. Every reply to a customer passes through their hands.
  2. Quality control Checks replies and documents before they reach the customer.
  3. Shop building Develops the shop so that ordering gets simpler.
  4. Copy and publications Writes product descriptions and articles in the customer's language.
  5. Documents Issues the proforma and the invoice once the customer confirms the order.
  6. Offers and transport Adds products to the catalogue and prices delivery to the customer's address.
  7. Sales and enquiry handling Answers the enquiry with a price and delivery cost, in the customer's language.

over 50 AI roles in seven departments — a human has the final say

What silence costs you

A customer waiting for a price keeps asking around — and buys where the reply came first. Work out how much in sales drains away at your company when enquiries sit unanswered.

Example: 120 enquiries per month, an order worth 8,000 zł, 15 out of 100 drop off

Lost sales per month
144,000 zł
Per year
1,728,000 zł

120 enquiries × 0.15 × 8,000 zł = 144,000 zł per month, which is 1,728,000 zł per year.

Formula: enquiries per month × the percentage that drops off (divided by 100) × average order value = lost sales per month; per year that amount times 12.

Do you want every enquiry at your company to get a price and delivery cost before the customer asks elsewhere? Let's talk about your enquiries.

Pass it to your boss

Does the decision belong to someone higher up? Below is a one-page summary: copy it, email it or print it — nothing goes through us.

Silvanos.ai × Casali Energy: a B2B shop run by AI

Silvanos.ai built for Casali Energy — a wholesaler of photovoltaics, energy storage and heat pumps — a B2B shop run by AI, in which a human only approves. The shop runs in production, and we build a shop like this in 4–6 weeks. The shop has 14,139 product cards in 8 categories and speaks to the customer in 13 languages; over 50 AI roles work in it, each with its own scope. The system prepares the customer's enquiry, price, transport and reply in their language — a human sends it.