The request comes in, the draft is already there.

An agent reads emails and attachments, recognises the customer and the items, checks prices and availability in the ERP and prepares the order and the reply. Whoever works in the office checks and confirms.

Sectors
manufacturing, distribution, wholesale
Used by
sales office, customer service
First prototype
2–3 weeks

The problem

Orders arrive in every shape: a three-line email, a PDF with twenty item codes, a photo of a paper form. Someone has to read them, work out who the customer is, find the items in the ERP and copy them in by hand.

The work itself is not hard, but there is a lot of it and it cannot wait. On busy days confirmations slip to the afternoon or the next day, and transcription errors come to light when the goods have already shipped.

How the data flows

  1. Where it comes from: Emails and attachments (PDF, Excel, photos); WhatsApp (messages and images).
  2. What happens, Agent: AI, Reads customer and items (even in the customer's words); Checks the ERP (prices, discounts, stock); AI, Prepares order and reply (order to confirm, draft).
  3. Who checks: Sales office (checks and confirms).
  4. Where it ends up: Confirmed order (in the ERP); Reply to the customer (email or WhatsApp).
  5. every correction refines the rules.
Steps marked AI are the model's. The rest is ordinary automation or people.

How we solve it

04
01

It reads whatever arrives

Text, PDF, Excel, images. It recognises customer, lines and item codes even when they are written the way the customer writes them rather than the way the catalogue names them.

02

It checks the ERP, not its memory

Prices, customer discounts, stock and lead times come from the ERP, with the permissions of whoever uses it. If a piece of data is missing it says so, instead of guessing.

03

It prepares, it does not send

The order is entered but not confirmed, and the reply to the customer stays a draft. The last click belongs to whoever looks after the customer, and every correction is logged to refine the rules.

04

It hands over when needed

Complaints, discount requests, items outside the catalogue, new customers. Whatever falls outside the expected cases reaches a person with the summary already done.

What we measure

03

These are the numbers we set in discovery, before writing any code. The starting value is measured on your data, and the prototype is judged against it.

Time to confirmation

From the order's arrival to the reply to the customer. It is the number the customer sees, and usually the first to move.

Drafts confirmed unchanged

How many orders go out as the agent prepared them. It is the most honest quality indicator, because it is decided by whoever checks them.

Lines copied by hand

Transcription errors on codes and quantities, before and after. They are few, but each one is a wrong shipment.

When it is not worth it

If orders are few or always the same, a well-made order form costs less and works better. The agent is worth it when the format is decided not by you but by the customer.

Services involved

Other use cases

11 scenarios
02 · LLM integration

Quotes and estimates

Specifications, drawings, three-line requests: for anyone working to order, every quote is a job in itself. AI finds similar past jobs and proposes line items, times and costs to validate.

03 · Process automation

Email triage

The info@ inbox receives everything and someone has to sort it by hand. AI reads every email, gives it a category and forwards it to the right department. It replies to nobody: it routes.

Tell us the process. We will tell you if it makes sense.

A free 30-minute call, with no commercial follow-up unless you want it.

Request a call

Companies that have already worked with us

  • Velux
  • ISU
  • MovePay
  • AB Group
  • Archiva Group
  • Onda SpA
  • PDE
  • Bynary System
  • COOPI
  • Gimasi
  • Creactives SpA
  • Rev AI
  • Odditt Inc.
  • Si Sociale
  • Stones
  • Turboden
  • Schindler