DOLMIR

01Intelligence · AI systems

Intelligence built around your company’s context.

A language model, on its own, knows everything and nothing about your company. A useful AI system is built around what the company knows: documents, data, rules, history, the people who decide. And it has a human gate wherever judgement is needed.

The problemThe information exists but it is scattered, and today someone reads it, interprets it and copies it across.

AI systems

layer 01

  1. Reads

    Emails, attachments, documents

  2. Evidence

    Every value with its source

  3. Verifies

    Customer records, catalogues, rules

  4. Stops

    Where judgement is needed

  5. Today

    Prototype on requests for quotation

When it is needed

  1. 01When the information exists but is scattered across emails, documents and systems.

  2. 02When someone today reads, interprets and copies — and that person is the most experienced in the office.

  3. 03When the right answer depends on context, not on a fixed rule.

What we build

  • Reading and understanding emails, attachments and documents

  • Data extraction with evidence: every value with its source

  • Checks against the company’s customer records, catalogues and rules

  • Replies and drafts prepared, never sent on their own

  • Internal assistants that consult the company’s data and cite what they read

  • A human gate wherever judgement is needed

Possible application

Reading with evidence

Illustrative scenario on a demonstration company: simulated data.

Reading with evidence

decision required

Officine Rossi S.r.l.

09:12

Request for quotation — 2,000 pcs SL-441

Customer

Officine Rossi S.r.l.

sender · customer records

Quantity

2,000 pcs

email text, line 1

Code

«SL-441» → SL-4410 or SL-4415

catalogue: two items match

Outcome

  • ✓Reply to the customer prepared
  • ✓Open question passed to a person

What exists today

Reading, extraction with sources, verification and a prepared reply are the DOLMIR Software prototype on the request-for-quotation process, tested only on simulated data. The conversational system with tools that you can question on this site runs on a demonstration company with simulated data.

What is possible, after the survey

Internal assistants on company documents, reading other kinds of document, other office processes: built to measure, after the survey.

What it is not

It is not a generic chatbot and it is not «AI that transforms the business». It is office work done with evidence, and with a person who decides.

The other layers

Tell us the problemAll solutions