A team of AI agents on every deal.

One AI core runs procurement, finance, construction and sales. Agents do the work, verifiable code does the money math, and a person makes the call.

A team of AI agents on every deal.

Every deal is run by a team of agents, each with its own role

Coordinator

Picks up the event, assigns the work and oversees the whole deal

Demand agent

Reads the request or specification and maps it to standard items; sees upcoming demand from the works schedule

Procurement agent

Finds and vets factories, pools orders, runs the tender and negotiations

Logistics agent

Runs the carrier tender, booking, documents, customs and landed cost

Finance agent

Prepares bank applications, compares terms and models the deal's cash flow

Limits control

Tracks guarantees, letters of credit, credit lines and budget risk

Quality agent

Checks each batch against the sample, handles claims and rates the factory

In procurement, agents run the entire path: request, item, supplier, request for quotation (RFQ), comparison, negotiation, finance, logistics, contract, approval, delivery, analytics.

AI does the work. Code does the math. People decide

AI does the work. Code does the math. People decide

AI does the work

It understands a request written in plain language, finds suppliers, prepares RFQs, breaks down quotes and documents, handles correspondence and flags anomalies.

Code does the math

Landed cost, financing cost, limits and savings are calculated by versioned rules. Rerun the numbers on the same data and you get the same result. The language model never touches the money math.

People decide

A commitment to a factory, bank or carrier exists only after a person signs. The agent brings a proposal with a confidence score, not a finished action.

You decide how much to hand over

  1. 1Suggestion
  2. 2Action with approval
  3. 3Autonomous work within set limits
  4. 4The full process

Each area of work starts at the level you choose and moves up once the results are confirmed. The platform is built so routine operations run without human involvement. People stay in the loop where lawyers, banks and carriers make the decisions.

Every purchase makes the next one sharper

Who quoted, who ran late, who shipped defects, which bank approved faster, which substitute the designer accepted. These facts build up with every deal. When an agent makes a mistake, the mistake becomes a rule written in plain language. The rule is visible, it can be reversed, and it is tested against past deals before it goes live. One client's data is never visible to other clients.

Every purchase makes the next one sharper

Magic comes to you with the numbers first

The AI core sees your upcoming demand, global prices, freight rates and bank limits. When buying now makes more sense, you get a notification with the amount and a ready-made RFQ.

Every agent action is logged

Companies don't buy a model. They buy a closed piece of work: a handled request, a verified invoice, a placed order. Between the model and the finished work sits a layer that Magic builds once and uses across every module and enterprise deployment: process memory, access rights, an action log and error rollback. Each industry adds its own reference data, integrations and scenarios.

Every agent action is logged

Technical facts

QuestionAnswer
Where is the data storedIn your environment or a dedicated one; isolated deployment is provided for
Are models trained on our dataNo. The ban on training models on client data is set out in the contract with the client
What happens if an agent makes a mistakeThe action is reversed inside Magic; for external steps, the log keeps the rationale and source data needed to correct it
Who sees whatRole-based access within your company; other clients don't see your volumes or prices
AI transparencyEvery agent action appears in the log with its rationale
LanguagesHebrew, English, Russian, Chinese
Infrastructure and securityCloud infrastructure, perimeter protection and threat monitoring; security processes are being built to recognized international standards; the list of standards is published after audit
Neutrality and rankingRanking rules are the same for all participants and are set out in the contract

Let us show you how Magic would run your deal.

Join as a pilot partner