Ars ex
Machina
Automatic

Automation for your business

Not a single step but the whole path, from enquiry to payment. Decisions stay with people. We started with our own wholesale business — now we automate client processes across many industries.

What we automate

Not a platform and not a box: we come into your processes as they are and take the manual work out of them.

01

Bots and AI agents

They triage new enquiries and send the person responsible a ready summary, turn calls into text, and nudge you when payments go suspiciously quiet. Scripts and integrations are our words to worry about, not yours.

02

Mail and documents

Incoming mail is machine-read: enquiries, invoices and statements land where they belong by themselves. AI reads the scans and drawings a person used to sift through.

03

Costing pipelines

Cost prices from a drawing or documents — minutes instead of days, using your rates and material prices. The result is a guide for the manager; the figure that goes into the quote is set by your specialist.

04

Integrations

We connect what you already use: CRM, spreadsheets, messengers, accounting. If you ever accidentally learn the word «webhook», we have slipped up somewhere.

How it works in different businesses

In our wholesale business the robots carry a delivery from the retail chain's order to the reconciliation with counterparties; at a manufacturing client, from the first enquiry to the quote; on our farm the dashboard is planning irrigation and robot work before the first planting (more in Apiary).

01

Wholesale supply to retail chains

Before: orders were retyped from portals and email, and shipment paperwork was put together from memory.

Now: each order arrives in the team chat as a card and is linked to the accounting records, the document pack is assembled to the chain's rules, and product documents are checked in government registers. A truck is booked once the goods are ready; bank statements and reconciliations run on schedule.

People decide: exceptions, the choice of driver, disputed amounts.

02

Made-to-order manufacturing

Before: enquiries and calls were sorted by hand, client records sat empty, and deals with drawings waited weeks to be costed.

Now: an enquiry reaches the person responsible already triaged, a call arrives as text, and the company record fills itself from government registers. A draft cost estimate comes out of the drawing in minutes; each morning there is a summary for sales, and when payments go quiet, a reminder.

People decide: whether to take on the enquiry and what price goes into the quote; the robot never writes to customers.

03

Farming and agribusiness

The field is still being prepared; first planting is in 2027. Before: the irrigation design sat in the engineer's PDF.

Now: zones, shifts, schedule and fertiliser sit on the field map, next to plot boundaries and satellite images. Tasks for the robots are set by field block, and a model shows how the fleet would cope. Valves and machines will carry the plan out once they are in the field.

People decide: what to water, where and when, and where to send the machines.

04

Your business

The path looks much the same everywhere: enquiry, work, documents, delivery, money, report. Every company has its own systems and rules, but the parts repeat.

We build your path from parts that already work and fit them to your systems: enquiry and email triage, a counterparty directory, bank statements, a first screen, watchdogs and a nightly database copy.

People keep working in their usual chats and CRM, and make the important calls themselves.

01 / Explore

Six steps.
One set of data.

This sample order shows how our running delivery workflow works. The retailer, product and quantity are fictional.

AXM–DEMO–104 Sample order

01 / Input

No need to retype the order.

In the running workflow, orders arrive through EDI. The system reads line items, quantities and delivery dates, preserving the link to the original order.

A person handles exceptions: an unknown product, a disputed quantity or an order change.

Incoming order / example

Sample retailer

Product
Apples
Quantity
120 boxes
Weight
1,200 kg

Source retained

02 / Shared record

The order gets one connected history.

Line items are matched with the accounting system. The order, register entry and upcoming shipment are linked together.

The person responsible agrees the matching rules. Unrecognised items need clarification.

Order record / example

AXM–DEMO–104

  • Product matched
  • Quantity checked
  • Original order available

No duplicate entry

03 / Retailer rules

Documents assembled by agreed rules.

The required documents depend on the retailer, product and delivery terms. The workflow selects files and generates labels from the order data.

A specialist defines the requirements. A missing source calls for a review, not a guess.

Document pack / example

Everything belongs to this order

  • Accounting documents
  • Product documents
  • Pallet sheets and labels

Contents listed

04 / Human decision

Exceptions reach the right person.

The record and document pack arrive in the team chat. The person responsible sees what was assembled and checks the data before use.

Automation removes data transfer work. A specialist remains responsible for document contents.

Team chat / example

Document pack for review

Order AXM–DEMO–104. Document contents and source data are available in one record.

Reviewed by a person

05 / Next part of the workflow

The shipment stays in view.

Documents and shipment details stay linked to the register. They can then support checks on deliveries, pallets and settlements.

Your workflow may be different. We agree its boundaries first, then adapt this approach to your operations.

Register / example

Order linked to shipment

  • Order data
  • Document pack
  • Record for reconciliation

History can be checked

06 / Reconciliation

Dispatched. Now reconcile.

Shipment details are compared with the register and bank statement. In this sample, the payment differs from the shipment amount: the discrepancy appears alongside its sources.

A difference does not yet establish a debt. The person responsible checks the payment reference, documents and adjustments, then decides what to record in the accounts.

Reconciliation register / sample amounts

AXM–DEMO–104

Shipment amount
RUB 120,000
Payment per statement
RUB 110,000
Discrepancy
RUB 10,000

Review by the person responsible required

The amounts are fictional. This example creates no accounting entries and makes no changes to the accounts.

This is a workflow illustration, not a connection to live orders. It uses no client data and sends nothing.

Automatic

The client's dashboard

Costing queue with the robot's verdict Costing card the manager can edit Rate book and metal prices Sales and acceptance analytics
Roles and permissions per project type What the robot did this week AI spend, per project Honest data age
Automation you can actually see

The bots work inside your chats and CRM, while the full picture lives in the AXM dashboard. The screenshot shows a manufacturer's costing queue: the robot reads the drawing, works out the cost price and shows how closely its estimates agree; the manager edits the draft and the total recalculates at once. The first screen shows what the robot did this week and what is waiting for your decision.

The rate book holds operation rates and material prices; some of those prices the robot pulls from metal traders' public price lists and tags with where and when they came from. Roles and permissions depend on the project type: a costing project and a wholesale one get different sections. AI spend is tracked per project, by operation and by model; our own measurements and tuning never count towards your spend. The screenshot here is a made-up client: real queues are seen only by their owners and the people the owner lets in.

Each business gets its own first screen: for a supplier to retail chains it shows what is on the road today and what is still missing for dispatch; for a farm, the field and the irrigation plan. People edit the robots' rules right here in the dashboard.

The automation view in the AXM dashboard: a costing queue with the robot's verdicts, cost prices and the specialist's own calculation; the client is fictional

How it works

01

Robots where the data is

The robots run on your server or on ours, whichever your security people prefer. What access to mail, CRM and spreadsheets is needed, we agree up front and write down: what we read, what we write, what we leave alone.

02

A feed of snapshots instead of direct database access

The dashboard never touches your CRM, mail or accounting system: the robots publish data snapshots, and the screen shows when a snapshot was taken. Edits made in the dashboard — rates, chain rules, corrections to a costing — go into separate tables the robots pick up on their next pass, and every action is logged.

03

AI only for reading and drafts

Neural networks read drawings, scans and emails and draft summaries — that is an external model, and what it gets to see we agree up front. The model estimates material, operations and hours; ordinary code turns them into roubles using your rates and prices — no magic: it behaves the same every time and leaves a trace in the log. A person signs off the robot's draft.

04

Who sees what

Access is run from a dedicated panel: accounts, projects, roles and per-section rights. A person can be created, blocked and signed out of every session, and each such action stays in the audit log. What someone actually sees depends on the project's workspace type and on their role in it.

05

Watchdogs at the handovers

Bad news should come from the system, not from a customer's phone call. If a flow stalls or data fails to reach the next step, the person responsible gets an alert; if a bot goes down, a watchdog brings it back. A quiet check makes noise only about new failures, not long-known ones. Every night all databases are snapshotted and self-checked, and a copy goes to a second machine.

03 / Evidence

It worked in our operations first.

Our automation grew out of wholesale fresh-produce deliveries. It is a running workflow with orders, accounting, documents and people in team chats.

In operation

Deliveries to retail chains

Automation carries each delivery from the retail chain's order to the money: orders and accounting records, document packs checked in government registers, truck and driver, letters of guarantee, bank statements and scheduled reconciliations with counterparties.

Context
Our own wholesale business
Scale
More than fifty automations running in the order–documents–shipment flow, per our internal audit of 3 Oct 2026
Status
Based on operational documentation as of 4 Oct 2026
Boundary
People review rules and exceptions. Implementation results are assessed separately for each process.
AXM automation dashboard with a sample calculation queue
A calculation-project dashboard. The interface works; the client and records shown in this image are demonstration data.

How we investigate data before implementation

04 / Getting started

Start where
you copy by hand.

01

Understand one process

We inspect the input, result and exceptions. We agree who approves the data and how to tell whether the process has improved.

02

Agree access and a pilot

We record what we read, what we change and which data may be shown to an external model. We test within a limited scope.

03

Launch and support

We train the people responsible and provide an action log and an error-handling procedure. Support terms are agreed in advance.

Tell us what data you transfer, from where and to where. That is enough for the first conversation.

Discuss your process