We implement AI that already works in our own logistics

A request gets a route across the network of legs, timing and ready haulage assignments; trips are assembled automatically and anything debatable goes to a planner for approval. Keeps the fleet and drivers, local delivery routing and the delivery-time matrix for the client portal. Replaces the workplaces of the previous transport system.

A cold-chain quality operating system: one loop from signal to investigation and recurrence prevention, plus field warehouse audits from a phone. An AI agent fills in client and auditor questionnaires from the company's procedures; a specialist checks the answers.

Our own temperature-control ecosystem: a line of thermal containers, sensors, data-collection gateways and a web portal. Sensors log temperature, humidity, opening and shocks along the whole route, and the data is read via Bluetooth.

Not tickets but a case: client, shipment, deadline, owner, evidence. Deadlines recalculate automatically, escalation raises itself, AI drafts the replies. The client rates how the case was handled, and AI reviews the rating together with the case history.

AI parses the incoming request: addresses, weight and dimensions, temperature mode, named thermal packaging, up to five routes per email. Pricing uses current tariffs, contract or standard, with thermal containers selected; out comes a finished quote with a delivery date.

The central exchange bus between systems: one message envelope, versioned contracts, idempotency, retries with backoff, a dead-letter queue and end-to-end tracing. No new infrastructure — it runs on the database you already have.

The driver self-registers at a kiosk on arrival, the truck joins the queue on the yard board, an operator puts it on a free gate, and a voice assistant immediately calls the driver with the gate number. No more manual registration at the barrier.

The layer through which assistants and agents work with accounting systems, CRM and external APIs over the open MCP protocol: permitted operations, not raw database access. The assistant files expense reports and per diems this way, from receipts to a payment statement after approval.

A single catalogue of working reports — finance, operations, logistics, clients — with access rights. Each card shows when its data was updated and whether the source responds, so a stale figure is not taken for a fresh one. No purchased BI platform or licences.

An electronic transport documents gateway: a finished trip becomes an electronic consignment note and goes to the state system through an EDI operator, by a click or automatically. The log shows which signing stage each note has reached.

Merges into one picture what is scattered across five disconnected systems: the issue tracker, repositories, time tracking, assistant requests and the project office. One registry of people, one event feed and dashboards on top.

Takes a request from any source, assembles the electronic forwarder order in the tax-authority format, signs it and sends it through an EDI operator. Installs alongside your live transport system and ships for installation inside your own environment.

An approvals engine for contracts: route, deadlines, owners, document pre-check and a decision timeline, showing who holds it and how long it has waited. Website offer requests arrive on their own, then go through a counterparty check, e-signing and client portal access.

A pharma warehouse system: receiving and dispatch against client orders, drug-marking codes scanned on a handheld or phone and checked in the state tracking system, exchange with the accounting system. Rewritten on a modern stack so it can be developed, not merely operated.

Computes cost and margin per lane and brings a prioritised action list: what to raise, what to hold for volume and where to check the carrier. It also hosts the freight procurement desk: a new subcontractor's form from the website is checked against open state registers and becomes a card, and carrier rates are compared with the market.

Automated checking of actual-cost documents: the carrier tariff on the trip date, the amount formula, mileage, duplicate requests and double payments. The result is a traffic light with an explanation; the decision stays with a person.

The mobile app for our couriers and subcontractors: the day's trips, route stops in order and on a map, loading and unloading check-ins, photos, and returns of thermal boxes, sensors and documents. It works offline too: check-ins and photos go out as soon as there is a connection.

Tags on thermal containers, trolleys of cooling elements and sensors: readers at warehouse doorways and handheld terminals record what left and what came back. No manual recounts and no lost reusable packaging.

Clients place orders, price a delivery, follow it stage by stage, see their warehouse stock, download documents and rate the shipment themselves. Fewer calls to the account manager over every detail.

A client emails an order on their own form. The service reads it, checks addresses, phones, temperature mode and contract, and creates the order in the accounting system. If data do not match, the manager sees why and fixes it in a form.

Keeps stock of cold packs and thermal boxes, checks sufficiency against orders and tomorrow's forecast, and moves packs through readiness zones. The dispatch gate blocks unprepared or reused packaging; AI proposes actions and a person confirms.

A client or carrier e-mails a truck pass request; AI parses it, checks the driver's details and issues a pass for the warehouse gate. If data is missing, the client gets a question; if parsing fails, an operator takes over.

A dedicated portal for couriers and carriers: assignments, route, documents and earnings. Fewer calls to the dispatcher and less paper in hand.

A team workspace with an AI assistant, in the browser and on the phone: shared access, roles, working rules, action permissions and scheduled tasks. Staff work under a corporate account with shared rules instead of a personal account in a public service.

Tariffs moved out of the accounting system into a separate service: rate grids, services, pickup zones, contract discounts and mark-ups, and a history of every tariff change. The client portal, the website calculator and automated quotes get price and lead time from one engine.

The air dispatcher's workspace: cargo is assigned to air waybills and flights, the trip goes to the transport system on its own, booking statuses and airport departures update automatically. If a trip is rejected, the reason and next step are shown — no spreadsheets.

Cargo labels in one consistent format; scanning confirms receipt and dispatch and rules out mix-ups.

A dashboard of delivered orders with something left open: no waybill or temperature report, no delivery scan, no sensor data. A management summary and a planner's work view show what to handle first — payment for the service and compliance depend on it.

Picks thermal containers for the dimensions and number of cargo pieces: how many, which ones, and whether the cargo may be turned over. The same calculation runs inside automated quotes, so the packaging in a price is computed, not guessed.

A bonus-metrics service for transport, warehouse and the commercial unit. Data is pulled from accounting systems on a schedule; for each employee the target, actual and bonus are visible, and driver mileage is checked against the trip before payment.

One home page for every HR service, with tiles, role-based access and quick search: an employee no longer needs to know which system holds the task. Managers get a meeting report on new starters and headcount, plus hiring cost and funnel trends.

Accounting-system configurations and their exchange logic live in repositories: every change has an author, a history and a way back. Each night the exchange logic in production is checked against the repository, so a change made around it shows up by morning.

Checks orders from the transport system: closing documents, signatures and stamps on the waybill, deadlines in the local time zone. The job goes to the right carrier's work chat; on pilot routes the bot asks about cargo arrival and escalates if nobody answers.

Warehouse photo capture of freight units: every shot is bound to the unit, time and order, so a dispute about cargo condition is closed with a fact, not a memory.

A financial model of the company on accounting data: cash, receivables, taxes, margin, a monthly revenue forecast and scenarios showing how a decision on tariffs, volume or costs will land in the result. On weekdays management gets a short morning summary of the main signals.

Versioned definitions of the messages systems exchange. A contract describes the message, not the service code, so a system can be replaced without rewriting everyone who talks to it.

One list of customs and international freight orders: event timeline, cargo status in the list, permits with deadlines, export-ban check, CMR form, client status e-mails after a preview and a manager's report. Changes flow back to the accounting system via the exchange gateway.

AI reads inbound client emails with thread history and attachments, identifies the topic and adds client data from CRM. Routine requests like a freight quote or truck pass are handled automatically; everything else reaches an employee as a case with context and a draft reply.

Tracking thermal boxes issued to and returned by couriers: who holds what, what is overdue and what never came back.

Reconciliation of registers and settlements across systems: discrepancies are found automatically and land in a work list with the reason attached.

Tracking the return of signed documents for delivered orders: plan, actual and overdue by client. Money stops hanging because of paperwork.

Last-mile delivery cost by lane and carrier, next to the sales price from current tariffs and the margin on it. You see where a tariff runs at a loss before the P&L shows it.

Gathers requests marked 'ready to plan' into FTL trips and creates the trip in the transport system — a fast screen for one job instead of a general-purpose accounting UI.

Warehouse shift metrics on a screen: operations, pace, backlog. The shift lead sees a dip now, not in tomorrow's report.

A request from the client portal becomes an order under a one-request-one-order rule; anything that does not assemble lands in a manual queue with the reason in plain words. The order goes to the logistics planner, the shipment status comes back to the card, and the full change history is visible. The accounting system is reached only through the exchange bus, never by writing straight into its database.

A management P&L assembled from accounting data instead of a hand-built spreadsheet that lands on the tenth of the month.

Receivables and overdue balances by client and ageing: you can see who to work with today, not at the end of the quarter.

A single page of key operational metrics for a screen in the office or the warehouse: it refreshes itself and can be read in passing.

HR sets up a review, the employee fills in a questionnaire via a personal link with no login, and duties are pre-filled from the job description. HR scores seven criteria and exports a document for management; the employee never sees the scores or HR's comments.

One registry of contractor agreements instead of a spreadsheet: terms, renewals, rate history, documents and alerts about what is about to expire.

Shared tokens, themes and components for every product: interfaces look and behave the same way, and a new screen is assembled from ready parts instead of being drawn from scratch.

Several agents check orders and payments every night: a service delivered but not invoiced, an order losing money on actual bills, a tariff left unrevised, a duplicate payment, a client who stopped ordering. Each finding comes with the calculation; a person decides.

An hourly picture of operations from the accounting and transport systems: orders, weight and volume, thermal boxes, trip revenue and costs, and execution of the transport plan. It also links straight to problem orders, delivery times and box documents.

A recruiters' platform instead of a shared spreadsheet: who started, how probation is going, what is overdue and who should be introduced to the team.

Every night it turns calls into text, links them to the customer in the CRM and tags topic, outcome and tone. A morning summary follows, and a tagging failure raises an alert: agreements are not lost and it is clear what customers call about.

A workspace for controlling temperature sensors per shipment: mode compliance, lost connection and free suitable sensors for replacement.

A live map of how systems actually exchange data: assembled from the data itself instead of drawn from memory. It shows the links nobody remembers any more.

Builds a passport for every service: what it does, where it runs, whether it is alive, what it talks to and who owns it. Passports rebuild themselves whenever something changes on the servers, so the question 'what do we even have' stops depending on an admin's memory.

The infrastructure layer: deployment, services and environment builds are described and repeatable. A new service is brought up from a template instead of being hand-assembled every time.

An interactive order report: plan versus actual, orders with no fulfilment, orders with no amounts and a year-on-year comparison.

A dashboard portal on how units work: load, output and trend, comparable across departments.

Regular departmental reporting on one method: comparable numbers instead of a different spreadsheet for every manager.

Computes a department's cash-collection metrics and distributes the bonus by accounting rules instead of by an argument at month end.

Accumulates, day by day, the metrics clients actually pay for: delivery time, documents and cold-chain compliance. The question 'did it improve?' is answered with a chart, not a feeling.

Three times a day it gathers labour-law changes relevant to the company, adds a short AI analysis and posts a summary to the HR team's chat. Weekly: a calendar of what comes into force. The analysis is preliminary; a specialist decides.

Our own website counters, campaigns and tags, leads and calls from the CRM, search queries and how visible the company is in AI assistants — in one workspace. A radar raises the alarm by itself: a lead with no source, a chat with no reply, a data feed that has stopped. A person investigates and decides.

Language models running on the company's own servers: data never leaves the perimeter, the cost per request does not depend on an external price list, and access goes through one gateway.

A spending request follows a route: department, financial control, treasury. The approver is set by expense item and department, with stand-ins during absences. An approved request reaches the accounting system as a finished document, with no re-keying.

Tracking returnable packaging: where it is, who holds it and how much. Losses stop being an invisible cost line.

A workspace for producing thermal boxes and sensors: production orders, bills of materials, release with serial numbers, label printing and moving finished goods to the warehouse. Everything reaches the accounting system through the exchange bus rather than by writing straight into the database.

Robots watch for requests nobody has answered. A client waiting in chat triggers a reminder to the operator and then, step by step, to managers; a missed call becomes a call-back task for whoever missed it, escalated if no conversation follows.

Screens applications from the recruiting system: each candidate gets a score, a verdict, a factor-by-factor review, risks and interview questions. The recruiter takes a candidate forward, puts them in reserve or declines with one click. Sourcing and a first interview come next.

Assembles each order from bus events: stages from intake to closure, skipped steps, cargo temperature and the status of electronic documents. You can see where an order stands, while domain data stays in the source systems.

The service takes tasks from the change queue itself, does them, deploys to a staging environment and verifies the result — both code and live interface. It does not invent answers to open questions: it asks the owner and continues after the reply.

The voice agent answers when the team could not pick up in time: it finds out what the call is about, the order number and a contact, and answers general questions from the website's content. The request immediately becomes a task for the owner and a case in the Service Desk — decisions and promises stay with people.

Corporate training in working with AI: courses, knowledge checks, practical tasks and personal progress. Plus an ideas bank: an employee describes where AI would help in their work, and the idea gets scored against criteria with recommendations; there is a monthly AI-solutions contest.

Answers questions from uploaded documents — reference guides, scripts, training materials — in the documents' own words and always shows the source. If the documents have no answer, it says so. The internal AI assistant runs on this knowledge base.

The strangler-facade approach: a new service stands next to the working system, both compute in parallel for a while and the load moves across piece by piece. No business downtime and no big-bang cutover.

Collects unposted carrier documents, matches them against operational facts and tariffs and assigns a traffic light with an explanation. Posting stays with a human.

Builds a map of corporate database structure: tables, domains, relations and lineage. Useful both to a person investigating and to an agent working with the data.

Every night AI reviews the changes that reached production services over the past day, looking for bugs and data corruption. Two independent agents then try to refute each finding against the code; in the morning the author gets only what both of them confirmed.

A robot dog in the warehouse takes voice commands, answers by voice and describes what its camera sees; the operator works from a console with a map and cameras and gets an end-of-shift report. Next come accompanied warehouse audits and an autonomous patrol route.

Receipts and dispatches for custody storage are created in a web workspace and travel to the warehouse over the exchange bus in the same format the accounting system used to send. If a document cannot be sent, the reason shows at once instead of being discovered later through a missing receipt at the warehouse.

A system for internal policies and employee initiatives: versions, sign-off, review deadlines and rewards for accepted ideas.

From task to change: a task enters the tracker, AI triages it and assigns a project, an agent makes the change in its own branch and opens a merge request, and a person accepts the result. Progress shows on the task card.

Choosing a delivery mode across several transport types: comparison by transit time, cost and temperature risk rather than by an account manager's habit.

One access point to the accounting core: external services reach 1C through a gateway by contract instead of connecting to the database directly.

Orchestration of an agent team: a lead agent hands tasks to workers, they run in parallel and report back, and questions that need a person reach them with a separate flag. It works over every module and proposes actions in plain language.

Warehouse documents are recognised and parsed automatically: a person checks the result instead of typing it off the page.

Inbound accounting documents are recognised, parsed and reconciled automatically. An accountant reviews instead of typing.