Home · Solutions · Supply chain
Solution · Supply chainOrders arrive in nine different shapes and leave as one clean record in your TMS
Transport orders into the TMS without retyping
Orders from e‑mail, customer portals and file drops are read, matched to the customer, the lane and the agreed rate, and created in the TMS; amendments update the order they belong to.
Executive summary
Ninety customers, ninety ways of sending an order, and a forwarder retyping every one into the TMS.
What Mientha builds is an intake layer in front of your TMS, not a replacement for it.
Orders exist in the TMS minutes after they arrive, including the ones landing at half past five, so the search for a carrier starts the same afternoon.
the TMS (order creation and amendment); SharePoint source archive; Power BI
Business problem
Freight forwarding
A forwarder sells coordination, and the transport order is where it begins. In practice it carries the parties, the stops with their date and time windows, the goods, the weight and the pallets, the equipment the vehicle needs, and the rate. Everything the company invoices afterwards copies what somebody typed into that form.
The channel is the customer's choice, not the forwarder's. Ninety customers means ninety habits: a PDF by e‑mail, a portal read off the screen, a spreadsheet with twelve loads on Friday afternoon, and a few EDI links that carry the header and leave the instructions in free text nobody parses.
Amendments are where the damage sits. A pickup moves by a day, a consignee changes, a pallet is added, and the message arrives as a reply to a thread or as a fresh PDF with the same reference. Under pressure the desk creates a second order instead of finding the first, and one shipment exists twice.
The pattern repeats every afternoon. Tomorrow's orders arrive between three and six, which is also when carriers have to be found, so entry queues behind placement. Nobody can say how many orders are in the building, and the answer to "can we take this customer on" is a question about headcount.
How it works today
This is the shape of an order desk before anything is automated, whatever the TMS.
- PersonA forwarder works down the unread list, deciding for each message whether it is a new order, an amendment, a question or a carrier reply
- PersonAttachments are opened and read; portal customers are handled by logging in and copying fields off the screen
- SystemAddresses, references, goods, weights, pallets, equipment and time windows are keyed into the TMS, and the rate is checked against a table in a shared folder
- WaitingOrders arriving after the afternoon peak wait for a free forwarder, sometimes until the next morning
- Risk of errorAn amendment is entered as a new order, or missed, and a load moves on a date the customer has already changed
- Risk of errorFree-text instructions, tail lift, temperature, ADR, pallet exchange, are read once by one person and never reach the carrier order
Why the current process costs more than it appears
Time that disappears before anyone measures it.
- Retyping can be measured; what follows a mistyped field cannot. An address copied wrongly from a PDF sends a vehicle to the wrong industrial estate, and the cost arrives as waiting time, a missed unloading slot and an apology.
- Amendments count as work only when somebody notices them. A duplicate order placed with a second carrier surfaces at the ramp or on the carrier's invoice, and by then the choice is to pay it or argue about it.
- Response speed decides who gets the next load, and nothing about the arrangement scales: customers using three forwarders on one lane compare who confirms first, while every new customer brings another format, another portal login and another set of house rules.
Cost of inaction
A customer does not complain about a slow confirmation. It gives the next load to the forwarder who answered first, and the loss appears at the annual volume review as a lane that quietly halved. Nor is the standing time in the table above, paid when a duplicated order puts two vehicles at one ramp.
The other thing that grows is dependency. Which customer writes the reference in the subject line, which one sends amendments from a different address, which one means Tuesday when it writes "ASAP": that knowledge sits with three experienced people and leaves the building with them.
A plausible organisation with realistic proportions. The figures are there to be recalculated on your data; they are not a client result.
A Polish road freight forwarder with four branch offices, about 140 staff and no fleet of its own, placing work with roughly 400 contracted carriers; 90 active customers, its own TMS and Microsoft 365 E3.
5,200 transport orders a month: about 55% as PDF or spreadsheet attachments in four shared mailboxes, 20% from customer portals, 15% as files on SFTP or SharePoint, 10% as EDI carrying the header only; roughly 1,100 amendments a month on top.
Twenty-six forwarders read each order, key it into the TMS, check the rate against a table in a shared folder and confirm by e‑mail. Amendments are handled by whoever recognises them.
About nine minutes per order across entry and the amendments that follow it, concentrated in the three hours when tomorrow's loads also have to be placed. Orders arriving after six are tomorrow's work.
Robots collect orders from the mailboxes, the portals and the file drops; extraction reads the fields a transport order needs; matching resolves the customer, the addresses, the lane and the contracted rate before anything is created. Ambiguous cases reach a forwarder in Microsoft Teams with the document beside the proposal.
In the modelled case the desk gets back 780 hours a month, orders arriving at half past five exist in the TMS before anyone has read them, and duplicated shipments stop being weekly. The figures are a model, not a client measurement.
Proposed solution
What Mientha builds is an intake layer in front of your TMS, not a replacement for it. Robots watch the four shared mailboxes through the Microsoft Outlook 365 connector, collect files from SharePoint and the SFTP drop, and sign in to the portals that offer nothing else. Every item becomes a queue entry in UiPath Orchestrator, so the backlog gains a number and an age.
Reading is half the work. UiPath Document Understanding extracts the fields with models trained on your customers' own orders, and a per-customer field map absorbs the fact that "ref", "nr zlecenia" and "PO number" mean the same thing to three senders. Extracted fields are matched against your master data, address book and contracted rate table before anything is written, and what survives is created in the TMS with its document filed against it.
Amendments are handled deliberately, because that is where the money leaks. Each order carries a fingerprint from the customer's reference, the loading place and the original date, so a later message matching one is written as an amendment with a record of which fields moved, not as a second shipment. What the rules cannot settle goes to Microsoft Teams: an unknown consignee, a rate outside the contracted table, a window that cannot be met. The forwarder sees the document beside the proposed order with the reason stated, and a correction that proves to be a customer habit becomes a rule.
UiPath Document Understanding with Validation Station; UiPath Orchestrator queues, triggers, retries and audit; UiPath Integration Service connectors for Microsoft Outlook 365 and Microsoft OneDrive & SharePoint; UiPath Action Center actionable notifications in Microsoft Teams; Power BI
The intake across mailboxes, portals, file drops and EDI; the per-customer field map including free-text instructions; the matching rules for customer, address, lane, rate and window; the amendment fingerprint and change log; exception routing in Teams; the source archive and dashboard
Order creation and amendment in your TMS through whatever interface the product offers, an API where published and a file exchange where not; portal reading by robot where a customer offers no other channel
How the automated process works
- AutomationA new message in the four shared mailboxes, a new file on SharePoint or SFTP and a scheduled portal check all create the same queue item with its source document
- AutomationExtraction classifies the item as a new order, an amendment or something else, and reads the fields the TMS form needs, including free-text instructions
- SystemRobots resolve the customer by tax number, the stops against the address book and the price against the contracted rate table, and check the window against the distance
- AutomationOrders passing every rule are created in the TMS with the reference, the stops, the goods, the equipment and the agreed rate
- AutomationA message matching an existing order is written as an amendment to it, with a change record naming the fields that moved
- PersonAnything the rules cannot settle becomes an Action Center task in Microsoft Teams: source document, proposed order, reason, decision
- AutomationThe customer is confirmed as soon as the order exists, and Power BI shows intake by channel, exception reasons and confirmation time
Human-in-the-loop model
Automation handles
- Collection from mailboxes, portals, file drops and EDI, with the source document kept beside every order
- Extraction and per-customer field mapping: addresses, references, goods, equipment and time windows
- Matching against the customer master, the address book and the contracted rate table, and creation in the TMS
- Amendments written against the existing order, the change record, and the confirmation back to the customer
People decide
- Whether to accept an order whose rate falls outside the contracted table, and at what price
- What happens to a time window that cannot be met, including what the customer is told
- New consignees, new loading places, and anything the address book does not recognise
- Whether an amendment touching a committed load is accepted, refused or repriced
Before and after
Systems and integrations
The stack is deliberately short: one engine, one execution layer, one place where a person decides.
Inputs
- four Outlook shared mailboxes
- customer portals
- SFTP and SharePoint file drops
- EDI messages
Automation layer
- UiPath Orchestrator
- UiPath Robots
- UiPath Document Understanding
- UiPath Integration Service
- UiPath Action Center
Target systems
- the TMS (order creation and amendment)
- SharePoint source archive
- Power BI
Human touchpoints: Action Center tasks in Microsoft Teams; the branch intake channel in Microsoft Teams; the daily desk summary
Technologies used
reads order fields from PDFs and spreadsheets with models trained on your customers' documents; Validation Station for low-confidence fields
Aqueue every order and amendment, run the matching rules, write to the TMS, retry and audit
Awatches the shared mailboxes and the file drops, archives every source document
Aexception tasks completed beside the source document, without leaving Teams
Athe order desk's exception queue and the daily intake summary per branch
Aintake by channel and customer, exception reasons, amendment rate, confirmation time
Athe target system, addressed through a published API or a monitored file exchange
CIllustrative economic model
Numbers you can check against your own data.
Entry time varies more by customer than by order: a repeat lane in a format the desk knows takes three or four minutes, an unfamiliar consignee with a hand-written window and a rate to verify takes closer to twenty. Nine minutes is the blend across ninety customers, with the amendments that follow an order counted inside the same nine, and €24 an hour is a fully loaded cost of a forwarder's post in Poland. Neither figure was measured at a client.
Run the numbers on your data
An illustrative estimate from your own inputs. It models released capacity; it is not a promise of savings.
Business benefits
- Orders exist in the TMS minutes after they arrive, including the ones landing at half past five, so the search for a carrier starts the same afternoon
- The forwarder's day moves from typing to placing loads and talking to customers, which is the part of the job that earns the margin
- Amendments update the order they belong to, so duplicated shipments and carriers arriving on the wrong day stop being weekly events
- Every order carries its source document, its fields and the rules applied, which is what a rate dispute or a damage claim needs months later
- A new customer is onboarded by adding a field map rather than a person, and confirmation speed no longer depends on who is on shift
The management view
- The head of forwarding can say at any hour how many orders arrived, how many are in the TMS and how many wait for a person, per branch and per customer
- Volume growth stops translating into recruitment on the order desk, which changes what a new customer costs to serve
- Exception reasons are counted and the house rules of ninety customers move out of three people's memory into configuration
Board-level KPIs
Security and governance
An auditor should be able to reconstruct every decision.
- Robots sign in to the TMS with their own named accounts, holding only the rights to create and amend orders; no forwarder's login is shared and every write is attributable to a queue item
- Mailbox access is granted through Microsoft Graph with permissions scoped to the four order mailboxes and nothing else; portal and TMS credentials sit in a credential store, rotated on your schedule
- Order documents, extracted fields and decisions remain in your Microsoft 365 tenant, and the robots that process them run from the EU region of UiPath Automation Cloud
- Rate tables and address rules are versioned with a named owner, because both decide what the company charges; changing either is an approved change, not an edit in a shared folder
Why now
The modelled 780 hours a month are consumed in the three hours when the desk is worth most, which is when tomorrow's loads also have to be placed; that is the real price of the arrangement
Poland performed the largest volume of road freight transport in the EU in 2025, 381.0 billion tonne-kilometres, 20.2% of the EU total (Eurostat, published 31 July 2026), and in a market that size a forwarder is compared on how fast it confirms
Extraction trained on your own documents, mailbox and file triggers and human tasks delivered into Teams are standard capability now; and since mandatory KSeF e‑invoicing began on 1 February 2026, a freight invoice can only be as clean as the order behind it
Relevant executive roles
How many customers the company can serve stops depending on how many people can type.
The afternoon peak stops deciding what gets done tomorrow, and the order desk stops being the constraint nobody names.
Ninety sets of house rules move into configuration, exception reasons become countable, and the desk's day goes to placing loads.
Every order is complete, matched to a contracted rate and evidenced by its source document, which is what a clean invoice and a defensible margin per shipment depend on.
Common questions and objections
They do not have to, and asking them is not part of this. The channel stays their choice: a PDF, a portal screen, a spreadsheet or an EDI message all become the same queue item, and the field map absorbs the differences instead of a person's memory.
Which is why nothing ambiguous is created. An uncertain field, a consignee outside the address book, a price outside the contracted table: each stops the order and puts it in front of a forwarder with the document beside it.
The intake, the field maps and the matching rules sit outside the target system; only the write step changes. Reading a year of your own orders is also the cheapest way to find which address data needs cleaning before the migration.
When this is not the right solution
- Intake below roughly a thousand orders a month, where a disciplined mailbox routine and two experienced people do the job for less
- Intake that is already genuinely structured, with EDI carrying the special instructions as well as the header; there the work is mapping and validation, not reading documents
- Orders arriving mainly through load boards such as Trans.eu or TIMOCOM, a separate channel with its own commercial rules that is normally kept out of a first release
A question for the next management meeting
Should the number of customers this company takes on next year be set by what the sales team can win, or by how fast the order desk can type?
Implementation approach
A scope without ambiguity, before anything is signed.
We deliver
- A read of one month of your intake as it arrived: channels, customers, formats, amendment behaviour and existing exception categories
- The per-customer field map built from your own orders, including the free-text instructions that matter to a carrier
- Extraction models trained on your customers' documents, with Validation Station for the fields the model is unsure of
- The matching layer: customer, address book, lane, contracted rate, equipment and time-window feasibility
- Order creation and amendment against your TMS, the change record, the source archive, the Teams exception queue and the Power BI dashboard
We need from you
- Three months of incoming orders exactly as they arrived, with the TMS records they produced
- A process owner on the order desk and someone who can decide rate and address rules
- Technical accounts for the mailboxes, the file drops, the portals and the TMS test environment, plus your current rate tables
Stages
Discovery
One month of intake read end to end: channels, formats, amendment behaviour, exception categories
Design
Field maps, matching and rate rules, the amendment fingerprint, exception routing, security model
Build
Extraction models, robots, the TMS interface, the Teams queue and the dashboard
Validation
A parallel run on real orders, the desk comparing what the robot proposed with what it would have typed
Go-live
Two or three customers first, under supervision, then the rest by channel and branch
Optimisation
Model feedback, new customer formats, rule tuning, monthly review of exception reasons
Departmental. Effort follows the number of intake channels, how far your customers' formats differ, and what your TMS accepts on its inbound interface.
Ninety customers decide how an order reaches you. One desk absorbs all ninety.
Send us one month of incoming orders exactly as they arrived, mailbox export included, with the TMS records they turned into. You get back a channel-by-channel read of what could have been created without a person, and the exception categories your desk would keep.
Show us how ninety customers send ordersThe neighbouring process usually has the same problem
Your answer goes out Thursday; the load was booked Tuesday with whoever priced it first.
View solution Supply chainEvery proof of delivery on file before the invoice is dueFreight is delivered, invoiced and disputed before anyone finds the signed document.
View solution Finance & accountingFreight billed the day it is delivered, not the month afterOrders wait for the month-end billing run, and the accessorials written on a driver's note are never billed.
View solutionIndustries we deliver this in most oftenTransport & logistics