Home · Solutions · Supply chain

Solution · Supply chain

The office stops asking where the paperwork is; the chase runs itself and escalates by age

Every proof of delivery on file before the invoice is due

Consignment notes and delivery receipts are collected from every channel, matched to the load and chased by age; documents carrying a reservation go straight to the claims desk.

DepartmentalMicrosoft TeamsHuman in the loopAI where it earns its place
890consignment notes and delivery receipts a month have to be chased at this illustrative operator, and nobody knows which ones until an invoice is already late.

Executive summary

Challenge

Freight is delivered, invoiced and disputed before anyone finds the signed document.

What changes

We build a document register that knows, for every load, which documents the contract requires back, which have arrived, through which channel.

Business value

Invoicing stops waiting for a search: the document is attached to its load the day it arrives, and the billing run reports what is genuinely missing.

Systems involved

the load and document record in the TMS; SharePoint archive; Power BI

Business problem

Transport documents

A transport document is not administration. It is the evidence that the service was performed, and most of what the company earns rests on it: the customer's payment, the subcontractor's settlement, the defence of a damage claim. Until the signed note is on file, a load is either not invoiced or invoiced with nothing behind it.

The documents come home through channels with nothing in common. Own drivers photograph the note at unloading, into whatever application the fleet uses. Subcontractors email scans when they get round to it, several loads per message. Originals still travel by post where a contract requires them, and consignees publish some copies in a portal somebody has to log into. A person becomes the index: open the attachment, read the consignment number, search the TMS, rename, file, flag the order as documented.

When a document never arrives, nothing happens at all. That load sits in the state of one nobody has looked at yet, with no owner, no age and no alert. One clerk holds this together at a few hundred loads a month; at several thousand, mostly subcontracted, memory stops scaling. Reminders go to the carriers somebody remembers, remarks stay unread until they arrive as deductions, and nobody can say how many delivered loads lack proof.

How it works today

The channels differ from company to company; the shape below does not.

  1. PersonOwn drivers hand in or photograph the note; subcontractors email scans to a general address, several loads per message
  2. PersonA clerk opens each attachment, reads the consignment number and searches the TMS for the load
  3. SystemThe file is renamed by hand, filed or attached in the TMS, and the order is flagged as documented
  4. WaitingLoads whose documents never came wait without a status, indistinguishable from loads nobody has processed
  5. PersonChasing happens in bursts before billing, from a list rebuilt in Excel and sent to the carriers who come to mind
  6. Risk of errorRemarks and missing signatures surface only if somebody reads the scan closely; when a claim lands, the note, the photographs and the driver's statement sit in three places
PersonSystemWaitingRisk of error

Why the current process costs more than it appears

Behind every exception is an hour nobody logged.

  • Matching one document takes two minutes and nobody records it, so the cost never appears as a budget line; across thousands of loads it is the settlement office's largest single task.
  • Chasing in bursts is the most expensive way to chase: a reminder sent three weeks after unloading reaches a planner who now reconstructs the trip from memory.
  • Remarks on a note are the cheapest information a carrier will ever get about a damaged load, and reading them late turns a defensible position into a deduction argued after the fact.
  • Nobody can answer the completeness question: asked how many delivered loads have no document on file this morning, most forwarding offices answer with a search rather than a number.

Cost of inaction

Twelve months of matching and chasing by hand≈ €195,360
The same year with a quarter more subcontracted loads≈ €244,200
Three peak seasons at today's document rate≈ €586,200

Missing documents are not lost money until somebody decides they are, and the person who decides is usually the customer. The rows above price only the search: they say nothing about freight invoiced late, a subcontractor paid for a trip that was never evidenced, or a claim that becomes indefensible because the reservation was read too late.

Those windows are fixed. Article 30 of the CMR Convention requires reservations for apparent damage at the time of delivery, allows seven days (Sundays and public holidays excepted) for damage that is not apparent and twenty-one days for delay; Article 32(1) sets a limitation period of one year. A process that first reads the remark three weeks after unloading spends part of every year outside those limits.

Illustrative scenario

A plausible organisation with realistic proportions. The figures are there to be recalculated on your data; they are not a client result.

Organisation

A Central European forwarding and transport group: 180 vehicles of its own, some 240 regularly used subcontracted carriers, a TMS, Microsoft 365 E3, a settlement office of eleven and a claims desk of two.

Volume

7,400 loads a month, about 62% moved by subcontractors. Documents return through four channels: the driver application, a shared mailbox, post for the originals some customers still require, and a few consignee portals. About 12% of loads, roughly 890 documents a month, have to be chased.

Current process

Attachments are opened, read and matched by hand to the load in the TMS, renamed and filed. Chasing runs off an Excel list rebuilt before each billing run, and reservations are found by whoever looks at a scan closely enough.

Bottleneck

Around six minutes per load of matching and chasing, the chased tenth taking several times the average, plus a standing block of loads that cannot be invoiced because the paper is between a cab and a mailbox.

Solution

Robots collect from every channel, read the consignment number, the dates, the signatures present and any remark, and attach each document to its load. What is missing is chased on a schedule that escalates by age; what carries a reservation reaches the claims desk the same day.

Potential outcome

In the modelled case the document gap stops being discovered at billing and becomes a queue with an age on it, the chase rate falls as carriers learn the reminder is automatic, and the office works on settlement and customer queries instead of matching. It is arithmetic on the assumptions below, not a client result.

Proposed solution

We build a document register that knows, for every load, which documents the contract requires back, which have arrived, through which channel, and how old the gap is. Collection uses the channels you already have: robots take files from the shared mailbox, from the library where scanned post is dropped, from the driver application's interface or export, and from portals where a login is the only way in.

Reading is done by a document model configured in UiPath Document Understanding on your own notes and receipts, alongside pre-trained models for the bills of lading and packing lists that travel with them. It returns the consignment number, the order reference, the dates, packages and weight, whether the consignee's stamp and signature are present, and the text of any remark. Matching stays deterministic: two independent keys in agreement attach the document automatically, one key alone becomes a validation task.

The chase is a schedule rather than somebody's memory. Each morning a robot ages the open items against the delivery date and the deadline in the transport order, then sends one message per carrier listing that carrier's own loads. Escalation follows age: the carrier, then the dispatcher who booked it, as a card in Microsoft Teams, then the settlement lead before the billing run.

Native capabilities used

UiPath Document Understanding models and Validation Station; UiPath Orchestrator queues, time triggers, retries and audit; UiPath Integration Service connectors for Microsoft Outlook 365 and Microsoft OneDrive & SharePoint; UiPath Action Center tasks completed inside Microsoft Teams; Microsoft SharePoint libraries with Microsoft Purview retention; Power BI reporting

What we build

The collection robot per channel, the document model and its field set, the completeness rules per customer, the matching logic and its fallback keys, the ageing and escalation schedule, the claims routing, the archive and the reporting

Custom integration

The TMS, through whatever interface the product offers, most often a scheduled file exchange or an API; collection from portals where the signed document sits behind a login

How the automated process works

  1. AutomationDocuments are collected on a schedule from the shared mailbox, the scanned-post library on SharePoint, the driver application and the portals; files holding several loads are split
  2. AutomationUiPath Document Understanding classifies each document and extracts the consignment number, the order reference, the dates, packages and weight, stamp and signature presence, and any remark
  3. SystemA robot matches the document to the load in the TMS on two independent keys, writes the reference onto the order and files the image with customer, carrier and load metadata
  4. PersonDocuments matching on one key only, or with an unreadable scan, become a validation task completed from Microsoft Teams in seconds
  5. AutomationEvery morning open items are aged against the delivery date and the deadline; each carrier receives one message listing its own outstanding loads, and items past the first threshold escalate to the dispatcher in Teams and then to the settlement lead
  6. PersonAny document carrying a reservation, a short delivery or a missing signature goes the same day to the claims desk, with the delivery date and the notification windows on the task
AutomationSystemPerson

Human-in-the-loop model

Automation handles

  • Collection from every return channel, splitting, classification and field extraction
  • Matching to the load, filing in the archive and updating the document status on the order
  • The daily ageing, the per-carrier reminders and the escalation ladder
  • The evidence record: which document arrived when, through which channel, and what it said

People decide

  • Whether a remark on a note is a claim, and what is notified to whom
  • Whether a subcontractor's settlement is released or held while a document is outstanding
  • Whether an illegible scan is accepted, corrected or refused back to the carrier
  • The rules: what each customer requires back, the deadlines, the escalation thresholds

Before and after

BeforeAfter
Document on file after deliverydays to weeks, by channelsame or next day for the collected majority
Loads chased by namebefore billing, from Excelevery morning, by age, per carrier
Reservation noticedwhen the customer deductsthe day the document is read
Loads with no document at billingdiscovered during the runknown and aged before it starts

Systems and integrations

Where a rule suffices we do not use a model. Where judgement is needed, a person decides.

Inputs

  • driver application interface or export
  • Outlook shared mailbox
  • scanned post in a SharePoint library
  • consignee portals

Automation layer

  • UiPath Orchestrator
  • UiPath Robots
  • UiPath Document Understanding
  • UiPath Integration Service
  • UiPath Action Center

Target systems

  • the load and document record in the TMS
  • SharePoint archive
  • Power BI

Human touchpoints: Action Center validation in Microsoft Teams; carrier escalation cards in Teams; the claims desk task

driver application interfaceUiPath OrchestratorUiPath Robotsthe loadAction Center validation in Microsoft Teams

Technologies used

UiPath Document Understanding (IXP)

reads notes and receipts: numbers, dates, weights, stamp and signature presence, remark text

A
UiPath Robots + Orchestrator

collect from every channel, queue each load, run the daily ageing, retries and audit trail

A
UiPath Integration Service (Microsoft Outlook 365, Microsoft OneDrive & SharePoint connectors)

watches the mailbox and the scanned-post library

A
UiPath Action Center

validation of unreadable fields and reservation triage, as actionable notifications

A
Microsoft Teams

missing-document escalation to dispatchers, reservation alerts to the claims desk

A
Microsoft SharePoint

the document archive with customer, carrier and load metadata, under Microsoft Purview retention

A
Power BI

document status and chase rate by customer, carrier, channel and age

A
The TMS (transport management system)

source of the load, its required documents and its deadline; target for the reference

C
Averified product capability (vendor documentation)Cillustrative model — the figures on this page

Illustrative economic model

Start by questioning the assumptions.

Illustrative model
7,400 loads a month × 6 minutes of matching and chasing= 740 h / month
740 h × €22 fully loaded hourly cost= €16,280 / month
× 12 months≈ €195,360 / year
Annual capacity released (illustrative)≈ €195,360

Chasing is not a separate line in anyone's timesheet, which is why the six minutes below blends the quick match of a document that arrived cleanly with the third reminder on a load nobody can find. The unit priced is a load and not a document, because one load can owe the office two or three pieces of paper and the search is the same. €22 an hour is a fully loaded settlement-desk cost in Central Europe.

Run the numbers on your data

hours released per month
of annual capacity released

An illustrative estimate from your own inputs. It models released capacity; it is not a promise of savings.

Business benefits

  • Invoicing stops waiting for a search: the document is attached to its load the day it arrives, and the billing run reports what is genuinely missing
  • The chase becomes specific and early, one message naming a carrier's own loads two days after delivery instead of a general reminder three weeks later
  • Reservations reach the claims desk while the delivery is recent and the notification windows open
  • Subcontractor settlement runs against a document status both sides can see, which takes the argument out of the monthly run
  • A customer asking for proof of delivery is answered from the archive in seconds, without a call to a depot

The management view

  • Document completeness becomes a number with an owner: how many delivered loads have no proof today, by customer, carrier and age
  • Chase rates per channel and per carrier turn a complaint about carriers who never send documents into a fact for the next rate conversation
  • Claims exposure is visible early, because every reservation is registered the day it is read rather than the day it is deducted
  • The process survives holidays and staff changes, because the register lives in the flow, not in one clerk's memory

Board-level KPIs

days from delivery to document on fileshare of loads invoiced without a documentchase rate by carrierreservations registered inside the notification windowcost per load

Security and governance

Security is designed with the process, not after it.

  • Each robot signs in as itself: mailbox access scoped to that mailbox alone, a TMS account allowed to write a document reference and nothing else, portal logins held in the Orchestrator credential store rather than a shared password file
  • Documents, extracted fields and decisions stay inside your Microsoft 365 tenant, and the robots that read them run from the EU region of UiPath Automation Cloud
  • Retention labels through Microsoft Purview keep notes for the period your contracts and insurers require, and dispose of them on a rule
  • Every change to a rule, a deadline or an escalation threshold is versioned, approved by the process owner and recorded with its author

Why now

01

From 9 July 2027, under Regulation (EU) 2020/1056, Member State authorities must accept regulatory freight information shared electronically through certified eFTI platforms. That neither bans paper nor makes the electronic consignment note compulsory, but it settles the direction, and the operator whose documents are already indexed is the one able to use it.

02

Mandatory KSeF started on 1 February 2026, moving the freight invoice date away from the office's own timing: an invoice held back for a missing document is now visibly late rather than quietly late. Meanwhile the modelled €16,280 a month is spent every month this stays manual.

03

What used to be a project is now configuration: models trained on your own notes, connectors watching the mailbox and the library, human tasks completed inside Microsoft Teams.

Relevant executive roles

COO

A delivery is finished when its evidence is on file, and this makes that visible per load and per carrier rather than once a month at billing

CFO

Freight invoiced on time and disputes fought with the document in hand; the receivable stops depending on how fast somebody found a scan

Head of Claims

Reservations are registered the day they are read, so the notification windows are used rather than discovered later

Common questions and objections

Our subcontractors will never send documents on time.

They will not become punctual, but they do answer a message naming their own loads two days after delivery. What changes first is your side: the gap is known on day two, and the chase rate per carrier becomes a number for the next rate conversation.

The scans we get are photographs taken in a dark yard.

Some will always need a person, which is why validation is part of the design rather than an admission of failure. The model reads what is legible and flags what is not, so a person spends seconds on a field instead of minutes on a file.

Does this mean we can stop using paper consignment notes?

No, and nothing here depends on it. The flow collects whatever comes back, a photographed paper note or an electronic one, and treats both alike. Whether an electronic note works on a given lane depends on the states involved, a separate question from making today's documents arrive, match and get chased.

When this is not the right solution

  • Below a few hundred loads a month, where one person knows every carrier and every open document, a register costs more than it earns
  • The transport order does not state which documents must come back or by when; with no rule to chase against, the contracts come first
  • Loads carry nothing in the TMS that the document also carries, so no matching key exists and the order records need cleaning first

A question for the next management meeting

How many days after unloading does the signed document reach this company, and how much revenue is waiting on the loads where it never does?

Implementation approach

We start with one slice of the process and extend only once it is proven.

We deliver

  • A read of one month of your returned documents: the channel, the elapsed days, what was unreadable, what never came
  • The document model, its field set and its validation rules, configured on your own notes and receipts
  • The completeness rule table: what each customer and each contract requires back, and by when
  • The matching logic with its fallback keys, and the exception routing into Microsoft Teams
  • The ageing and escalation schedule, the claims routing, the archive with retention and a runbook

We need from you

  • Three months of loads with the documents that came back and the dates they arrived
  • A process owner in the settlement office and a named contact on the claims desk
  • Technical accounts for the mailbox, the TMS and the portals, with the access the robots need and no more
  • The contractual document requirements and return deadlines from your customer and carrier agreements

Stages

Discovery

One month of documents by channel, the chase rate, the exceptions, the contractual requirements

Design

The register, the completeness rules, the matching keys, the escalation ladder, the security model

Build

Collection robots, the document model, the TMS integration, the Teams touchpoints, the archive

Validation

A parallel run on historical loads, then live beside the current process until the register agrees

Go-live

One branch's subcontracted loads first, then the own fleet, with hypercare and chase tuning per carrier

Departmental. Effort is driven by the number of return channels, the quality of the scans that come back from the road, and how much of the document requirement exists as a rule rather than as habit.

Twelve loads in a hundred owe you a document nobody is chasing by name.

Send us one month of returned documents with the loads they belong to: the channel each arrived through, the date it arrived, and the loads that stayed empty. You get back a chase rate broken down by channel and carrier, and a written view of what could be collected without a person.

Break down your chase rate by carrier

The neighbouring process usually has the same problem

Industries we deliver this in most oftenTransport & logistics

Browse all 115 solutions