Home · Solutions · Other solutions

Solution · Other solutions

Every statement matched against the ledger, so clerks only see what differs

Supplier statements reconciled the day they arrive

Each supplier statement is read whatever its layout, matched line by line against the SAP vendor ledger, and only the differences reach a clerk, with the evidence attached.

Quick winMicrosoft TeamsHuman in the loopAI where it earns its place
350supplier statements reach this illustrative hotel group every month, and roughly two thirds are filed without anyone opening them.

Executive summary

Challenge

Your suppliers tell you every month what they think you owe. Most of those letters go unread.

What changes

Every statement that reaches the mailbox is picked up and turned into lines.

Business value

Credits the supplier issued and you never booked become a task the day the statement arrives, instead of a discovery during next year's dispute.

Systems involved

SAP S/4HANA vendor accounts; the SharePoint reconciliation record; Power BI coverage report

Business problem

Accounts payable

A supplier statement is a free audit of your payables, and most of it is thrown away. Statements arrive by email as PDFs and spreadsheets in whatever layout the supplier's system produces, monthly, often covering the whole group at once. Reconciling one means exporting the vendor ledger for the period, lining up two lists that use different reference schemes, and working out why forty lines do not agree.

Teams do it for the largest suppliers, when a dispute forces the issue, and before year end because the auditors ask. They never do it for the long tail, which is exactly where credit notes go missing. The consequences are quiet: credits issued and never taken, invoices lost between a property and accounts payable that resurface as dunning letters, duplicates a statement would have exposed, and disputes that run for months because each side works from its own list.

The work survives untouched because it is tedious, because the benefit per statement is invisible, and because nobody is measured on it. A hundred small unknowns a month eventually become a payables balance that nobody in finance will defend without qualification.

How it works today

  1. PersonA supplier emails a monthly statement to the AP mailbox and it is filed in a folder
  2. PersonFor the largest suppliers a clerk exports the vendor ledger from SAP for the period
  3. PersonThe clerk aligns statement lines with ledger lines by document number, then by amount, by eye
  4. WaitingUnmatched lines are investigated one at a time across forty properties, over days
  5. PersonMissing invoices are requested from the supplier or the property; credits found are booked
  6. Risk of errorThe remaining statements stay unopened, and whatever they contain stays uncounted
PersonWaitingRisk of error

Why the current process costs more than it appears

Nobody planned this work; it accumulated.

  • Credits the supplier has already issued sit on their books until they expire there, and money nobody claims never appears in any report as a loss.
  • Invoices missing on your side become month-end accruals estimated by hand, then dunning letters, then a rushed reversal and an apology.
  • Lines that exist in your ledger but not on the statement are usually duplicates or misdirected payments, so the one control that would find them is the one that never runs.
  • Disputes consume the AP manager and the supplier's patience alike, and a linen or produce supplier that stops delivering turns a paperwork problem into one that guests notice.
  • Auditors reconcile the largest balances at their own rates and note the ones nobody reconciled, which is an invoice and a finding in the same letter.

Cost of inaction

Twelve months of statements that mostly stay in the folder≈ €103,000
Three years of coverage stuck at a third≈ €310,000
When new properties take the flow to 450 statements a month≈ €133,000

Nobody counts what was never claimed, which is why this survives every cost review. A credit note issued into a void does not generate a document on your side, does not appear in any ledger and eventually disappears when the supplier purges old balances. The same is true of the invoice that never arrived: it produces nothing at all until it produces a dunning letter.

Growth makes the arithmetic worse. Every property added brings its suppliers and their statements, while the team that opens a hundred a month stays the same size, so coverage drifts from a third towards a quarter without any decision being taken. Twelve months on, the payables balance is a little less certain, the goodwill a little thinner, and the cash left with suppliers a little larger.

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 hotel group with forty properties in Spain and Portugal, buying from about 4,000 suppliers: food and beverage, linen and laundry, energy, maintenance contractors and agencies. Accounts payable is centralised on SAP S/4HANA with each hotel as a profit centre, and invoices reach the team through the properties.

Volume

Around 350 supplier statements arrive each month as PDFs and spreadsheets. The team reconciles roughly a third of them, the largest suppliers and those already in dispute, at about three quarters of an hour each.

Current process

Export the ledger, align two lists with different references, chase what does not match. Credits discovered during disputes suggest that more are sitting on statements nobody has opened.

Bottleneck

The alignment itself. Two lists, two reference schemes, forty properties' worth of deliveries and one person matching by eye, which is why coverage stops at a third.

Solution

Every statement is read on arrival by UiPath Document Understanding whatever its layout, matched line by line against the SAP vendor ledger by number, amount, date and tolerance, and closed automatically when it agrees. Differences become a clerk task in Microsoft Teams with both documents side by side and a drafted note to the supplier.

Potential outcome

In the modelled case coverage goes from a third of statements to all of them while the hours spent fall, unbooked credits become a task the same day, and year-end reconciliations exist for every supplier rather than for the largest ten. The figures are a model, not a measurement.

Proposed solution

Every statement that reaches the mailbox is picked up and turned into lines. UiPath Document Understanding, using Generative Extraction for the layouts nobody has seen before, produces document type, number, date, amount, currency and reference for each row, and sends only the rows it is unsure about to a person. There is no template per supplier to build and no template per supplier to maintain when they change their system next spring.

The robot then pulls the vendor ledger for that supplier and period from SAP and matches the two lists: by document number, by amount and date, by tolerance, handling partial payments, prefixes and a payment run that writes references differently from the supplier's system. Every line ends in one of a small set of states: matched, on the statement only, in the ledger only, amount differs, credit note not booked, payment not applied. A statement whose lines all match closes itself with a record, which is most of the reason coverage can reach everything.

What is left is a queue of decisions rather than a queue of documents. A statement with differences creates a task in UiPath Action Center, completed in Microsoft Teams, showing both documents side by side, the classified differences and a drafted note listing exactly what is missing on each side. Credit notes on the statement that SAP does not have raise a request and a provisional line in the tracker. Coverage, open differences by type and age, and credits recovered are reported in Power BI, and the record for each supplier is kept in SharePoint as the evidence an auditor will ask for.

Native capabilities used

UiPath Document Understanding with Generative Extraction for varied layouts and Validation Station for uncertain lines; UiPath Action Center tasks with attachments, completed in Microsoft Teams; UiPath Orchestrator queues, triggers, credential store and audit log; SharePoint lists with versioning

What we build

The line-matching engine with reference normalisation, partial payments and tolerances; the classification of every difference type; the clerk task with side-by-side evidence; the supplier note templates; the coverage and recovery report

Custom integration

Statement pick-up and supplier notes through the Microsoft Outlook 365 connector; vendor line items, payments and credit-note postings through the SAP BAPI and OData connectors; the status list through the Microsoft OneDrive & SharePoint connector

How the automated process works

  1. AutomationA statement arriving in the AP mailbox is picked up, queued and identified by supplier and period
  2. AutomationDocument Understanding reads it into lines whatever the layout, with uncertain lines flagged rather than guessed
  3. PersonFlagged lines go to Validation Station for a short check before matching begins
  4. SystemThe robot pulls the vendor ledger for that supplier and period from SAP and matches by number, amount, date and tolerance
  5. AutomationFully matched statements close themselves with a reconciliation record filed in SharePoint
  6. PersonStatements with differences become an Action Center task in Teams: both documents side by side, differences classified, a note to the supplier drafted and ready to send
AutomationPersonSystem

Human-in-the-loop model

Automation handles

  • Picking statements from the mailbox and reading any layout into structured lines
  • Pulling the vendor ledger for the right supplier and period from SAP
  • Matching line by line and classifying every difference into a named type
  • Closing matched statements with a record and drafting the note that goes back to the supplier

People decide

  • What each difference means: chase it, book it, dispute it or accept it
  • Every credit note booking and every adjustment, because the reconciler never posts
  • Which recurring difference is worth a conversation with a property or a supplier
  • The tolerances, the difference categories and what "close enough" means for this ledger

Before and after

BeforeAfter
Statements reconciledabout a third, chosen by size or disputeall of them, on the day they arrive
Effort per reconciliation44 min of aligning two lists12 min on the ones that actually differ
Unbooked credit notesfound during a dispute, months laterraised as a task the same day
Evidence for the auditora spreadsheet for the largest balancesa record per supplier and period in SharePoint

Systems and integrations

Everything below runs on licences and systems you already hold, or would need anyway.

Inputs

  • supplier statements as PDFs and spreadsheets in the AP mailbox
  • the SAP vendor ledger for the supplier and period
  • property-level invoice records

Automation layer

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

Target systems

  • SAP S/4HANA vendor accounts
  • the SharePoint reconciliation record
  • Power BI coverage report

Human touchpoints: Validation Station for uncertain lines; difference tasks in Microsoft Teams; the supplier note sent from the AP mailbox

supplier statements as PDFsUiPath OrchestratorUiPath RobotsSAP S/4HANA vendor accountsValidation Station for uncertain lines

Technologies used

UiPath Document Understanding (IXP)

reads statements in any layout through Generative Extraction; Validation Station for uncertain lines

A
UiPath Robots + Orchestrator

pull the ledger, run the matching, hold the queue, keep secrets in the credential store and log every step

A
UiPath Action Center in Microsoft Teams

the difference task with both documents attached and the drafted supplier note

A
SAP S/4HANA (BAPI and OData connectors)

vendor line items, payments and credit-note postings

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

statement pick-up, supplier notes, the reconciliation status list

A
Power BI

coverage, open differences by type and age, credits recovered by month

A
Averified product capability (vendor documentation)

Illustrative economic model

A model, not a promise.

Illustrative model
350 statements a month × 8.4 minutes released each= 49 h / month
49 h × €30 fully loaded hourly cost= €1,470 / month
× 12 months= €17,640 / year
Annual capacity released (illustrative)≈ €17,640

Coverage is the variable that matters here, so the calculator prices the move from a third of statements to all of them rather than the work on any single one: today 105 statements are reconciled at 44 minutes each, which is 77 hours a month, and in the target state all 350 are read while 140 of them produce differences that cost a clerk 12 minutes, which is 28 hours. The 49 hours between them, spread across all 350 statements, are the 8.4 released minutes below. €30 an hour is an assumed fully loaded cost, and the credit pool is deliberately outside the calculator, because recovered cash and released minutes are two different arguments.

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

  • Credits the supplier issued and you never booked become a task the day the statement arrives, instead of a discovery during next year's dispute
  • Missing invoices are requested before the dunning letter, so accruals rest on the supplier's own list rather than on an estimate
  • Disputes shorten because the supplier receives one note that states exactly what differs on each side, rather than a request to send their list again
  • Duplicates and misdirected payments surface as "in ledger only" lines, which turns a routine document into a detective control at no extra cost
  • Year-end reconciliations already exist, for every supplier and with evidence attached, so the audit conversation starts from a file rather than a promise

The management view

  • Coverage becomes a number the AP manager can quote: statements received, reconciled, open differences by type and age, credits recovered by month
  • Recurring differences point at root causes rather than at people: a property that does not forward invoices, a supplier that bills the wrong entity, a payment run that references documents its own way
  • New team members are productive in days, because the queue holds decisions with the evidence attached rather than documents that must first be understood

Board-level KPIs

statement coveragecredits recovered per monthopen differences by agedays to resolve a differenceshare of supplier balances with a current reconciliation

Security and governance

The automation holds exactly the rights it needs, and not one more.

  • The reconciler and the poster are never the same actor: the robot reads the vendor ledger with a read-only SAP account, and every credit note or adjustment is booked by a clerk or on a clerk's recorded decision
  • Mailbox access is granted through Microsoft Graph application permissions scoped to the AP mailbox alone, and the robot's secrets are held in the Orchestrator credential store rather than in a workflow
  • Statements carry supplier bank details and sometimes personal names, so extraction runs under the UiPath AI Trust Layer with an allow-listed model and masking on any generative step
  • Reconciliation records are retained in SharePoint under Microsoft Purview labels as audit evidence for supplier balances, and processing stays in the EU region of UiPath Automation Cloud and within your Microsoft 365 tenant's EU Data Boundary

Why now

01

Statements were unreadable by machines for a simple reason: every supplier's layout differs and nobody would build four thousand templates. Extraction that reads an unfamiliar multi-page document without one removed that obstacle, and it now runs pre-trained

02

The other two pieces were always there. The vendor ledger has been in SAP all along, and mailbox triage can identify a statement on arrival rather than waiting for a clerk to notice it in a folder

03

Because all three pieces are ordinary, this is one of the few finance controls that can go from nothing to full coverage in weeks rather than quarters

Relevant executive roles

CFO

The payables balance is confirmed against every supplier's own view, and cash left behind with suppliers comes back rather than expiring on their books

AP Manager

Coverage moves from a third to all statements while the team works only the differences, with the evidence already attached

Head of Internal Audit

A reconciliation control that runs on every supplier replaces year-end sampling, and its record is the evidence

Common questions and objections

Our suppliers' statements are all different.

That is precisely why the design uses extraction rather than templates. The model reads the layout it is given, flags the lines it is unsure about, and does not need rebuilding when a supplier changes their system.

We already reconcile the suppliers that matter.

The credits and missing invoices sit in the long tail that nobody reconciles. The model here assumes only six statements in a hundred carry an unbooked credit, and that assumption alone outweighs the time saved.

Some suppliers do not send statements.

Most send one monthly already. For the rest the robot can request one, and every statement that arrives makes the balance a little more defensible than it was.

When this is not the right solution

  • A small supplier base where each statement is already reconciled by the clerk who knows the account and its history
  • Suppliers who do not issue statements and will not start, since there is nothing to read
  • A vendor ledger whose payment references are so inconsistent that matching would fail more often than it succeeds; the referencing has to be fixed first

A question for the next management meeting

Somewhere in last month's unopened statements is a credit this company has already earned and never claimed: who in finance would find it, and by what date?

Implementation approach

The first week looks the same at every client: we look at the data.

We deliver

  • A reconciliation of one month of your historical statements in discovery, with the credits and missing documents it finds and the accuracy per layout
  • The matching rules, tolerances and difference categories, agreed with the people who will work the queue
  • Document Understanding configuration for statements, plus the SAP ledger and Outlook integrations
  • The clerk task with side-by-side evidence, the supplier note templates and the SharePoint reconciliation record
  • The coverage and recovery report, training, a controlled start with the largest suppliers and run-phase support

We need from you

  • One month of statements with the matching SAP vendor line items, so the first reconciliation is on your own data
  • Access to the AP mailbox for a service account and read access to vendor line items in SAP
  • A named owner in accounts payable for tolerances, difference categories and the supplier note wording

Stages

Discovery

One month of statements reconciled on your data; findings and accuracy per layout reported back

Design

Matching rules, reference normalisation, tolerances, difference types, task and note templates

Build

Extraction, matching engine, SAP and Outlook integrations, the Teams task and the status report

Pilot and scale

The largest suppliers first with every difference reviewed, then all suppliers and the note automation

Quick win. Effort follows how consistently payment references are written on both sides, the spread of statement layouts, and how much tolerance the ledger needs before a difference is worth a person's time.

Two thirds of last month's statements were never opened.

Send us ten supplier statements from last month with the matching SAP vendor line items. We reconcile them and come back with the differences classified, the credits found and the accuracy per layout.

Reconcile ten statements with us

The neighbouring process usually has the same problem

Industries we deliver this in most oftenManufacturing & industryRetail & e‑commerceServices & ITShared services

Browse all 173 solutions