Home · Solutions · Supply chain
Solution · Supply chainOrders in SAP minutes after they arrive; the order desk handles only the exceptions
Customer orders from PDFs and emails straight into the ERP
Purchase orders that arrive as PDFs, spreadsheets and emails are read, checked against prices, stock and credit, created in SAP and confirmed to the customer without anyone retyping them.
Executive summary
Customer POs still get retyped from PDFs into SAP and confirmed hours later. Stop paying your order desk to type.
The flow we deliver runs on the UiPath Platform and on your existing Microsoft 365 tenant, and it leaves the EDI channel exactly as it is.
Clean orders are in SAP within minutes of arrival, so the afternoon cut-off stops deciding whether a customer is served today or tomorrow.
SAP S/4HANA (SD sales orders, customer-material records, pricing, availability, credit); Outlook (confirmations); SharePoint archive
Business problem
Order management
Every producer with a broad customer base runs two order channels. The largest retail chains send EDI orders that create themselves in SAP. Wholesalers, regional chains, HoReCa distributors and export customers send whatever their purchasing system prints: a PDF purchase order, an Excel order sheet or a few lines in an email, typed once by the customer and again by the order desk.
Retyping is the smaller part. Each order needs article numbers translated into material codes, quantities converted between pieces, cases and pallets, the price compared with what sales agreed, availability checked and credit blocks noticed, all before an afternoon cut-off after which the order ships a day later. The desk feels it as pressure, sales as price questions in Teams, logistics as late picking lists, finance as credit notes.
Peaks make it worse. Promotions and pre-holiday weeks multiply the volume exactly when temporary staff do not know the customers; a case entered as a piece becomes a short delivery, a fill-rate penalty and a second truck, and a confirmation sent hours later means the customer plans on assumptions.
How it works today
The pattern repeats across producers and distributors regardless of the ERP; the transaction names change, the cross-reference spreadsheet does not.
- PersonOrder-desk staff open the shared mailbox orders@ and sort PDFs, Excel sheets and email orders by region
- PersonThe clerk finds the sold-to and ship-to in SAP, translates article numbers with a cross-reference spreadsheet, converts units and types the lines into VA01
- WaitingLines priced differently from the SAP condition wait for the key account manager to answer in Teams; orders on credit hold wait for finance
- SystemSAP runs the availability check; when stock is short the clerk decides whether to shorten, split or substitute, often after calling the customer
- Risk of errorPallets typed as cases, wrong delivery dates and duplicates from a resent email surface at picking or at the customer's goods receipt
- PersonConfirmations go out one by one from Outlook; orders that arrive after the 14:00 cut-off are entered next morning and lose a day of lead time
Why the current process costs more than it appears
Behind every exception is an hour nobody logged.
- Nine minutes of typing per order is the visible part; the lookups around it, article numbers, units, delivery addresses and promotional prices, live in different places and are never timed.
- A wrong unit costs far more than the order it was typed on: a short delivery, a fill-rate penalty, a second truck, a credit note and a dispute over the mistake.
- Cut-off times turn minutes into days: an order that misses the picking wave by ten minutes ships a day later, and the customer's shelf absorbs the delay.
- Key account managers answer the desk's price questions several times a day, so negotiated prices live in email threads instead of SAP conditions.
- Customer knowledge is undocumented: it lives in private cross-reference files and with the two longest-serving clerks; in peak weeks temporary staff work without it, and the error rate rises with the volume.
Cost of inaction
The desk keeps up on most days, and that is the problem: the cost stays invisible, paid in overtime, in temporary staff each autumn and in a cut-off everybody has learned to work around. Growth arrives as new customers without EDI, so the manual share grows with the business.
What accumulates outside the desk's budget is harder to see: credit notes for entry errors, fill-rate penalties, a second truck and the hours sales spends answering price questions are all booked elsewhere. And the knowledge that keeps the error rate down, the cross-reference files and the two clerks who know which customer means pallets, is in no system and leaves with them.
A plausible organisation with realistic proportions. The figures are there to be recalculated on your data; they are not a client result.
An FMCG food and beverage producer with two plants and a central warehouse, 1,200 employees, SAP S/4HANA and Microsoft 365 E3; an order desk of six serves about 900 active customers. The eight largest retail chains order through EDI and stay outside this scenario.
4,000 non-EDI orders a month, about twelve lines each; roughly 55% PDF purchase orders, 25% Excel order sheets, 20% free-text emails. Volume rises by a third in promotional weeks.
Orders are saved from the shared mailbox, typed into VA01 with a cross-reference spreadsheet, checked for price and availability and confirmed by email once the SAP order exists; a 14:00 cut-off separates same-day picking from next-day.
About nine minutes per order, price questions that wait for sales, a pile of unentered orders before long weekends; confirmations go out the same afternoon at best.
Orders are captured from the mailbox, PDFs are read by UiPath Document Understanding and Excel sheets by customer template, codes are resolved from SAP master data, the order is simulated for price, availability and credit, then created in SAP and confirmed; unknown codes, price differences and shortages reach the desk as tasks in Teams.
In the modelled case clean orders are in SAP and confirmed within minutes, the desk types only the exception share (20 to 25% in the first months) and the practical cut-off moves later because entry no longer waits for a free person. Illustrative, not a client result.
Proposed solution
The flow we deliver runs on the UiPath Platform and on your existing Microsoft 365 tenant, and it leaves the EDI channel exactly as it is. The shared mailbox is watched through the UiPath Integration Service connector for Microsoft Outlook 365, and every incoming order becomes a queue item in UiPath Orchestrator. PDF purchase orders go to UiPath Document Understanding, whose pre-trained Purchase Orders model returns the header (PO number, dates, buyer, delivery address) and the lines (product codes, quantities, units, unit prices). Excel order sheets are read directly, one mapping per customer template. Free-text email orders become a pre-filled form for the desk with the customer already identified; we deliberately put no language model in this flow.
Robots then do what the clerk did by hand, from master data instead of memory. The customer is resolved from buyer name, VAT number, sender domain and delivery address; article numbers are translated through SAP customer-material info records and a cross-reference table the desk owns; units are converted with the material's conversion factors; a PO number seen before for the same customer stops a duplicate. The order is simulated in SAP before it is created, so price, availability and credit status are known while the document is still in the queue, and orders that pass every rule are created and confirmed from the shared mailbox with the SAP order number, confirmed quantities and dates.
Everything that fails a rule becomes a task for a person in Microsoft Teams, with the PDF, the extracted lines and the SAP proposal side by side: unknown article numbers, prices outside tolerance and shortages are decided by the order desk in UiPath Action Center, low-confidence fields open in Validation Station, and credit-blocked orders follow the release flow credit control already uses. Orchestrator keeps the queue, the retries and the log of every order from email to confirmation.
UiPath Document Understanding pre-trained Purchase Orders model and Validation Station; UiPath Orchestrator queues, triggers and audit; UiPath Integration Service connectors for Microsoft Outlook 365 and Microsoft Teams; UiPath Action Center tasks with actionable notifications in Microsoft Teams
The intake workflow, customer and material identification rules (cross-reference, units, duplicates), the simulate-then-create logic with price and availability tolerances, exception routing, confirmation templates, the end-of-day summary and the desk runbook
SAP S/4HANA sales-order simulation and creation plus customer, material, price, availability and credit lookups through UiPath SAP activities (BAPI); per-customer Excel order-sheet templates
How the automated process works
- AutomationEach new email in orders@ is picked up on arrival: attachments are separated, the sender is matched to a customer and every order becomes a queue item in Orchestrator
- AutomationDocument Understanding reads the PDF: PO number, dates, delivery address and every line with product code, quantity, unit and price; Excel sheets are read through the customer's template
- SystemRobots map article numbers to SAP materials and units, check the PO number for duplicates and simulate the order in SAP for price, availability and credit
- AutomationClean orders are created in SAP and the confirmation with order number, quantities and delivery dates leaves the shared mailbox within minutes
- PersonUnknown codes, prices beyond tolerance, shortages and free-text email orders reach the desk as Action Center tasks in Teams; low-confidence fields open in Validation Station
- AutomationCredit-blocked orders go to the existing release flow; the desk's Teams channel receives an end-of-day summary: entered, confirmed, waiting and why
Human-in-the-loop model
Automation handles
- Capture and extraction of every order, whatever the format
- Customer, material and unit resolution, duplicate detection and the SAP simulation
- Creation of clean orders in SAP, the confirmation to the customer and the archive
- The queue, reminders on open tasks and the end-of-day summary
People decide
- Unknown or ambiguous article numbers, and whether the new mapping is kept for next time
- Prices outside tolerance, together with the key account manager who agreed them
- Shortages: shorten, split, substitute or postpone, following the customer's known preferences
- Credit-blocked orders, which stay with credit control, and every change to the rules
Before and after
Systems and integrations
Everything below runs on licences and systems you already hold, or would need anyway.
Inputs
- Outlook shared mailbox (PDF purchase orders, Excel order sheets, email orders)
- customer cross-reference table on SharePoint
Automation layer
- UiPath Orchestrator
- UiPath Robots
- UiPath Document Understanding
- UiPath Integration Service
- UiPath Action Center
Target systems
- SAP S/4HANA (SD sales orders, customer-material records, pricing, availability, credit)
- Outlook (confirmations)
- SharePoint archive
Human touchpoints: Action Center tasks in Teams; Validation Station; order-desk Teams channel with the end-of-day summary
Technologies used
pre-trained Purchase Orders model reads header and line data from PDF orders; Validation Station for low-confidence fields
Aqueue every order, apply the mapping rules, simulate and create orders in SAP, retry, log and audit
Awatches the shared mailbox, fetches attachments, sends confirmations and the channel summary
Aexception tasks for the order desk, completed without leaving Teams
Athe order desk's channel for tasks and the daily summary
Acustomer, material, price, availability and credit lookups; sales-order simulation and creation
AIllustrative economic model
A model, not a promise.
No client stopwatch produced these numbers; they are ranges we meet on order desks. Nine minutes per order averages simple repeat orders with promotional orders that need lookups, a price question and a manual confirmation; €24 is a fully loaded hourly cost for an order-desk role in Central Europe. The result is capacity the desk gets back, not a headcount figure.
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
- Clean orders are in SAP within minutes of arrival, so the afternoon cut-off stops deciding whether a customer is served today or tomorrow
- Confirmations with real quantities and dates reach customers within the hour, and the calls asking whether the order arrived stop
- Unit and quantity mistakes are caught at intake against the customer-material record, not at picking or at the customer's dock
- Price differences are visible before delivery, not as deductions on the remittance advice weeks later
- Promotional peaks are absorbed by robots, and the desk's time goes to shortages, new listings and customers with a problem
The management view
- Every non-EDI order is traceable from the email to the confirmation, with the rule it passed or the person who decided it
- Cut-off compliance and time-to-confirmation become measured figures per customer group, not impressions
- Negotiated prices are either in SAP conditions or flagged as exceptions; verbal agreements stop entering orders silently
- New customers and higher volumes are a matter of adding mappings, not people
Board-level KPIs
Security and governance
Control is not an add-on.
- Robots use a dedicated SAP user that can display master data and simulate and create sales orders; it cannot change prices, conditions or credit limits
- Access to the shared mailbox goes through Microsoft Graph with application permissions limited to that one mailbox; secrets live in the Orchestrator credential store, never in workflows
- Order PDFs and the fields read from them are processed in UiPath Automation Cloud's EU region and stored in your Microsoft 365 tenant, so nothing leaves the EU; the extraction is a document model, not a generative one, so no order content goes to a language model
- A price outside tolerance is never corrected by a robot; it becomes a task with a named decision-maker, and every decision is logged with its evidence
- Business-contact data in orders (names, phone numbers, delivery addresses) goes to SAP and the archive only, never to logs; the archive follows your retention rules
Why now
Retailers keep shortening lead times and tightening fill-rate penalties, so a day lost at the cut-off now carries a price, not only a service-report entry
The modelled €14,400 a month of desk capacity goes into retyping while seasonal staff who know your customers are harder to find each year
A pre-trained Purchase Orders model, the Outlook 365 connector and Action Center tasks in Teams are standard components; the build is the rules and the SAP logic, not the document engine
Relevant executive roles
Service level stops depending on whether the desk kept up with the mailbox before the picking wave
Peaks and new customers no longer mean overtime or temps, and the desk's work shifts from typing to solving
Negotiated prices live in SAP or surface as exceptions, and key account managers stop being the desk's help line
Credit notes for entry errors and fill-rate penalties become visible and fall; credit blocks are handled in the flow, not around it
Common questions and objections
The long tail is where the desk's time goes: many small customers, many formats, no standard. EDI onboarding for a few hundred wholesalers is neither realistic nor wanted; a document model reads what they already send, and EDI is not touched.
Nothing is guessed. A line is created only when the article number resolves through the customer-material record or the cross-reference the desk maintains; otherwise it becomes a task, and the mapping the clerk chooses is kept for next time.
They reach the desk as a pre-filled form with the customer identified, and the SAP steps run once the lines are confirmed. If free-text orders turn out to be a large share, a second phase can add UiPath Communications Mining; we do not start there.
When this is not the right solution
- Non-EDI volume below a few hundred orders a month: a customer portal or a disciplined mailbox routine costs less than automation
- Customer-material records and pricing conditions in SAP are not maintained, so most orders would become exceptions; the master data process comes first
- Orders are mostly negotiated by phone and priced individually, so there is no document to read and no rule to apply until the commercial process is standardised
A question for the next management meeting
If our order desk stopped retyping orders that customers have already typed into their own systems, how many hours a month would come back, and what is a day lost at the cut-off worth in service level?
Implementation approach
The first week looks the same at every client: we look at the data.
We deliver
- An order-mix profile from one month of your order traffic: channels, customer templates, cross-reference quality, exception categories and cut-off behaviour
- Configuration of the Document Understanding Purchase Orders model on your customers' documents, with Validation Station feedback
- The rule layer: customer identification, article and unit mapping, duplicate detection, price and availability tolerances
- SAP simulation and order creation, confirmation templates, Excel order-sheet mappings and the exception flows in Action Center and Teams
- A pilot on one customer group, followed by rollout in waves, hypercare and a runbook the desk keeps
We need from you
- Order history for three months, with the SAP sales orders and confirmations that came out of it
- A process owner on the order desk and a sales contact for the price tolerances
- Service accounts for SAP (test and production) and for the shared mailbox
- The current state of customer-material info records and the desk's cross-reference files
Stages
Discovery
Sample analysis, channel split, mapping quality, exception categories, cut-off rules
Design
Target flow, tolerances, customer identification rules, confirmation content, security model
Build
Document model, robots, SAP integration, Excel templates, Teams and Action Center touchpoints
Validation
Parallel run on real orders, exception handling, acceptance by the order desk and sales
Go-live
Controlled start per customer group with supervision and hypercare
Optimisation
Monitoring, mapping growth, tolerance tuning, further customer groups
Departmental. Effort depends on the number of customer templates, the state of customer-material records, the variety of pricing agreements and how many rules the desk applies from memory.
Where does your order desk lose its afternoons?
Send us one month of non-EDI order volumes by channel and customer group, plus five typical purchase orders. We return a one-page automatable-share estimate and an exception taxonomy for your desk.
Total the retyping on your order deskThe neighbouring process usually has the same problem
Every third customer contact asks where an order is; the answer already sits in SAP and the carrier portal.
View solution Supply chainOrder confirmations and delivery dates chased automaticallyBuyers spend their mornings chasing suppliers for dates, and planners plan on dates that are already wrong.
View solution Supply chainProduct and price master data published to every channelStop retyping every new product and price change into the ERP, the PIM, the web shop and each marketplace.
View solution Supply chainExport documents ready while the truck is loadedYour ERP holds every figure the customs pack needs. Somebody types them in again anyway.
View solution Case studyFaster release of blocked ordersValid orders move faster — so the company earns faster.
View case study Case studyOrder-to-Cash control towerEvery blocked order, disputed invoice and breached credit limit flows into one process: an agent diagnoses, a robot executes, a human approves with one decision on Teams.
View case studyIndustries we deliver this in most oftenManufacturing & industryRetail & e‑commerce