Home · Solutions · Procurement
Solution · ProcurementThe last quote arrives, and the ranked comparison is in Teams within the hour
Supplier quotes compared within the hour
Supplier quotes are read in any layout, normalised into one comparison and ranked; the buyer decides in Microsoft Teams and the purchase order follows without retyping.
Executive summary
Stop paying buyers to retype prices, units and lead times from PDFs into comparison sheets.
We deliver the sourcing flow on the UiPath Platform and the Microsoft 365 tenant you already run.
The comparison is ready within the hour of the last quote or the deadline, so awards and orders leave the same day.
comparison workbook in Excel Online on SharePoint; SAP S/4HANA (purchase orders via BAPI); SharePoint evidence archive
Business problem
Sourcing
An engineering-to-order company buys differently from a serial manufacturer. Every machine carries project-specific parts: a welded frame from a drawing, machined shafts, a special motor with a long lead time. Policy asks for three quotes above a threshold, because competition on each order defends project margin and lead time decides whether the assembly hall waits.
The quotes come back in the supplier's own format: a PDF from the supplier's ERP, an Excel sheet with quantity breaks, a price typed into the email, a scanned letterhead. Each states its own currency, unit and payment terms, tooling quoted separately or not at all, and a lead time counted from order or from drawing approval. Before anything can be compared it must be brought to one basis, and today that basis is a personal Excel workbook filled by hand.
Buyers spend a third of their day transcribing instead of negotiating. Project engineers wait for an award before they can plan assembly. Nobody can tell from a system how many RFQs are open or why an award went where it did. When the order book grows, RFQs grow with it, buyers do not, and the three-quote rule is the first thing quietly relaxed.
How it works today
This is the flow we usually find in project purchasing before automation.
- PersonThe buyer receives a requisition from a project engineer, picks three to five suppliers from the approved list and from memory, and emails the RFQ with the drawings
- WaitingQuotes trickle back over one to two weeks as PDFs, Excel files, email text or a scanned letterhead; missing lead times and tooling costs are chased by phone
- PersonThe buyer opens each quote and retypes unit price, quantity break, currency, lead time, validity, Incoterms and payment terms into a comparison workbook
- Risk of errorQuotes in different currencies, units and delivery terms are compared on the visible price; freight, tooling and payment terms are corrected in the buyer's head, if at all
- WaitingThe comparison waits for the last supplier or the deadline while the project engineer asks in Teams whether the parts can be ordered
- PersonThe buyer emails the comparison to the requester, settles the choice at the Monday sourcing meeting and types the purchase order into SAP ME21N; the workbook stays in a personal folder
Why the current process costs more than it appears
The cost grows where nobody is looking.
- Retyping is the smallest part of the hour. Bringing four quotes to one basis, per piece, in euro, freight and tooling included, takes longer than reading them and is done from scratch for every RFQ.
- Chasing eats the day. Reminders to silent suppliers are sent when the buyer remembers, so decisions wait for the last quote instead of being taken on what has arrived.
- Awards drift towards the familiar supplier. A buyer with thirty open RFQs orders from whoever answered first in a readable format, and the competition the RFQ was meant to create never takes place.
- Late awards move delivery dates. A bracket ordered a week late is a machine assembled a week late, with expediting fees, overtime in the assembly hall and a customer conversation nobody planned.
Cost of inaction
Eight buyers absorb the RFQs each new contract adds, by relaxing the three-quote rule, so the "awaiting comparison" list stays at sixty to ninety and the money that competition was supposed to defend goes to the supplier who answered first, with no report able to show how much.
The quieter cost is portability. A buyer who leaves takes the workbook, the contacts and the memory of what was compared, and when an auditor asks for the evidence behind last quarter's awards, the answer is a week of searching rather than a register.
A plausible organisation with realistic proportions. The figures are there to be recalculated on your data; they are not a client result.
An engineering-to-order machinery maker with 700 employees, two plants, SAP S/4HANA and Microsoft 365 E3; eight buyers serve around forty project engineers.
260 RFQs a month with three to five quotes each, roughly 1,000 quote documents; about half PDFs from suppliers' systems, a third Excel files or email text, the rest scans.
RFQs go out from personal mailboxes; each buyer retypes every quote into a workbook per project, awards are settled at the Monday sourcing meeting and purchase orders typed into SAP afterwards.
About 55 minutes of comparison work per RFQ, spread over days while the last quote is awaited; a standing backlog of 60 to 90 RFQs "awaiting comparison".
RFQs are registered in a SharePoint list and sent by a robot; every quote is read by UiPath IXP Generative Extraction whatever its layout and normalised into one workbook per RFQ; silent suppliers are chased automatically; the buyer awards from a decision task in Microsoft Teams and the robot creates the purchase order in SAP.
In the modelled case the comparison is ready within the hour of the last quote or the deadline, buyers negotiate instead of transcribing, and every award has a dated comparison behind it. Illustrative figures, not a client result.
Proposed solution
We deliver the sourcing flow on the UiPath Platform and the Microsoft 365 tenant you already run. The RFQ register is a SharePoint list with line items, invited suppliers, deadline and the weighting of price, lead time and terms. A new item triggers a robot that emails the RFQ from the sourcing mailbox, one message per supplier with the RFQ number in the subject; replies return to the same mailbox, where the Microsoft Outlook 365 connector picks each one up, matches it to its RFQ and files the original on SharePoint.
UiPath IXP Generative Extraction reads the quote; this is the one place where a language model is used. The schema is defined once, from supplier and validity to unit price, quantity breaks, lead time and exclusions; the model reads free-form PDFs, Excel exports and letter-style quotes without a template per supplier and returns a confidence score for every value. Values below the threshold or outside plausibility rules agreed with purchasing become a Document Validation action in UiPath Action Center that the buyer settles in seconds. Everything after that is deterministic: currencies converted with finance's rate table, prices per piece, lead time in calendar days, tooling spread over the quantity, freight added per Incoterm, all written to one comparison workbook in Excel Online on SharePoint with identical columns for every supplier.
Suppliers who have not quoted get a reminder before the deadline; quotes with mandatory fields missing get a clarification email. When the last quote arrives or the deadline passes, the robot ranks the offers by the RFQ's weighting and creates the decision task: an Action Center task completed directly in Microsoft Teams, with the top three suppliers on normalised totals, lead times and terms, a link to the workbook, and the options award, negotiate, extend or cancel. On award, the robot records the decision and its reason, creates the purchase order in SAP S/4HANA and informs supplier and requester. The model reads documents; every rule stays with purchasing.
UiPath IXP Generative Extraction with confidence scores; UiPath Action Center Document Validation actions and app tasks as actionable notifications in Microsoft Teams; UiPath Orchestrator queues, triggers and audit log; UiPath Integration Service connectors for Microsoft Outlook 365, Microsoft OneDrive & SharePoint (SharePoint lists, Excel Online) and Microsoft Teams
The RFQ register, intake and matching, the extraction schema with threshold and plausibility rules, normalisation and the workbook template, reminders and clarifications, ranking rules, the decision task, the award record and reporting
Purchase order creation and vendor and material lookups in SAP S/4HANA through UiPath SAP activities (BAPI); other ERPs through their API or user interface
How the automated process works
- PersonThe buyer registers the RFQ in the SharePoint list: line items, invited suppliers, deadline and the weighting of price, lead time and terms
- AutomationA robot emails the RFQ with the drawings from the sourcing mailbox, one message per supplier with the RFQ number in the subject
- AutomationEach reply triggers intake: the quote is matched to its RFQ, filed on SharePoint and read by IXP Generative Extraction whatever the layout
- SystemRobots normalise currency, units, quantity breaks, lead time and Incoterms into the comparison workbook
- PersonValues below the confidence threshold or outside the plausibility rules go to a Document Validation action, where the buyer confirms or corrects them
- AutomationSilent suppliers receive a reminder before the deadline; quotes with mandatory fields missing receive a clarification email
- PersonWhen the last quote arrives or the deadline passes, the buyer gets the ranked comparison as a decision task in Microsoft Teams and awards, negotiates, extends or cancels; the requester sees the ranking in the project channel
- AutomationOn award, the robot writes the decision and its reason to the RFQ record, creates the purchase order in SAP and notifies supplier and requester
Human-in-the-loop model
Automation handles
- Sending RFQs, matching replies to RFQ numbers and filing every original quote
- Reading quotes in any layout and normalising prices, units, currencies, lead times and terms
- Reminders, clarification requests and the running count of quotes per RFQ
- Ranking by the agreed weighting, the award record and the purchase order in SAP
People decide
- Which suppliers are invited and which weighting applies
- Uncertain or implausible values, confirmed or corrected in a validation task
- The award itself, including negotiation rounds and departures from the ranking, with a recorded reason
- Rule changes: thresholds, weighting templates, currency and Incoterms tables stay with purchasing
Before and after
Systems and integrations
The stack is deliberately short: one engine, one execution layer, one place where a person decides.
Inputs
- sourcing shared mailbox in Outlook
- RFQ register (SharePoint list)
- requisition data and drawings on SharePoint
- finance exchange-rate table
Automation layer
- UiPath Orchestrator
- UiPath Robots
- UiPath IXP Generative Extraction
- UiPath Integration Service
- UiPath Action Center
Target systems
- comparison workbook in Excel Online on SharePoint
- SAP S/4HANA (purchase orders via BAPI)
- SharePoint evidence archive
Human touchpoints: decision task in Microsoft Teams; Document Validation action; project channel post in Teams; clarification emails to suppliers
Technologies used
reads quotes in any layout: header, terms, line prices, quantity breaks, lead times; confidence score per value
ADocument Validation actions for uncertain values; the award decision as an app task completed in Microsoft Teams
Aone queue item per quote, normalisation, reminders, ranking, retries, audit log
A"email received" trigger on the sourcing mailbox; RFQs, reminders and clarification requests
ARFQ register as a SharePoint list, comparison workbook in Excel Online (one template, one file per RFQ), filed quotes
Adecision task as an Action Center actionable notification; project channel post for the requester
Avendor and material checks; purchase order creation after award
AIllustrative economic model
Numbers you can check against your own data.
Fifty-five minutes per RFQ is what opening three to five quotes, retyping, converting units and currencies, chasing and formatting adds up to on the sourcing desks we see; the range is illustrative, not a client measurement. €32 is a fully loaded hourly cost for a buyer in Central Europe. We show buyer capacity released, not headcount removed, and leave out savings from better awards because they depend on your categories.
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
- The comparison is ready within the hour of the last quote or the deadline, so awards and orders leave the same day
- Buyers spend their hours on negotiation and supplier choice, not on retyping, unit conversion and chasing
- Every RFQ is compared on one basis, normalised price with freight and tooling, lead time in days and payment terms, so no quote wins because of a unit or a currency
- Reminders and clarifications go out on time for every RFQ, so more quotes arrive before the deadline
- Each award carries a dated comparison, the original quotes and the reason, ready for the auditor or a supplier query
The management view
- Sourcing throughput becomes measurable: open RFQs, quotes per RFQ and days from RFQ to award, per buyer and category
- Three-quote discipline is applied by the flow itself, and departures from the ranking are visible with their reasons
- Savings from competition can be evidenced: awarded price against the average and the highest quote, per category
- Growth of the order book is absorbed without adding buyers, and supplier terms sit in the register, not in personal workbooks
Board-level KPIs
Security and governance
Security is designed with the process, not after it.
- The robot reads and sends mail only from the sourcing mailbox: its Microsoft Graph permissions are limited to that mailbox with Exchange RBAC for Applications
- Quotes, register and workbooks stay in your Microsoft 365 tenant within the EU Data Boundary; extraction and orchestration run in the UiPath Automation Cloud EU region, and the language models behind the extraction are chosen and audited under your AI Trust Layer policies
- Robot credentials live in an Orchestrator credential store connected to Azure Key Vault; the SAP account can create purchase orders but not release them, so award and approval stay in separate hands
- Every quote, extracted value, validation, reminder and award is logged with timestamp and user, and access to the register and workbooks follows SharePoint permissions per purchasing group
Why now
Order books in machinery move in steps: when the next large contract lands, RFQ volume rises within a quarter, and the choice is between hiring buyers who retype quotes and a flow that reads them; the buyer capacity at stake in the modelled case is about €7,600 a month
Project controllers, auditors and owners ask more often for the evidence behind awards, and a register with a dated comparison per RFQ answers without a mailbox search
Generative extraction of unstructured documents is now a governed platform capability with confidence scores and a validation interface, up to 50 pages per document, so an unknown quote layout no longer needs a template
Relevant executive roles
Awards leave purchasing the day the last quote arrives, so sourcing stops being the reason a machine is assembled late
Every award is evidenced against its alternatives, savings from competition become measurable, and RFQ growth no longer turns into buyer headcount
Buyers negotiate instead of transcribing, the three-quote rule is applied every time, and supplier terms sit in the register rather than in personal workbooks
Common questions and objections
That is why we use generative extraction rather than templates: the schema describes what a quote contains, and the model reads whatever layout is in front of it. Values it is unsure about go to the buyer for a one-click confirmation, and the first weeks of validations tune the threshold.
The ranking is not produced by a model. It is the weighting purchasing defines, applied to normalised numbers, with every input visible in the workbook; buyers can depart from it with a recorded reason.
Email text is read like any other document, and a phone quote is typed by the buyer into the workbook, where it joins the same ranking. The point is one comparison basis per RFQ, not one input channel.
When this is not the right solution
- Fewer than roughly fifty RFQs a month, or items single-sourced by design, where the register alone, without extraction, is enough
- Sourcing already runs in a supplier portal or an e‑sourcing tool that returns structured bids; the comparison is then a report, not a document problem
- Line items cannot be identified across suppliers because there are no part numbers or drawings, so a person must interpret each quote anyway
A question for the next management meeting
Where would we look today to find the competing quotes behind last quarter's purchase orders, and the reason each award went where it did?
Implementation approach
A scope without ambiguity, before anything is signed.
We deliver
- A sample of your RFQs with the quotes they drew: layouts, currencies, units, missing fields, time from last quote to award
- The RFQ register in SharePoint with statuses, deadlines and weighting templates
- The extraction schema, confidence threshold and plausibility rules tuned on your suppliers' documents, normalisation logic and the workbook template
- Reminder and clarification emails, ranking rules, the decision task in Teams and the project channel post
- Purchase order creation in SAP, the award record, reporting and a runbook for purchasing
We need from you
- Three months of RFQs with the quotes received and the resulting purchase orders
- A process owner in purchasing for the weighting templates and rules
- The sourcing mailbox, a SharePoint site and SAP test and production accounts for the robot
- Your supplier contact list and the source of finance's exchange rates
Stages
Discovery
Sample of RFQs and quotes, layouts, rules, volumes, the buyers' current workbooks
Design
Register, extraction schema, normalisation and ranking rules, decision task, security model
Build
Robots, extraction project, workbook template, mailbox and Teams touchpoints, SAP posting
Validation
Replay of historical RFQs against the buyers' own comparisons, accuracy review, acceptance
Go-live
One commodity group or one buyer first under hypercare, then the whole team
Optimisation
Confidence thresholds, rule tuning, reporting, further categories or entities
Quick win. Effort depends on the variety of quote layouts and currencies, the number of line items per RFQ, and whether purchase orders are created through BAPI or the SAP user interface.
Buyers retype prices from PDFs into a comparison sheet, one quote at a time.
Send us twenty recent RFQs with the quotes received and the comparison sheets built from them. We return a short read-out of what the extractor reads unaided, what it would ask a buyer to confirm, and a first estimate of the hours released.
Compare twenty of your quotesThe neighbouring process usually has the same problem
Purchase requests get lost between mailboxes, approvers and buyers. Stop retyping approved requests into the ERP.
View solution ProcurementSupplier onboarding and due diligence in days, not weeksStop losing three weeks and the compliance evidence every time a plant needs a new supplier.
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 solutionIndustries we deliver this in most oftenManufacturing & industryServices & IT