Skip to content
Cloud Mechanics
Procurement Agent iconAI Agent

Procurement Agent

Procurement Agent that automates procurement workflows using your data and tools.

The challenge

Requisitions arrive as free text and get routed by whoever recognises the category. Buyers rekey quotes into a comparison spreadsheet, off-contract spend leaks through unnoticed, and proving why one supplier won means reconstructing the decision from an email thread months later.

The outcome

A Microsoft Foundry agent classifies the requisition against contracts held in Azure AI Search, Azure AI Document Intelligence reads the quotes, and Azure Functions normalise them for a like-for-like comparison. Catalogue and framework orders raise automatically; off-contract spend reaches a buyer with the risk assessed.

At a glance

Type
ai agents
Category
business

Next step

Move from solution to engagement.

Build This Solution

01 — Architecture

End-to-end architecture

A requisition enters, is classified to a category and contract, and routed down the right sourcing path. Quotes are read by Document Intelligence and normalised for a like-for-like comparison, supplier risk and sanctions are screened, and a named buyer awards anything off-contract or over threshold.

REQUESTERRequester / Buyerneed & approvalINTAKEDynamics 365 /ERPrequisitions & POsAzure Logic Appsintake & routingSOURCINGAzure APIManagementsupplier APIsAI DocumentIntelligencequotes & contractsAI & AGENTMicrosoft FoundryAgentsource & compareAzure OpenAImodelssummaries & termsAzure AI Searchcatalogue & policyAzure AI ContentSafetyprobity guardrailsDATA & RECORDSAzure SQLspend & suppliersAzure Cosmos DBrequest stateAzure Key VaultERP credentialsAPPROVAL & INSIGHTBuyer approvalaward & exceptionsApproval workflowdelegation matrixPower BI / Fabricsavings & complianceDEVOPS & DELIVERYGitHubsource control & CIDockercontainer buildContainer Registryversioned imagesAgent releasedeploy with rollback
Figure 1 — End-to-end reference architecture for a source-to-order agent on the Microsoft stack.
  • Requester: Requester / Buyer
  • Intake: Dynamics 365 / ERP, Azure Logic Apps
  • Sourcing: Azure API Management, AI Document Intelligence
  • AI & agent: Microsoft Foundry Agent, Azure OpenAI models, Azure AI Search, Azure AI Content Safety
  • Data & records: Azure SQL, Azure Cosmos DB, Azure Key Vault
  • Approval & insight: Buyer approval, Approval workflow, Power BI / Fabric

02 — Workflow

Process & decision workflow

How a need becomes a compliant purchase order — classify, source, compare and check, then branch. Catalogue and framework purchases raise automatically against contract, while off-contract or over-threshold spend goes to a buyer with the options, the risk and the rationale.

1RequestNeed, budget, category and urgencycaptured2ClassifyMapped to a category, contract and costcentre3SourceCatalogue, framework or quote routeselected4CompareQuotes normalised on price, terms andlead time5CheckSupplier risk, sanctions and policyvalidated6AwardPurchase order raised against the rightcontractOn contract& in budget?Path 1 · catalogue or frameworkAuto-raise the orderPurchase order issued against contractPath 2 · off-contract or over thresholdBuyer approvalOptions, risk and rationale presentedOrder recordedPrice, terms and approverretainedSavings capturedBenefit and compliance reportedCloseorder placed
Figure 2 — Request → classify → source → compare → check → contract branch → auto-raise or buyer approval.
  1. Request: Need, budget, category and urgency captured
  2. Classify: Mapped to a category, contract and cost centre
  3. Source: Catalogue, framework or quote route selected
  4. Compare: Quotes normalised on price, terms and lead time
  5. Check: Supplier risk, sanctions and policy validated
  6. Award: Purchase order raised against the right contract
  7. Path 1 · catalogue or frameworkAuto-raise the order: Purchase order issued against contract
  8. Path 2 · off-contract or over thresholdBuyer approval: Options, risk and rationale presented

03 — Components

Key Microsoft components

Procurement decisions have to be defensible — sourcing, comparison, screening and award all stay on the Microsoft stack.

  • Dynamics 365 / ERP iconDynamics 365 / ERPRequisitions, purchase orders and the system of record.
  • Azure Logic Apps iconAzure Logic AppsIntake, approval routing and supplier notifications.
  • Azure API Management iconAzure API ManagementSecure gateway for supplier, catalogue and ERP APIs.
  • Azure AI Document Intelligence iconAzure AI Document IntelligenceReads quotes, contracts and supplier documentation.
  • Microsoft Foundry Agent Service iconMicrosoft Foundry Agent ServiceClassification, sourcing and comparison orchestration.
  • Azure OpenAI models iconAzure OpenAI modelsTerms summarisation and like-for-like comparison notes.
  • Azure AI Search iconAzure AI SearchRetrieval over catalogues, frameworks, contracts and policy.
  • Azure AI Content Safety iconAzure AI Content SafetyProbity guardrails on supplier communication and criteria.
  • Azure Functions iconAzure FunctionsNormalisation, threshold rules and savings calculation.
  • Azure SQL iconAzure SQLSpend, supplier master, awards and approval records.
  • Azure Cosmos DB iconAzure Cosmos DBRequest state, sourcing events and comparison history.
  • Azure Key Vault iconAzure Key VaultERP and supplier portal credentials and secrets.
  • Azure Blob Storage iconAzure Blob StorageQuotes, contracts and award evidence packs.
  • Power BI / Microsoft Fabric iconPower BI / Microsoft FabricSpend, savings, compliance and supplier dashboards.

04 — AI

What the agent consumes

The capabilities the agent applies to every requisition, and the line it does not cross.

AI capabilities embedded in the agent

  • Requisition classification
  • Catalogue matching
  • Quote extraction
  • Like-for-like comparison
  • Supplier risk screening
  • Sanction and exclusion checks
  • Retrieval-augmented generation
  • Terms summarisation
  • Delegation routing
  • Maverick-spend detection
  • Savings attribution
  • Workflow orchestration

AI responsibility boundaries

The agent sources, compares and prepares; a named buyer awards anything off-contract, over threshold or single-sourced. It never selects a supplier on unpublished criteria, changes bank details or bypasses the delegation matrix. Every award records the options considered, the criteria applied and the approver.

05 — Personalization

Personalization & evolving process

The same methodology applies to every agent in the catalog. Tune the categories, the sourcing template, the thresholds and the value model — the page structure stays identical.

Category & supplier profile

Define the spend categories, frameworks, preferred suppliers, currencies and regulatory obligations in scope. The personas are the requester, the category buyer, finance and the compliance officer.

06 — Impact

Key outcomes & business impact

Starting targets for the value case — validate each one against the customer baseline during discovery.

  • On-contract spend85%+Requisitions routed to the right agreement automatically.
  • Cycle time−50%Catalogue and framework orders raise without a queue.
  • Realised savingsTrackedBenefit measured against the baseline, not the quote.
  • Award traceability100%Every award stores its options, criteria and approver.

Illustrative improvement index

Manual baseline = 100. Illustrative targets, not a commitment — confirm against the customer baseline.

10050Requisition cycle time10040Off-contract spend10045Sourcing hoursManual baselineAI-assisted target
  • Requisition cycle time: manual baseline 100, AI-assisted target 50
  • Off-contract spend: manual baseline 100, AI-assisted target 40
  • Sourcing hours: manual baseline 100, AI-assisted target 45

Related & recommended

Derived automatically from our solution knowledge graph.

Related AI agents

Other agents that pair well with this one.

Related professional services

How we design, build and secure it.

Related quick wins

Ready-made Azure AI to start fast.

Technologies

What powers this solution.

Related managed services

Keep it running and optimised.

Ready to move from challenge to solution?

Talk to a Cloud Mechanics expert or build your solution in minutes.