Skip to content
LYKAN AI STUDIO
Contact

Document and approval workflows · Azure

Every figure traceable to its source document.

A workspace designed for procurement teams handling vendor agreements, purchase orders, invoices, and approvals. It connects document viewing, extracted information, search, and a recorded human decision.

A possible implementation uses Nuxt · FastAPI · Azure SQL. Illustrative example, not a client engagement.

Sample workspace · Procurement

Vendor agreement · source

1. Services are delivered per the attached schedule.

2. Payment is due within forty-five (45) days of invoice receipt.

3. Either party may terminate with thirty (30) days written notice.

Extracted fields · invoice

Vendor Matched

Who it serves

  • ▸Analysts reviewing invoices and agreements
  • ▸Approvers who need context before deciding
  • ▸Finance and legal teams who need an audit trail

The operational challenge

Invoices, purchase orders, and contracts live in different places. Checking one against another is manual, and the reasoning behind an approval is rarely recorded.

A guided product journey

  1. 01

    A document enters the queue

    The review queue shows assigned documents, due dates, status, and priority.

  2. 02

    Fields are extracted, one item flagged

    The system extracts key fields and identifies a potential mismatch for review.

  3. 03

    The analyst checks the source

    Source document beside extracted fields, with supporting contract language highlighted.

  4. 04

    Resolve or comment

    The analyst resolves the mismatch or returns it for correction with a comment.

  5. 05

    A visible approval flow

    The item moves through approval with decision history always in view.

Capabilities

Review queue

Assigned documents, due dates, status, priority

Document workspace

Source document beside extracted fields

Match review

Invoice, PO, and vendor information compared

Contract library

Search, versions, vendor grouping, key dates

Question panel

Answers linked to supporting passages

Approval flow

Comments, decisions, return for correction

Audit view

Who changed or approved what, and when

Applied AI

  • ▸Extract key fields from invoices and agreements.
  • ▸Identify potential mismatches for review.
  • ▸Search contracts by meaning as well as keyword.
  • ▸Answer a question from the available documents, with a link to the relevant passage.

The reviewer can always inspect the original document beside an AI-produced answer. The decision is theirs and is recorded.

How the system is designed

Experience

  • Nuxt frontend

Services

  • FastAPI services
  • Azure AI Document Intelligence
  • Azure AI Search + language model

Data

  • Azure SQL (workflow data)
  • Azure Blob Storage (source files)
  • ▸Answers are generated from retrieved passages, so each one carries a link to its source.
  • ▸Workflow state lives in a relational store, separate from the search index.
  • ▸Hosting and access controls are chosen during project scoping.

Where people stay in control

  • ▸Source always visible next to AI output
  • ▸Return-for-correction path on every approval
  • ▸Audit view of every change and decision

Reviewing documents at scale? Let's talk.

Share the problem at a level you are comfortable with. We can discuss sensitive details in a direct conversation.

Start a private conversation