You will get Email-to-ERP Order Entry Agent
Project details
Customers email orders in all kinds of formats — PDFs, Excel, free text — and staff used to retype each one into the ERP. This agent reads the email, structures the order, validates every line, and creates the sales order in ERP automatically.
Monitor the mailbox for incoming order emails and attachments.
Extract content from PDFs, images, and Excel using Azure Document Intelligence.
Use OpenAI to structure the raw extraction into clean line items (product, quantity, unit, price).
Validate each line against master data — match the product, confirm pricing and units, check quantities.
Flag anything ambiguous (unknown product, price mismatch) for a quick human confirmation; auto-process the rest.
Create the sales order.
Send a confirmation summary and log every order for traceability.
Monitor the mailbox for incoming order emails and attachments.
Extract content from PDFs, images, and Excel using Azure Document Intelligence.
Use OpenAI to structure the raw extraction into clean line items (product, quantity, unit, price).
Validate each line against master data — match the product, confirm pricing and units, check quantities.
Flag anything ambiguous (unknown product, price mismatch) for a quick human confirmation; auto-process the rest.
Create the sales order.
Send a confirmation summary and log every order for traceability.
AI Algorithms
Large Language ModelAI Applications
AI Content Creation, AIOps, Text RecognitionAI Development Language
PythonAI Tools
Azure OpenAI, Microsoft 365 CopilotAI Models
GPT-4What's included
| Service Tiers |
Starter
$1,500
|
Standard
$4,000
|
Advanced
$8,000
|
|---|---|---|---|
| Delivery Time | 10 days | 20 days | 30 days |
AI Model Integration | |||
Batch Normalization | - | ||
Database Integration | |||
Detailed Code Comments | |||
Image Upscaling | - | - | - |
MLOps | - | - | - |
Model Deployment | |||
Model Documentation | - | - | - |
Model Monitoring | - | ||
Model Testing & Optimization | |||
Model Tuning | - | - | |
Natural Language Processing | - | - | - |
NLP Tokenization | - | - | - |
Pre-Training | - | - | - |
Prompt Engineering | |||
Setup File | |||
Source Code |
About Andrej
AI Agent Architect | MCP Servers, Tool Orchestration, Automation
Kumanovo, North Macedonia - 4:57 pm local time
I work on the seam between an LLM and a real system, which is where these projects
usually die. Right now I'm building a production MCP server that gives agents one
governed interface to enterprise data, with per-caller authentication, scoped
permissions and rate limits enforced at the server boundary. Before that I spent
fourteen months putting Dynamics 365 agents into live client
operations: order management, payflow, fraud detection.
What clients bring me:
Tool design and MCP servers. Tool granularity, naming, context budgeting,
deterministic error surfaces. When an agent picks the wrong tool or invents
arguments, that's almost always the cause, and almost never the prompt.
Agent orchestration. Routing requests to the right tool, chaining multi-step calls,
memory design, and fallback logic for when a tool fails at 3am.
Production hardening. Error handling, versioning, observability, evaluation. The
distance between a demo and something you can leave running.
RAG over your own data, where retrieval quality rather than the model is the actual
problem.
Voice agents and workflow automation. ElevenLabs, Twilio, n8n, Make. Same discipline,
different surface.
Underneath all of it: Python, Azure, PostgreSQL, Docker, and a data engineering
background, so when the real problem turns out to be your pipeline I don't hand it
back to you.
Worth saying plainly: I'm new to Upwork, so there's no review history here yet. The
work above happened at companies rather than on this platform, and I'd rather tell you
that than have you notice it.
Send me what's breaking and I'll tell you honestly whether I'm the right person for it.
Steps for completing your project
After purchasing the project, send requirements so Andrej can start the project.
Delivery time starts when Andrej receives requirements from you.
Andrej works on your project following the steps below.
Revisions may occur after the delivery date.
Kickoff call to map your current order flow end to end.
Send 15–30 real sample orders covering all formats.