Case Study · 2026 Live in production
Client
Redacted
Client identity under NDA · sector & metrics disclosed with permission
ClientNorth American Manufacturing Group
SectorIndustrial Manufacturing · Multi-Entity
End usersAP, procurement, and finance teams across entities
Built byEnvyro · 2026
Status● Live · 24 / 7

A confidential case study · AP automation · ERP-native

From 6,000 invoices a month to one exception queue.

A back-office AI agent that ingests vendor invoices from mailboxes and EDI feeds, extracts and normalizes line items, three-way matches against POs and goods receipts, and posts straight into the ERP — escalating only true exceptions.

AP Automation Multi-Entity Three-Way Match ERP Write-Back Straight-Through Under NDA
0 / mo
Invoices processed across entities, end to end
0s
Avg. invoice-to-posting time
0%
Straight-through processed — no human touch
~3 FTE
Of AP work returned to the business each month
01 · The Challenge

Every supplier sent a different shape. AP normalized them all by hand.

Six thousand invoices a month, across four+ legal entities, from hundreds of suppliers — each in a different format, each requiring its own coding logic, each tying up to a different PO and goods receipt.

01 / 04

Every supplier sent a different format

PDF, scanned image, email body, EDI. AP normalized them by hand — re-orienting to whichever layout the supplier happened to use this month.

02 / 04

Multi-entity coding

The same vendor invoiced into four+ legal entities, each with its own GL, cost-center, and project logic. Per-invoice, per-entity, by hand.

03 / 04

Three-way match was the real bottleneck

Not the keying — the matching. POs, goods receipts, invoices, across three systems and three tabs. Reconciled by eye, every time.

04 / 04

Month-end close held hostage by the AP backlog

Close slipped a week, every month, because of the queue. Finance leadership knew the date but never trusted it.

02 · The Solution

One pipeline. Every format, every entity.

Envyro partnered with the group to deploy a multi-entity AP agent that ingests, normalizes, matches, and posts every invoice into the ERP — leaving the AP team only with the genuine exceptions.

Format doesn't matter — PDF, scan, email body, or EDI all flow through one pipeline. Coding is supplier- and entity-aware. Three-way matching happens automatically. Anything that breaks the match lands in a queue with the exact reason attached.

Built by Envyro · Live across every entity in the group.

Multi-format ingestion

IMAP mailbox + EDI feeds + OCR for scans. One pipeline takes them all — no per-supplier setup, no manual triage.

Line-item-level extraction

Supplier-aware normalization down to the SKU. Quantities, units, prices, taxes — all extracted as clean structured output.

Automated three-way match

Invoice ↔ PO ↔ goods receipt, fully automated. Tolerance rules per entity. Break reasons captured the moment they happen.

ERP write-back with coding

Entity, GL, cost center, and project resolved automatically at runtime — invoices land in the ERP coded complete.

"
Close used to be the week we lost. Now AP looks at exceptions, not invoices — and close lands on the same day every month.
MF
North American Manufacturing Group
Finance leadership · live with Envyro-built platform
03 · Deployment

One pipeline. Every entity.

A single pipeline serving every entity in the group — per-entity ERP credentials, coding rules, and GL mappings resolved automatically at runtime. No per-supplier setup.

1 pipeline
Every entity, every supplier
0
Per-supplier setup
24 / 7
Always-on
1
Step 01

Ingestion + entities

Mailboxes and EDI feeds connected. Per-entity ERP credentials and coding rules vaulted and resolved at runtime.

2
Step 02

Extract + match

Line-item extraction tuned against real supplier samples. Three-way match engine wired to POs and goods receipts with per-entity tolerances.

3
Step 03

Live + observed

Production traffic flowing. Every invoice logged, every match-break captured, every edge case fed back into the model.

04 · Production Data

Where every invoice actually lands.

A representative month — roughly 6,000 invoices, the vast majority straight-through processed and posted clean. Exceptions surface with the original document, partial match, and exact break reason attached.

0
Invoices / month
41 sec
Avg processing
4 outcomes
Routing classes
Live
In production
Where every invoice lands
By volume
Representative month · ~6,000 invoices · outcome distribution
~96% straight-through · ~4% surfaced with exact break reason No silent drops · every invoice logged
05 · The Validation Gate

Ninety-six in a hundred posted clean. The rest get a warm handoff.

Nothing is silently dropped. Nothing is hallucinated into the ERP. Every invoice either matches — or it asks. The exception queue carries the original document, the partial match, and the exact break reason.

Auto-posted
0%

Matched, coded, and posted automatically

Validated against PO + goods receipt, coded per entity, written to the ERP and archived. AP never had to open these.

Reviewed
0%

Surfaced to AP with full context

Original document, partial match results, and the exact match-break reason — AP picks up exactly where the pipeline stopped.

Nothing is silently dropped. Nothing is hallucinated into the ERP. Every invoice either matches — or it asks. That single rule is what makes the pipeline safe to run at production volume.

06 · How It Works

From invoice arriving to posted clean.

A single pipeline carries every invoice through five stages — ingestion, extraction, normalization, matching, and write-back — in under a minute, regardless of format or entity.

~41 sec
End-to-end processing per invoice, including OCR and ERP write-back.
Zero manual entry
Matched invoices land in the ERP coded complete, without anyone touching them.
Full audit trail
Every invoice archived with a structured filename. Every action logged and attributable.
Step 01
Invoice arrives

Mailbox or EDI feed. Any format. Pipeline picks it up the moment it lands — no polling lag.

Step 02
Format extracted

OCR for scans, layout-aware parsing for PDFs, structured pull for EDI. Format normalized before extraction begins.

Step 03
Line items normalized

Supplier-aware mapping to internal SKUs and GL codes. The right entity, GL, and cost center resolved per line.

Step 04
Three-way matched

Invoice matched against PO and goods receipt, with per-entity tolerances. Break reasons captured if not.

Step 05
Posted & archived

Written to the right ERP entity with full coding. Document archived with structured filename and audit entry.

07 · Before / After

The same invoice — at a tenth of the time.

What a single invoice used to mean for AP, versus what it means now. The shape of the job is the same; only the exceptions reach a person.

Before · per invoice
8 – 12 min
  • Open the email, save the attachment somewhere
  • Re-read whichever supplier's layout it happens to be
  • Hand-key header and line items into the ERP
  • Open the PO and goods receipt in another tab
  • Three-way match by eye, fix what doesn't tie
  • Code the right entity, GL, and cost center
After · per invoice
< 1 min
  • Pipeline picks up email or EDI instantly
  • Format normalized: PDF, scan, body, or EDI
  • Line items extracted and mapped supplier-aware
  • Three-way matched against PO + goods receipt
  • Posted to the right entity with full coding
  • Only true exceptions reach AP, with the reason attached
08 · The Impact

The back office runs the same — just faster, longer, and on the record.

AP hours come back. Close lands on time. Duplicates die at ingestion. And every action — match, post, exception — is timestamped and attributable.

i.

~3 FTEs of AP work returned

Redeployed onto vendor management and analysis. The team got smaller, the volume kept growing — and close still got faster.

ii.

Month-end close shortened by ~4 business days

Backlog stopped being the gating item. Close lands on the same date every month — and finance leadership can plan against it.

iii.

Duplicate catches paid for the build in Q1

Caught at ingestion, not after payment. The first quarter alone covered the project cost outright.

iv.

Audit trail per invoice

Every action timestamped and attributable. Every document archived with a structured filename and fully searchable.

09 · Technology Stack

Reliability, observability, and clean integration with the ERP.

Trigger
Mailbox (IMAP) + EDI feeds — per-message event ingestion, no polling lag.
Identity
Per-entity resolver — ERP credentials, coding rules, and GL mappings loaded at runtime.
Document Handling
PDF + OCR + email body + EDI — single normalization layer across formats.
AI Layer
LLM line-item extraction · supplier-aware coding · validated structured output.
ERP
Multi-entity write-back · automated three-way match against PO + goods receipt.
Exception Queue
AP review with original document, partial match, and exact break reason.
Archive & Logging
Structured filenames · per-action audit log · fully searchable.
10 · About Envyro

Production-grade AI agents — not demos.

Envyro is a specialized AI agency designing, deploying, and maintaining custom AI agents and pipelines that work in production. We stay on the call as your systems evolve.

SaaS · Collision Repair

Nexsyis

Shop management platform · AI email pipeline embedded into the stack.

Commercial · Maritimes

Office Interiors

Office equipment & service · bilingual voice AI for inbound calls.

Public Sector · Durham, NC

Durham County

350K+ residents · 24/7 GenAI resident support across municipal services.

Real Estate · NYSE

Veris Residential

$1.6B NYSE-listed REIT · resident-services AI across the portfolio.

Let's talk

Got an AP queue holding close hostage?

Tell us where the volume sits. We'll show you what a production-grade AP pipeline into your ERP looks like — and what the next two weeks could return.

matea@envyro.io 519 · 658 · 3579 envyro.io