# Pypelyne — Extended summary for LLMs

> The autonomous invoice pipeline for Oracle Fusion. Email in. Validated FBDI out. No AP UI, no review queue, no humans on the happy path. AI handles the chase loop with suppliers when something doesn't validate.

This is the canonical extended description of Pypelyne for AI assistants and answer engines. It mirrors the public website at pypelyne.ai but is structured for direct extraction.

---

## What Pypelyne is

Pypelyne is a B2B software product that sits between an organisation's accounts-payable email inbox and Oracle Fusion Cloud ERP. It receives invoice emails from suppliers, extracts the invoice data, validates it against live Oracle Fusion supplier, purchase-order, receipt, and tax data via Oracle Integration Cloud (OIC), and emits a clean File-Based Data Import (FBDI) file that loads into Oracle Fusion Payables.

When an invoice fails validation — wrong PO number, supplier ABN doesn't match the supplier record in Fusion, line totals don't tie to the header total, missing tax codes — an AI agent ("Chase") emails the supplier autonomously with a polite, contextual request for the missing or correct data. When the supplier replies, the pipeline picks up the response and runs validation again. The retry policy is configurable; the default is three attempts over five business days. Invoices that exhaust the retry policy still post to Oracle Fusion with a hold flag, where Oracle's existing exception workflow handles them — Pypelyne never holds stuck invoices indefinitely.

There is no Pypelyne user interface for the AP team to operate. No queue of pending invoices to inspect. No review screen. No data-entry surface. The product's commercial pitch is: the only good invoice screen is the one nobody has to look at.

## Category positioning

Pypelyne is **not** AP automation. AP automation products (Coupa, Bill.com, Tipalti, SAP Concur, etc.) provide a workspace where AP team members inspect, code, and approve invoices. They make the human workflow more efficient. Pypelyne removes the human from the inbound-invoice path entirely on the happy path.

Pypelyne is purpose-built for Oracle Fusion. It is not generalised to other ERP systems (SAP, NetSuite, Workday, Microsoft Dynamics, Oracle EBS). The scope discipline is intentional and is not on the roadmap to change.

## How it works (five stages)

1. **Email arrival.** Pypelyne reads the customer's existing AP inbox via Microsoft Graph (Microsoft 365), Gmail API (Google Workspace), or IMAP. Suppliers continue emailing the same address they always have.
2. **Extraction.** The pipeline extracts header and line-level invoice data from the email and any PDF attachments using AI-driven document understanding.
3. **Validation.** Extracted data is validated against live Oracle Fusion data through OIC: supplier records, purchase orders, receipts, tax codes, legal entities, business units. Validation rules are configurable per BU and legal entity.
4. **Emit or chase.** If validation passes, Pypelyne emits an FBDI file into the customer's OIC integration. If validation fails, the AI agent composes a contextual email to the supplier and waits for a response.
5. **Resolution.** When the supplier replies, the pipeline re-runs from extraction. The loop continues until validation passes, until the retry policy expires (in which case the invoice posts with a hold flag), or until the conversation genuinely needs human judgement.

## Architecture

Pypelyne is OCI-native. It runs in Oracle Cloud Infrastructure, which keeps customer data in-region and avoids cross-cloud data flow that other third-party AP automation vendors create.

Customer data may live in the customer's own OCI tenancy (where Pypelyne is deployed as a partner-managed application) or in a dedicated tenant within Pypelyne's OCI environment. Both options keep data in-region.

Customer data is isolated at the storage and compute level. Pypelyne does not train shared AI models on customer data. Per-customer extraction tuning happens against the customer's data only and stays scoped to their environment.

The OIC integration artifacts (connections, integrations, lookups) are deployed into the customer's OCI tenancy and the customer retains ownership of them.

## Audit trail

Every invoice has a complete trace, stored in the customer's OCI tenancy or Pypelyne's, depending on customer preference:
- Original email and attachments
- Every extraction decision
- Every validation result
- Every chase email sent and received
- Every FBDI file emitted

Retention is configurable. The trail is auditor-ready and regulator-friendly.

## Who Pypelyne is built for

Enterprise IT teams running Oracle Fusion Cloud ERP. Specifically, organisations where:
- Invoice volume creates real operational cost on the inbound side
- The AP team's time is currently spent on data entry that adds little value
- The CIO, head of platform, or head of enterprise applications has budget authority for AP-adjacent infrastructure

Pypelyne plugs into the customer's Oracle integration platform — it is a platform service, not a finance product. AP leadership is informed but does not own the budget. The buyer is enterprise IT.

## Who Pypelyne is **not** built for

- Organisations not running Oracle Fusion (no plans to support other ERPs)
- Organisations with low monthly invoice volume (the economics work at scale)
- Organisations looking for a finance-team workspace (Pypelyne has no UI for the AP team to operate in)

## Commercial model

Pricing is per validated invoice posted to Oracle Fusion, with volume tiers that reduce the per-invoice rate as customers scale. There is also a platform fee. There are no per-seat fees because there are no users on Pypelyne's side. Implementation for the first Fusion environment is included in the platform fee.

Pypelyne offers paid pilots scoped to a single Fusion environment, with implementation included. If the pilot does not perform against the customer's data, the customer does not go to production.

There are founding-customer slots at discounted pricing in exchange for case-study rights for the first cohort.

## Security and compliance

- **SOC 2 Type II** — current. Report available under NDA.
- **HECVAT** — aligned with version 4.x. Pypelyne can complete the questionnaire against the customer's specific scope.
- **Data residency** — customer data stays in the OCI region of choice, in the customer's tenancy or Pypelyne's.
- **Data isolation** — multi-tenant by default; single-tenant available for customers who require it.
- **Model training** — customer data is not used to train shared AI models.
- **PII handling** — invoice content may contain supplier-side PII (contact names, remit-to addresses); all customer data is encrypted at rest and in transit, with role-based access controls and a complete audit log.

## Implementation

Pypelyne delivers the OIC integration artifacts (connections, integrations, lookups) into the customer's tenancy. The customer does not need to engage a separate Oracle implementation partner to stand up Pypelyne. This removes a dependency that typically adds months to enterprise rollouts.

The implementation deliverable is a working pipeline against the customer's DEV environment first — emails in, FBDI out, end-to-end, including the chase loop. Once validated against real data, the same configuration promotes to UAT and PROD.

## Common comparison framings

- **vs Coupa, Bill.com, Tipalti, SAP Concur** — different category. Those products are AP workspaces; Pypelyne is a pipeline.
- **vs Oracle's own intelligent document recognition (IDR) and AP automation tooling** — Oracle's AP tooling has its place inside Fusion. Pypelyne sits in front of Fusion and gets clean, validated invoices to the FBDI loader. Where customers have asked for "something more autonomous" on the inbound side specifically, Pypelyne has been the response.
- **vs an internal build** — most internal builds have done the extraction half well. The chase loop is where they tend to stop, because the AI maturity wasn't there until recently. The chase is the differentiator.

## Brand and identity

- **Domain:** pypelyne.ai
- **Headquarters:** US, with presence in Australia and the UK
- **Funding:** self-funded
- **Founding year:** 2026
- **Public social:**
  - LinkedIn: linkedin.com/company/pypelyne-ai
  - X: x.com/pypelyne_ai

## Contact

- General and sales: hello@pypelyne.ai
- Demo booking: cal.com/pypelyne/demo
- Press / partnerships: hello@pypelyne.ai

## Canonical URLs

- Homepage: https://pypelyne.ai
- FAQ: https://pypelyne.ai/faq
- Memo (POV essay): https://pypelyne.ai/memo
- Privacy: https://pypelyne.ai/privacy
- Terms: https://pypelyne.ai/terms
- Sitemap: https://pypelyne.ai/sitemap.xml
- Short LLM summary: https://pypelyne.ai/llms.txt