Settyl

I want to share something specific about how the procurement cycle at industrial manufacturers could work differently — and what the 4-agent procurement cluster in Lasya AI actually does, step by step.

The current state at most industrial manufacturers is predictable. An RFP goes out. The category manager tracks responses across email threads and a shared spreadsheet. Supplier proposals arrive in different formats — PDF attachments, portal submissions, emailed Excel files. Each response is manually reviewed, scored against a set of criteria that varies by who's doing the scoring, and compared in a spreadsheet that will be wrong three days after it's updated.

Risk assessment — financial health of the supplier, sanctions exposure, geopolitical concentration, ESG posture — is a separate exercise, done quarterly, in a different spreadsheet, by a different team. It rarely connects to the sourcing event in real time.

When a supplier is selected, someone creates a purchase order. They open the ERP. They re-key the supplier name, the line items, the pricing, the delivery terms, the GL codes. They submit it for approval. The approval sits in someone's queue. The PO posts — usually 2–3 days after it should have.

This is a 4–6 week cycle for a process that should take 4 minutes.

At Settyl, the Hub 01 Autonomous Procurement cluster runs the full sequence as four agent steps — sharing context through a single working memory so that nothing is re-keyed and nothing is re-prompted.

The Intake Agent receives the supplier proposal — email, PDF, portal upload, any format — and extracts the full structure: vendor identity, line items, pricing tiers, commercial terms, SLA commitments. This happens in seconds.

The Compliance Agent takes what the Intake Agent extracted and scores it against your enterprise contract template. SLA gaps. Pricing deviations. Regulatory mismatches specific to your industry. Non-standard liability clauses. The audit is automated and auditable.

The Risk Agent runs simultaneously across five dimensions: historical supplier performance from your ERP data, financial health from public filings, sanctions screening against live global watchlists, geopolitical exposure by country of origin, ESG posture from relevant indices. One output: a single risk index. Not five reports.

The PO Agent takes all of this context — supplier, terms, compliance status, risk index — and generates a compliant purchase order payload with context-based ERP field mapping. The PO writes directly to SAP or Oracle via the Lasya Connector. No IT ticket. No ERP login. No manual entry.

4 minutes from supplier email to compliant PO in ERP. 99.3% first-pass compliance rate. 62% reduction in maverick spend. Zero ERP migrations required — the connector sits beside your existing system.

Bring one supplier email to the live session at settyl.com

Most enterprise AI projects don’t fail because the models are weak.

They fail at the handoffs.

One AI handles procurement.
Another handles logistics visibility.
Another handles invoice matching.

Each performs well inside its own silo.
But the moment work moves between systems, humans step back in:
• Re-entering PO data
• Reconciling freight invoices
• Posting transactions into ERP
• Managing operational exceptions manually

That’s where the real operational leakage happens.

Lasya AI was built to solve the coordination problem — not just automate isolated tasks.

Instead of disconnected point AI tools, Lasya operates as an agent mesh across procurement, logistics, and finance:
→ Autonomous procurement workflows
→ Multimodal logistics orchestration
→ Freight audit & finance automation
→ ERP-connected ex*****on loops

AI shouldn’t stop at recommendations.
It should complete the operational loop.

Watch the new Settyl product story: The Cost of the Handoff

https://youtu.be/4wnk3YdcCjg

Settyl | Lasya AI | The Operating System for Autonomous Supply Chains Most enterprise AI systems fail at one place:The handoff between departments.Procurement AI. Logistics AI. Finance AI.Each works independently.But humans sti...

Numbers are only meaningful when they're translated into operational reality.

So let's do that with 98% PO matching accuracy.

Scenario: 500 invoices processed per month.

Manual matching (15% average error rate):
→ 75 invoices with mismatches
→ 75 vendor disputes or queries per month
→ 75 rework cycles: locate the PO, find the GRN, cross-reference the contract
→ 75 delayed payments affecting vendor relationships
→ Approximately 150–200 analyst hours spent on rework monthly

Lasya AI (98% accuracy):
→ 10 exceptions per month — each flagged with full context
→ 65 fewer disputes, calls, and rework cycles
→ ~160 analyst hours saved every month
→ Vendor payments on time, vendor relationships intact

That's the operational reality of the accuracy gap.

And at enterprise scale — 1,500+ branches processing thousands of invoices across multiple ERP instances — the numbers compound significantly.

Lasya achieves 98% through a combination of:
→ Template-agnostic OCR reading any invoice format with high accuracy
→ AI-driven 3-way matching: PO, GRN, and invoice cross-referenced simultaneously
→ Rate contract validation at line-item level
→ Context-rich exception flagging for efficient human resolution

This is what accounts payable looks like when the manual layer is removed.

🔗 https://www.settyl.com/solutions/finance-accounting