Procurement runs on email and spreadsheets. We can do better.

Your PO status is in the ERP. Vendor emails are in Outlook. Contracts are in a shared drive. Shipment tracking is in a carrier portal. Spend data is in spreadsheets. Every procurement director knows this.

Nobody fixes it because the fix used to require building a single integrated platform: a Coupa implementation, 18 months, and a seven-figure license. So it doesn't get fixed.

Things fall through the cracks. The $12K overcharge nobody had time to cross-reference. The contract that auto-renewed at bad terms because the renewal date lived in a PDF nobody opened. The shipment that slipped three weeks without anyone flagging it because the carrier sent the delay notice to a shared inbox nobody monitors.

This is not a technology problem. It is an information problem. The data exists. It is scattered across five systems and twelve email threads and nobody has time to connect it.

Procurement teams at mid-market manufacturers are especially exposed. They produce more procurement documents per dollar of revenue than any other vertical: purchase orders, invoices, bills of lading, packing slips, material test reports, carrier emails, supplier amendments. Each one is a place where money can leak. The same vertical that generates the most documents has the thinnest margins. And the enterprise platforms that were built to solve this problem were never built for them.

The result is that most mid-market procurement operations run on a combination of their ERP, Outlook, a shared drive, and institutional memory. When the person who remembers the contract terms leaves, the institutional memory leaves with them. When the AP clerk who cross-references every invoice against the PO goes on vacation, overcharges go uncaught.

I have talked to dozens of procurement directors at manufacturers. Every one of them describes the same experience: they know things are slipping through. They know they are overpaying on some invoices. They know contracts are auto-renewing at terms they would renegotiate if they had time. They do not have time.

The problem is not laziness. The problem is volume.

LLMs changed what is possible.

A model that can read a PDF, an Excel file, and an email thread, understand what they mean, and connect them: that capability did not exist three years ago. Templates and rule engines could not do this. The previous generation of procurement software was built on extracting structured data from structured templates. If the invoice format changed, the template broke. If the vendor sent a rate confirmation via email instead of through the portal, the system never saw it.

The new generation can read everything that flows into a procurement operation in any format and connect it on the fly. A vendor sends a price-increase notice buried in the third paragraph of a reply to an unrelated thread. An old system ignores it. A new system reads the email, extracts the price change, links it to the affected contract and the open POs under that contract, and flags every upcoming invoice that will be affected.

That is not a feature. That is a new category.

The document extraction companies (ABBYY, Rossum, Hypatos) built good products for structured extraction. They read one document at a time, pull fields into structured data, and hand it to your ERP. But they do not read email. They do not cross-reference. They do not build a persistent graph of every entity in your procurement operation. They were built for a world where AI could recognize fields on a template. We live in a world where AI can read and reason.

The procurement software companies (Coupa, Ariba, Zip) built systems of record. Good ones, for large enterprises with 18 months and a million-dollar budget. They are now bolting AI onto platforms designed around manual data entry and approval workflows. The AI is an add-on to the architecture, not the foundation. When the foundation is a rule engine, the AI can only do what the rules allow. When the foundation is AI, the system can do whatever the data supports.

Procurement Intelligence sits on top of your existing systems. ERP, email, shared drives, carrier portals, spreadsheets: it reads everything that flows in, connects POs to shipments to contracts to invoices to vendors to items, and tells your team what to act on before money moves.

No templates. No per-vendor configuration. No 18-month implementation. Forward an email.

That sentence is doing real work. The entire onboarding for a Kynthar Pilot is: forward your procurement emails to an address we give you. Your AP team forwards 30 days of routine procurement traffic. No IT project. No ERP integration. No training sessions. The system reads what comes in, classifies it, extracts the relevant information, resolves it to the right entities, and starts building the graph.

Within 48 hours you have a connected view of your procurement operation that would have taken a traditional platform months to build. Not because we skip steps, but because the AI does the work that used to require templates, field mapping, and per-vendor configuration.

The value stack builds in layers:

1. Unified visibility. See the status of every PO, shipment, contract, vendor, and item in one place, always current.
2. Attention routing. The system tells you what needs attention right now. Late shipments, expiring contracts, pending approvals. You do not go looking for problems. Problems come to you.
3. Automatic intelligence. Cross-references documents that humans would never have time to compare. The overcharge on page 3 of the invoice that contradicts clause 4.2 of the contract. The carrier rate increase that does not match the amendment. The vendor who invoiced for 500 units when the BOL shows 480 received.
4. Compounding insight. The system gets smarter with every document. More vendor history means better anomaly baselines. More contracts means better compliance checks. It compounds like an analyst who never forgets anything.

That is Procurement Intelligence. It is not better document processing. It is a fundamentally different capability: plug it into your business and it understands what is happening across all your procurement data, in any format, without being told what to look for.

Apex Manufacturing started its Pilot on a Tuesday. By Thursday morning, Kynthar had surfaced an overcharge on a routine PO: $12,847 against a contract with a price-protection clause the AP team did not have time to cross-reference. That is the kind of finding that pays for an annual contract three times over, surfaced in 48 hours, with zero process change. No new software to learn. No workflow redesign. Just a forwarded email address and two days.

Procurement Intelligence is not document extraction (ABBYY, Rossum). It is not AP automation (Bill, Stampli). It is not procurement software (Coupa, Ariba, Zip). It is the intelligence layer that reads what those systems cannot and tells you what they do not.

Document extraction reads one document at a time and pulls fields. Procurement Intelligence reads every document, connects them across the entire operation, and surfaces what matters.

AP automation routes invoices through approval workflows. Procurement Intelligence catches the overcharge before the invoice enters the workflow.

Procurement software is a system of record. Procurement Intelligence is the system that reads what the system of record cannot see: the email threads, the PDF attachments, the Excel exports, the carrier notifications, the contract clauses buried in shared drives. It connects all of it and tells your team what to act on.

Procurement teams are not lazy. They are outgunned by volume. The answer is not a bigger team. The answer is a system that reads everything.

The best procurement technology disappears into your workflow. Forward an email. Get findings. No training manual. No change management. No 18-month implementation.

Every document that flows through your operation is either helping you or hiding something from you. The first step is making sure nothing hides.

We built Kynthar because nobody else will build this for the mid-market. The enterprise vendors will not build down. The document extractors will not build up. The horizontal AI companies will not build vertical. Someone has to build the intelligence layer that 50,000 US manufacturers need and nobody is shipping to them. That is us.