CONTRACT DATA CAPTURE
You cannot search for a clause you did not know to look for.
Maximum coverage on a real contract family: what one read of a fifty-thousand-contract estate holds, and what was delivered.
- ESTATE
- about 55,000 contracts, phase one 8,500, a German energy group
- READ ONCE
- about 50,000 documents staged for phase one
- CAPTURED
- 123 fields per document on average, median 70
- DELIVERED
- 75 fields per contract dataset, into Microsoft Dynamics
- HELD
- 1,042,665 source observations, 52,937 reusable concepts
One contract is six documents.
Contracts are business rules your systems never see or obey. A master agreement, its amendments, its technical annexes and its attachments are one contract, and the rule that matters is usually in the annex. Talonic reads the whole family, resolves it into one dataset, and dates what is in force.
MASTER AGREEMENT
The frame. Parties, term, governing terms.
ADDENDUM
Where the price escalation clause actually lives.
PO AND SOW
What was ordered, at what rate, for how long.
INVOICE
What was charged. The side every clause is checked against.
What one read holds.
Fields captured per document
Average across the Milestone 1 batch, median 70. The read was never scoped to a field list.GETEC · Milestone 1 batch · July 2026
123
Fields delivered per contract dataset
The governed core the digitalisation and the migration needed. Schema V3, frozen at Milestone 1, into Dynamics through SAP BTP.GETEC · Schema V3 · Milestone 1
75
Source observations held
Every value read from every page in the live tenant, each with its page and position, whether or not a department has asked for it yet.GETEC · live tenant · 26 August 2026
1,042,665
Reusable concepts in the registry
What the registry organised those observations into on its own. Machine-generated labels, not the customer’s vocabulary.GETEC · live tenant · 26 August 2026
52,937
Per document and per contract dataset are different denominators and do not divide into each other. 123 is fields per document; 75 is fields per contract dataset. The comparison is one of scale, not a ratio. Full receipts, with populations and dates, are on the GETEC case page.
Three steps, and the second department pays nothing.
Capture everything
no clause list up front
Build the relationships
Make it available
- Capture everything. No clause list up front. The AI reads every page and takes every value it finds. This is the expensive half, and it is paid once.
- Build the relationships. It resolves the whole family, links the addendum to the agreement it amends, and dates what is in force.
- Make it available. Query the results with AI, or use the rules to drive your systems. The first department gets its 75 fields. The next department asks a question the first never scoped, and the answer is already there.
The same read, delivered in a different shape, is the PHOENIX case: one batch of 1,130 contracts inside a programme of 22,000 across thirty countries and ninety-six operating entities, delivered in the import shape their procurement system expects, with every cell it could not determine shipped as a documented blank.
Bring one contract family.
A master agreement, its amendments and its annexes. We read it once, hand back every value with its page, and show you what a second department would already have.
Note: All figures on this page are the GETEC and PHOENIX case figures already published on their case pages, with the same dates, populations and denominators.