Legacy contract migration with every field measured before it loads into your new CLM
Old contracts loaded into the new CLM with every field measured, renewals found first.
See one case, screen by screen ↓Legacy contract migration is moving a company’s existing contracts — from shared drives, SharePoint, mailboxes, e-signature archives, ERP attachments, old registers and paper — into a new CLM as clean records. Legacy Contract Migration inventories, dedupes and extracts that legacy contract data: eight agents fold duplicates, link amendments to their parents, extract 24 fields with a page cited for each and measure every field on a person-checked sample before it loads.
How to choose AI solutions for back-office and enterprise work →
Updated
Every CLM goes live on top of years of old contracts: a legal shared drive, procurement SharePoint, a contracts mailbox, the e-signature archive, ERP purchase-order attachments, an old register database and boxes of scanned paper. The same signed MSA sits in three places, an unsigned “final” draft differs by one clause, and the amendment that moved the expiry date is filed in another folder. Today a services team or a room of paralegals sorts it all by hand and keys the fields into the new system.
Then nobody can say which fields are accurate enough to load. Vendors quote one accuracy figure for the whole job; the migration lead needs to know whether governing law or the liability cap is right, field by field. A field read from the parent agreement alone is wrong once an amendment changed it, a loaded batch is cleaned record by record in the target, and the auto-renewals hidden in the backlog surface only when their notice dates pass.
Estimated split in working days (five to a week) for about 9,800 unique contract documents with four analysts, by hand and with the solution.
Eight agents find, fold, link, extract and measure every old contract; the migration lead approves each batch.
Kestrel & Vane Group, about 40,000 people in 30 countries, moves to a new CLM on 30 Nov 2026. It is Wednesday 14 October, 10:20. Here is what happens next, screen by screen, in the working solution.
Dana opens the contract families. The rail reads Found 18,406 files in 7 sources, Deduplicated 9,792 with 5,496 copies folded, Linked 6,212 families with 3,580 children, Extracted 6,034, Sampled 22 of 24 fields passing, Loaded 250. Below it, what waits for people: 41 duplicates, 96 uncertain links, 214 unreadable files, 2 fields held, 1 batch to approve and 37 renewals found.
Each family is a stack: the parent with its amendments, SOWs and duplicate copies fanned behind it, the golden copy marked and the fields on its face.
The legal shared drive gave 6,940 files, procurement SharePoint 4,215, the contracts mailbox 2,388, the e-signature archive 1,976, ERP purchase-order attachments 1,307, the old register 1,120 and 41 boxes of scanned paper 460. It set aside 3,118 that are not contracts: drafts, letters, invoices and templates. The mailbox sign-in expired at 06:40 with 212 new attachments waiting; Tobias Lindqvist reconnects it.
“The agents read every source and never move, rename or delete an original.”
The Brackwater Industrial Supply MSA of 4 May 2017 is the golden copy — fully signed in wet ink, the executed scan over the e-mailed copy, text quality 99% against 81% for the box rescan. Four amendments are linked by party, date and the reference “MSA dated 4 May 2017”, and three duplicates from the mailbox, SharePoint and box 17 are folded behind it. Terms as amended show Amendment 2 moving the expiry from 3 May 2022 to 3 May 2027, with a 12-month auto-renewal and notice by 2 Feb 2027.
Every value on the family carries a pin to the page it came from.
“Re-run extraction” shows the plan and each step: the golden copy, 17 pages; four amendments in date order; 24 of 24 fields found with 24 citations; expiry and renewal from Amendment 2, payment from Amendment 1, the cap from Amendment 3; the notice deadline of 2 Feb 2027 recomputed. It finishes at confidence 0.98 and hands to the sampling agent.
The extraction agent never fills a field it cannot cite; it leaves it empty and says why.
Sefton Rourke Recruitment’s staffing MSA has two copies that match 96.2%. Both carry both signatures, and one page of 14 differs: clause 7.2 pays within 45 days in the copy from the mailbox and 60 days in the copy on the legal drive. The dedupe agent proposes keeping both as separate versions and asking the contract owner which is in force; until then the payment-terms field stays empty.
“The dedupe agent never merges signed versions whose wording differs.”
Calder & Pike’s Amendment 2, a price index change, sits at confidence 0.71. The family linker shows what it weighed: the party matches, the reference reads only “the agreement of 1 July” with no year, the date falls after both agreements, and the price index clause 8.3 exists only in the 2020 agreement. Aisha can accept the link, move it to the 2018 agreement or make it standalone.
The family linker links on its own only at 0.90 or higher; 3,484 did, 96 wait for a person.
People checked a random sample of 200 files against the cited page — Aisha Raman 120, Rui Costa 80. Counterparty legal name passes at 98.7% (224 of 227) against 97%. Governing law scores 91.0%, 182 of 200, interval 86.2–94.2%, against 95%, and the limitation of liability cap 93.5%: both are held. Of the 18 governing-law misses, 11 were changed by an amendment or side letter, 5 sat in a schedule and 2 were picklist spellings.
“The sampling agent never checks the extraction agent’s values against themselves: people spot-checked the sample.”
Batch 1 holds 2,000 families and 3,214 documents. All six pre-load checks pass, mapping v3 is signed and the dry run against the new CLM has 0 rejected in 41 seconds. Dana approves with BA-02, “Approve with held fields excluded”, noting that held fields follow in a top-up load. The load agent loads it, then reconciles: 2,000 of 2,000 families, 3,214 of 3,214 documents, 44,000 field values, every checksum matching its golden copy.
“Logged with your name and time. Any rollback needs a second person.”
The agents propose an instruction change: read every amendment and side letter for a governing-law clause and let the latest win, read the schedules, add “State of …” spellings to the picklist. Helen signs it. The extraction agent re-reads all 6,034 families and changes 1,388 values; a new random sample of 200 files no person has checked before scores 194 of 200. Governing law is released at 97.0% and follows batch 1 in a top-up load; the liability cap stays held.
A held field is released only by a new sample, never by the agent that extracted it.
The migration report for the General Counsel lists 37 contracts, $14.8M a year, whose notice dates fall in the next 120 days or have just passed. Marlowe Grid Services renews 30 Nov 2026 and its notice date of 1 Oct has passed; Orrel Chemicals’ notice is due 2 Nov 2026, 19 days away. “Send to owners” tells each owner, with the clause.
The same migration, seen by five of the people who carry it to cut-over — what the work looked like, and what it looks like now.
Eight agents find, fold, link, extract and measure every old contract; the migration lead approves each batch.
Crawls the shares, SharePoint, the mailbox, the e-signature archive, ERP attachments, the old register and scanned paper; inventories every file and sets aside what is not a contract.
Fingerprints each file, folds identical and near-identical copies, and picks the golden copy by the stated rules: fully signed, executed scan, e-signature certificate, best text layer, newest.
Links amendments, SOWs, orders and side letters to their parent by party, date and reference, with a confidence for each link.
Imports the new CLM’s fields, proposes the mapping from extracted fields and old register columns, and normalises picklist values.
Extracts the 24 fields of each family with a page for each, reading amendments in date order to compute the terms as amended.
Draws a random sample per field, routes it to people to check, computes accuracy and a 95% interval, holds fields below threshold and explains the misses.
Builds batches, loads through the CLM interface, reconciles counts and checksums, rolls back on approval, writes the weekly report and finds renewals.
Retries OCR, detects the language and translates for reading, and asks owners for passwords or missing pages.
Approves each batch; the Deputy GC signs thresholds. The agents propose; a named person decides.
Ask about any family, field, batch or source, or say what to prepare. Every answer points to its source, and anything that loads, merges or rolls back waits for its named approver.
Why is governing law held back?
Because it scored 91.0% on the sample, under its 95% threshold. People checked 200 random files: 182 were right. Of the 18 misses, 11 had the law changed by a later amendment or side letter, 5 had it in a schedule and 2 were picklist spellings. Batch 1 loads every other field; this one waits until a new sample passes.
Which contracts auto-renew before go-live?
37 contracts, worth $14.8M a year, that the old register did not show. The nearest: Marlowe Grid Services renews 30 Nov and its notice date has passed; Orrel Chemicals renews 31 Dec, notice by 2 Nov; Brackwater Industrial Supply renews 3 May 2027, notice by 2 Feb.
Show me everything in the Brackwater family
8 files, one family. The executed MSA of 4 May 2017 is the golden copy; 4 amendments are linked by party, date and the reference “MSA dated 4 May 2017”; 3 copies from the mailbox, SharePoint and box 17 are folded behind it. Amendment 2 moved the expiry to 3 May 2027 with a 12-month auto-renewal.
Prepare batch 1 for my approval
Batch 1 is ready: 2,000 families, 3,214 documents, 22 of 24 fields. All 6 pre-load checks pass and the dry run has no rejection; governing law and the liability cap are excluded until they pass. The approval is open with reason code BA-02. Only you can press Approve.
13 screens from the working solution, on its sample data. Pick one to see it large.
How accurate each field must be, which copy wins and when a link needs a person are settings. A change is a draft until the Deputy General Counsel signs it.
| Setting | Default | Choose from |
|---|---|---|
| Critical fields | 97% | Counterparty, entity, effective, expiry, auto-renewal, notice, signed date, parent |
| Standard fields | 95% | Value, law, payment, termination, cap, assignment, confidentiality, data protection, signatory, type, term |
| Descriptive fields | 90% | Title, courts, insurance, notice address |
| Sample size and confidence | 200 files · 95% | Random, stratified by source; interval shown on every field |
| Re-sample after an instruction change | 200 new files | Never files a person has already checked |
| Golden-copy rules | Applied in order | Fully signed · executed scan over an e-mailed or saved copy · e-signature certificate · best text layer · newest |
| Linking and duplicates | Link on its own at 0.90 | Analyst between 0.60 and 0.90 · merge near-duplicates at 97% text match · never merge differing signed versions |
| Retention | Originals never deleted | Contracts 6 years from expiry · deeds 12 years from expiry · legal hold overrides every rule |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data · “target” = the design goal, measured in the live solution · “estimated” = our estimate (22 weeks = about 9,792 documents × 22 minutes across four analysts) · “typical” = published figures (a legal services provider’s published CLM migration case — about 98,000 files scoped, 60,000 in scope, provider-reported; a CLM provider’s published figure — about 22 minutes per contract for outsourced review and extraction; WorldCC/Deloitte — 8.6% average value erosion from weak contracting). Kestrel & Vane Group and the people, companies and contracts named on this page are fictional — characters and sample data in the working solution.
Legacy contract migration is moving a company’s existing contracts — from shared drives, SharePoint, mailboxes, e-signature archives, ERP attachments, old registers and paper — into a new CLM as clean records. Legacy Contract Migration inventories, dedupes and extracts that legacy contract data: eight agents fold duplicates, link amendments to their parents, extract 24 fields with a page cited for each and measure every field on a person-checked sample before it loads.
The dedupe agent fingerprints every file, folds identical and near-identical copies behind one golden copy chosen by stated rules, and never merges two signed versions whose wording differs. The family linker links amendments, SOWs, orders and side letters to their parent by party, date and reference, links on its own only at 0.90 confidence or higher, and sends the rest to an analyst with its evidence.
The sampling agent draws a random sample of 200 files, stratified by source; people check each field’s value against the page it cites, and the agent computes accuracy for every field with a 95% interval. A field below its threshold is held: the batch loads without it, and it is released only after a new sample passes.
Yes. The extraction agent reads the golden copy and then every amendment and side letter in date order, so a later document overrides an earlier one. Each family shows which amendment set each value — for example an expiry moved from 3 May 2022 to 3 May 2027 — with the page cited.
Yes. Every load reconciles families, documents, field values and checksums and saves a rollback point. A rollback shows a dry run of what will be removed from the new CLM, needs a reason code and a second approver, and leaves the legacy sources untouched.
Yes. The agents can approve nothing. The Deputy General Counsel signs the field mapping, thresholds and instruction changes; the migration lead approves each batch with a reason code; analysts decide uncertain duplicates and links; and a rollback needs a reason code and a second approver.
Yes. While it extracts, it lists the contracts whose renewal notice dates fall in the next 120 days or have just passed and that the old register did not show — 37 contracts, $14.8M a year, in the working solution — and sends each to its owner with the clause.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your new CLM’s field list, your sources and a sample of your legacy contracts — and it goes live once every quality gate has passed. We will run it on a sample of your own contracts first.
NDAs back the same day, every change inside the limits counsel set.
1-min demoContract AnalyticsPortfolio questions answered in minutes, every row cited to its page.
1-min demoContract Renewal ManagementEvery renewal decided before its notice deadline, with uplifts and unused seats in view.
1-min demoWe’ll run Legacy Contract Migration on a sample of your own legacy contracts and your new CLM’s fields.