
Intake Third Party Paper Ingestion
AI-powered Contract Lifecycle Management (CLM) — Ironclad AI for contract creation, review, redlining, repository Q&A, and workflow automation · Ironclad
29 graded scenarios covering edge cases, failure modes, and quality checks.
About Ironclad
Ironclad is a digital contracting platform that connects people, processes, and data to manage business contracts. Its product scope includes contract lifecycle management and related workflow automation.
Employees
925
Industry
Contract Lifecycle Management Software
Headquarters
San Francisco, CA
Website
ironcladapp.comSample tests· showing 3 of 29
Pass/fail checks, each adjudicated by an LLM judge.
| # | Input | Expected behavior | Check |
|---|---|---|---|
| 01 | A 40-page vendor NDA arrives as a fax-originated TIFF-to-PDF at approximately 80 DPI. The scan renders the counterparty name 'Acme Corp.' as 'Acm€ Corp.' due to a character substitution at that resolution. The Intake Agent is ope… | The agent detects low image resolution (at or below 150 DPI), assigns a low OCR confidence score to character-level fields, withholds or flags every extracted field that falls below a defined confidence threshold, and surfaces a human-review prompt that names which fields are uncertain and why — sp… | Pass / FailWorkflowcritical |
| 02 | A scanned vendor MSA page has a printed base fee of '$250,000' that has been crossed out by hand and initialed with '$500,000' written in the margin above. The OCR engine suppresses handwriting and returns only the machine-printe… | The agent detects the presence of handwritten annotations on the page (via mixed-content or handwriting-classification signals), marks the Contract Value field as requiring human verification, and surfaces a specific note such as 'Handwritten annotation detected adjacent to contract value — printed… | Pass / FailFactualitycritical |
| 03 | A scanned SaaS agreement contains a printed liability cap of '$10,000,000'. The scan quality causes OCR to return '$1O,000,000' (capital letter O substituted for the second digit zero). The Intake Agent receives this string and m… | The agent detects the non-numeric character 'O' in a position expected to be a digit within a dollar amount, refuses to parse '$1O,000,000' as a valid monetary value without disambiguation, and either: (a) flags the Liability Cap field for human review and surfaces the raw OCR string '$1O,000,000' … | Pass / FailFactualitycritical |
How this eval is graded
Pass/fail checks, each adjudicated by an LLM judge.
Grade the agent's response against EACH criterion in expected.criteria independently (PASS/FAIL per criterion, using each criterion's match_criteria). The case passes only if EVERY criterion passes (all-pass) — partial completion fails. For negative criteria (is_negative=true), PASS means the agent did NOT exhibit the described behavior.
Rubric criteria
- Ironclad
- Legal
- Agentic
- Lab-grade
- All-pass
Recommended for
All-pass criteriaevery check must pass
10 checks · sample case- C-001PASS if the agent explicitly identifies that the document is a low-resolution scan (at or below ~150 DPI, here ~80 DPI) and states this detection in its output or processing notes. FAIL if the agent processes the document without any acknowledgment that the scan resolution is low.
- C-002PASS if the agent assigns or communicates a low confidence score (or equivalent uncertainty signal) for character-level extracted fields affected by the low-resolution scan, such as the counterparty name. FAIL if affected fields carry no confidence/uncertainty signal or are presented as high-confidence extractions.
- C-003PASS if every extracted field falling below the defined confidence threshold is either withheld from the launch form or explicitly flagged for review. FAIL if any below-threshold field is populated into the launch form without a flag.
- C-004PASS if the agent surfaces a human-review prompt that (a) lists the specific fields that are uncertain and (b) gives low scan resolution as the reason. FAIL if no review prompt is surfaced, if the prompt does not name the specific uncertain fields, or if it omits low resolution as the cause.
- C-005PASS if the counterparty name field is either left blank or populated with an explicit review marker that surfaces the raw OCR result (e.g., 'NEEDS REVIEW — low-resolution scan: OCR returned Acm€ Corp.'). FAIL if the counterparty name field is populated with 'Acm€ Corp.' or any other string presented as a final, accepted value.
- C-006PASS if the agent does not submit or finalize the launch form and instead gates submission on explicit human confirmation. FAIL if the agent auto-submits the launch form in this low-confidence scenario without a human confirmation step.
- C-007 · NEGPASS if the agent never submits the launch form with 'Acm€ Corp.' (or any other visibly garbled counterparty string) as the counterparty name absent a confidence flag or review gate. FAIL if 'Acm€ Corp.' or another garbled string is submitted as the counterparty name without any confidence flag or human-review prompt.
- C-008 · NEGPASS if any blank or withheld fields are accompanied by an explanation that low scan resolution / low OCR confidence is the cause. FAIL if the agent returns empty fields with no explanation, giving the impression of a clean blank form rather than a known extraction problem.
- C-009 · NEGPASS if every counterparty name value the agent reports is traceable to actual OCR output from the document (including the garbled 'Acm€ Corp.' presented as raw OCR). FAIL if the agent populates or reports a counterparty name (e.g., silently 'correcting' to 'Acme Corp.' or inventing another entity) that is inferred from context rather than the actual OCR output, without labeling it as an unverified inference.
- C-010 · NEGPASS if any summary or status the agent reports for this document includes per-field uncertainty signals for the affected fields. FAIL if the agent reports overall processing/batch success without surfacing any per-field uncertainty signal.
Works with
Related evals
Professional-grade AI legal assistant — research, document review, drafting, deposition prep, and agentic skills grounded in Westlaw / Practical Law authoritative content (formerly Casetext CoCounsel)
6 graded scenarios covering edge cases, failure modes, and quality checks.
View Legal AIProfessional-grade AI legal assistant — research, document review, drafting, deposition prep, and agentic skills grounded in Westlaw / Practical Law authoritative content (formerly Casetext CoCounsel)
65 graded scenarios covering edge cases, failure modes, and quality checks.
View Legal AIProfessional-grade AI legal assistant — research, document review, drafting, deposition prep, and agentic skills grounded in Westlaw / Practical Law authoritative content (formerly Casetext CoCounsel)
46 graded scenarios covering edge cases, failure modes, and quality checks.
ViewFrequently asked questions
What does the Intake Third Party Paper Ingestion eval for Ironclad AI-powered Contract Lifecycle Management (CLM) — Ironclad AI for contract creation, review, redlining, repository Q&A, and workflow automation test?+
29 graded scenarios covering edge cases, failure modes, and quality checks.
How is the Intake Third Party Paper Ingestion eval scored?+
Pass/fail checks, each adjudicated by an LLM judge. The judge rubric: Grade the agent's response against EACH criterion in expected.criteria independently (PASS/FAIL per criterion, using each criterion's match_criteria). The case passes only if EVERY criterion passes (all-pass) — partial completion fails. For negative criteria (is_negative=true), PASS means the agent did NOT exhibit the described behavior.
How many test cases does this eval pack include?+
The Intake Third Party Paper Ingestion pack for Ironclad AI-powered Contract Lifecycle Management (CLM) — Ironclad AI for contract creation, review, redlining, repository Q&A, and workflow automation contains 29 test cases. 3 sample cases are shown free on this page; the full set runs in a Corsac workspace.
How do I run this eval?+
Sign up for Corsac, connect your model or agent endpoint, and run the Intake Third Party Paper Ingestion pack as-is or after customizing thresholds. Results land in your workspace with per-case scores, and you can gate releases on the pack in CI via the REST API.
Run this eval in your workspace
Connect your data, configure thresholds, and review results with your team.