How work really gets done.
Operational workflows, approval trails and legal documents from real companies, prepared so enterprise agents learn the process, not just the paperwork.
Available as- 01 · Dataset
- 52+ companies' ordered approval trails
- 02 · Environment
- Agent tasks over approvals, operator-graded
Licensed, stocked, ready to ship.
Plus custom collection for what doesn't exist yet.
Operational workflows and approval trails
Text, documents and structured records across the network
Procurement, finance, HR, legal, sales, support and more
Intake, review, approval and archive
Every document human-checked before delivery
Quality, information security, AI and cloud security
Process data, end to end.
Documents arrive with the trail around them: who asked, who reviewed, what changed and who signed.
- Approval trails
- Request, review, approval and archive events with roles and timing.
- Legal documents
- Contracts, policies and revisions with clause-level structure.
- Operational workflows
- Procurement, finance, HR and support processes reconstructed step by step.
- Decision labels
- Why something was approved, escalated or rejected.
- Agent tasks
- Realistic multi-step tasks with graders for enterprise agents.
- Redaction
- Counterparties, people and figures removed and human-checked.
Workflow record. Values are illustrative.
From scattered trail to record.
One purchase approval, through the five stages every workflow goes through before it ships.
License
Workflows, documents and approval trails come from companies under agreement, exported from the systems where the work actually happened.
- Who
- Companies
- Gate
- No agreement, no ingest
De-identify
People, vendors and account numbers are replaced with consistent tokens, so the same person stays the same person across the whole trail.
- Who
- Model + reviewer
- Gate
- Tokens consistent across the trail
Reconstruct
Emails, document versions, approvals and system events are put in order as one trail, so an agent sees how the work moved, not just the final file.
- Who
- Data engineering
- Gate
- Every decision linked to its trigger
Expert review
Operations and legal specialists label steps, outcomes and the policy behind each decision, including the rejections.
- Who
- Ops & legal experts
- Gate
- Every decision labelled
Package
Trails ship as event logs with their documents, or as agent task suites built from them.
- Who
- Delivery
- Gate
- Datasheet with every delivery
Built for frontier model tasks.
Scope, review criteria and delivery format are agreed per engagement, from the first sample to the final eval.
- Enterprise agents
- Agents that follow real approval chains and policies.
- Document understanding
- Contracts and forms with their revision history.
- Process mining
- How work flows, stalls and escalates across teams.
- Legal reasoning
- Clause review, redlines and decision rationale.
- Agent evaluation
- Graded multi-step tasks drawn from real operations.
What buyers ask us first.
Something else? Ask a specialist →
01Where does this data come from?
From operating companies that license their workflow, approval and document data to us under agreed terms.
02Which workflows are covered?
Eight workflow families, including procurement, finance, HR, legal, sales and support, each traced from intake through review and approval to archive.
03How is confidential information handled?
People, counterparties, prices and account details are redacted and human-checked. Handling, retention and access requirements are agreed during scoping.
04Can you build agent tasks from it?
Yes. We turn real workflows into multi-step tasks with graders and expert-reviewed reference solutions.
Same standard in every domain.
Conversational
Multi-turn conversations, ranked and reviewed for assistant training.
- 01 Dataset
- Multi-turn dialogue, intent-labelled
- 02 Environment
- Multi-turn evals, rubric-graded
Private code
Repositories, commits, pull requests and review history.
- 01 Dataset
- 3B+ lines with commit, PR and review history
- 02 Environment
- Agentic engineering tasks, engineer-graded
Healthcare
Clinical notes, records and imaging, de-identified and expert-reviewed.
- 01 Dataset
- 10M+ de-identified, patient-linked records
- 02 Environment
- Clinical case tasks, clinician-graded
CAD & 3D
Drawings, assemblies and meshes with validated geometry.
- 01 Dataset
- 5M+ typed, native-format CAD files
- 02 Environment
- Part classification and rebuild checks
Audio & speech
Synchronized audio with transcripts and event labels.
- 01 Dataset
- 5.7M+ transcribed, diarized assets
- 02 Environment
- Transcription evals per language and accent
Egocentric video
First-person footage with action and object labels for physical AI.
- 01 Dataset
- 200K+ hrs synced video, IMU, action labels
- 02 Environment
- Action and manipulation benchmarks
See real company operations data first.
A representative sample under NDA, in the format your pipeline reads. Run it through your own evals first.
- Sample records, as delivered
- Datasheet and schema
- Review criteria
- JSONL · Parquet · PDF + text
{"trail":"legal_08","company":"[ORG_17]",
"doc":"msa_rev03","clauses":42,
"events":[{"stage":"intake","role":"ops"},
{"stage":"review","role":"counsel","edits":6},
{"stage":"approval","role":"cfo"}],
"redaction":"verified","status":"documented"}