Organizational Data

Agents fail at the hand-offs.
We capture the hand-offs.

Cross-system threads.One task, every tool it touched, in order.

SourcesEmail, chat, ERP, CRM, ticketing, docs
LinkingActor, role and thread IDs across systems
Outcome labelsResolved, escalated, approved, time-to-close
PrivacyPII redacted, consented, auditable

What's in the dataset

Structure

Org graphs with real roles

Reporting lines, teams, permissions and ownership, so models learn who to ask, who can approve and who gets blocked.

Workflows

End-to-end process traces

Procurement, support triage, finance close, hiring and incident response, captured as step-by-step sequences with timestamps.

Documents

Policies, SOPs and the docs people cite

The internal knowledge that decisions reference, linked to the threads where it was used.

Annotations

Expert labels on intent and quality

Operators mark what went right, what went wrong and what a better next action would have been.

Sample record

One procurement thread, simplified. Delivered as JSONL.

thread.jsonl{
  "thread_id": "PO-88213",
  "org": "mfg-co-014",  // pseudonymous
  "systems": ["email", "erp", "chat"],
  "actors": [
    { "id": "a17", "role": "buyer", "dept": "procurement" },
    { "id": "a03", "role": "controller", "dept": "finance" }
  ],
  "events": [
    { "t": 0,     "sys": "email", "actor": "a22", "type": "request" },
    { "t": 5400,  "sys": "erp",   "actor": "a17", "type": "create_po" },
    { "t": 91800, "sys": "email", "actor": "a03", "type": "approve" }
  ],
  "outcome": { "status": "approved", "hours": 25.5 },
  "pii_redacted": true
}

Fields

Built for SFT, RL and evals

Every thread carries its actors, systems, ordered events and outcome, so you can turn it into demonstrations, reward signals or held-out test cases.

FormatJSONL + attachments
GranularityEvent-level, timestamped
DeliveryS3 / GCS, versioned

How we source it

Consent and provenance first, at every step.

Partner companies opt in

Organizations license workflow data under contract, with scope agreed system by system.

Redact and pseudonymize

Automated PII removal followed by human review. Names, emails and identifiers are replaced consistently.

Link, label and ship

Threads are stitched across systems, labelled by operators and delivered with a provenance manifest.

Which industries do you cover?

Manufacturing, logistics, professional services and SaaS today. Tell us the domain you need and we'll scope it.

Can we get exclusive data?

Yes. Custom collections can be licensed exclusively for a fixed period.

How is privacy handled?

Data is collected under contract with the source organization, then redacted and pseudonymized. Every delivery ships with a provenance manifest you can audit.

Need a specific workflow?

Scope a Dataset

Agents fail at the hand-offs.We capture the hand-offs.