Categories, not inventory
The data we source
These are the categories we look for when we vet a company. What's available at any moment depends on the companies we're working with, so we source to your brief.
Gmail, Outlook / Microsoft 365
Multi-year threads with customers, vendors and teammates: quotes, scheduling, negotiation, escalation.
Useful for: Drafting, triage, long-context reasoning, multi-party coordination.
Team chat
Slack, Microsoft Teams
Channels and threads where decisions, incidents and handoffs happen. Taken from the workspace owner's native export.
Useful for: Coordination, incident response, decision-making under ambiguity.
Support tickets
Zendesk, Intercom, Freshdesk, Jira Service Management
Problem-to-resolution histories, escalations, macros and internal notes.
Useful for: Customer-support agents, troubleshooting, policy-following.
Docs, wikis and SOPs
Google Drive, SharePoint, Confluence, Notion
Procedures, checklists, specs, runbooks and the drafts behind them.
Useful for: Procedure-following, document understanding, grounding.
Code and engineering history
GitHub, GitLab, Bitbucket
Repositories with commits, pull requests, reviews and issues. Secrets are stripped; third-party code is flagged.
Useful for: Coding agents, code review, software-engineering tasks.
CRM and sales records
Salesforce, HubSpot
Pipelines, quotes, activity logs and notes, with contact details removed.
Useful for: Sales and operations workflows, structured-plus-text reasoning.
Calls, screen recordings and workflow traces
Gong, Zoom, Loom, task captures
Multi-step tasks as they were actually done, where consent and notice can be confirmed.
Useful for: Computer-use agents, RL environments, speech.
Industries
We look hardest in industries where work is process-heavy and poorly covered by public data:
- Agencies and professional services
- Managed service providers and IT services
- E-commerce
- Software
- Workflow- and CRM-heavy businesses: logistics, construction and engineering, legal
- Finance and healthcare administration, after special review (no patient or customer financial records)
We usually look for 20+ employees and multiple years of records. Restaurants and retail are usually not a fit.
What we document for each lot
- Source systems, export method and date range
- Company size band, industry and status (active, winding down, closed)
- Approximate volumes per category
- What was excluded, and why
- De-identification method and who performed it
- Known restrictions on use or onward transfer
Need something specific?
Tell us the categories, industries and era. We'll tell you what we can source.