What data AI labs license from companies

The general shape of it

• Documentation. Internal documentation, SOPs, playbooks, knowledge bases, process documentation, templates, and operational frameworks.
• Workflow records. Project management workflows, task execution histories, CRM and workflow metadata, business process artifacts.
• Communication. Internal communications and messaging where work is coordinated, plus email in some programs.
• Systems of record. Documents, CRM records, code, and finance or HR records from the tools a company chooses to connect.

By function

Every partner frames this differently, but the underlying work is the same. These are the functions that come up most:

• Software and engineering. Code review, debugging, and deployment across real repositories and tools.
• Finance and accounting. Account reconciliation, invoice processing, month-end close.
• Customer support. Ticket diagnosis and resolution across multiple systems.
• Sales and CRM. Deal progression from outreach through close.
• Legal and contracts. Agreement review, redlining, negotiation.
• Operations and logistics. Order processing, exception handling, fulfillment.

Frontline and deskless work is its own category

If your workforce is on a factory floor, a ward, a warehouse, or an airport apron, the artifacts look nothing like the list above. micro1 runs a separate frontline operations program for exactly this, and asks for shift huddle notes, handwritten logbooks, whiteboard photos, sticky notes, Excel task trackers, and shift handoff records from Microsoft Teams.

That program is sized on users rather than employees, at 7,000 to 20,000 Microsoft Teams users, and it leans towards companies headquartered outside the United States. If that describes you, the general 30 to 200 employee criteria are the wrong ones to measure yourself against.

What stays out of scope

Personal data is the thing partners work hardest to remove, because a lab does not want it either. Mercor states its pipeline masks 60+ types of personal, health, and business identifiers before delivery, describes the masking as irreversible, and says no lab buyer or third party sees raw content. Those are their statements about their own process, not something we have audited.

The more important point is procedural: you choose what is in scope. Connecting a tool is a decision you make per tool, and the scope is written into the agreement before anything moves. If a system is off limits, it stays off limits.

Where we sit in this

Nowhere near your data. We never collect, store, view, or process it. We take a short profile of your company, work out which partners fit, and make the introduction. Everything about scope, anonymization, and terms happens between you and the partner.