20xwork reads how each employee does the steps of their job in the work tools your organisation connects, and turns that into guidance at the moment a step starts. To do that we process three kinds of data.
Work signals. Activity records from connected tools: for example, opportunity updates in Salesforce, AI assistant usage logs your organisation exports to us, or process definitions from a process-mining tool. From these we derive signals such as "next step names an owner and a date" or "the record was updated the same day". We store the derived signal and a reference to the source record. We do not store the full body of emails, documents or chat messages.
Recommendations and outcomes. The card shown to a person, the check run before a write, and whether the person kept, adapted or dismissed it. These are stored so the person's profile and your organisation's return can be measured.
Account data. Name, work email, role, team and manager, provided by your organisation's admin or your identity provider, to know who is who.
Transient request data. When a person starts a task in Claude, Claude Code, Copilot or another assistant with 20xwork connected, the task text and the fields of the write about to be made are sent to our service to build the card and run the check. They are processed in memory and are not retained after the response.
We collect no data from personal accounts, personal devices or tools your organisation has not connected. See What 20xwork reads, and what it never reads.