Privacy commitments

1. SteadMark reads the team channels you connect, working from conversations already happening.

SteadMark connects only to the Slack channels you authorize, and it does not read channels you have not connected. Direct messages get the same treatment as channels. Patterns merge into aggregates without attribution.

What enters our system from a channel is a pattern label, a confidence number, a hashed author ID, and a timestamp. We keep patterns, not an archive of your messages, and we minimize what we store.

2. Every recommendation aims at your next conversation, framed as mode-recovery, never as evaluating the person.

Every recommendation is built for the manager's next conversation. The output names what to say and why it lands. It does not score the person it describes, it does not rank them against peers, and it does not produce a verdict about who they are. Recommendations frame as mode-recovery. The product describes what helps someone re-engage, not what is wrong with them. Every recommendation passes the read-aloud test before it ships: if it would embarrass the person it describes when read aloud to them, it does not ship.

The same posture extends to attribution. Author identities are hashed before any pattern is stored, so distinct-participant counts work without identity. Named attribution is suppressed on every manager-facing and executive surface. They see the pattern label, the source, and the date. They do not see who. The one exception is a manager-initiated conversation prep tool that names the team member the manager has chosen to meet with, because that surface exists only to help them say it well.

3. We do not rank your team members, and managers never see them labeled by personality type, performance, or risk.

No personality model lives in the schema. There is no MBTI, Big Five, DISC, or equivalent typing. No employee performance score exists. No ranking of team members against each other is produced anywhere in the product.

Per-employee mode classification (engaged or self-preservation), pattern type, and confidence scores are computed inside the system and surfaced only on the change-function dashboard behind a role gate. The line manager surface never carries those labels.

SteadMark does maintain a per-employee adoption-phase value (such as early supporter or reluctant adopter) and a per-employee compliance-status value (such as in progress or not yet compliant) for the rollout the manager is leading. These describe the person's relationship to the specific change, not their character or their job performance. The manager who owns the rollout chooses or accepts these values. They are visible to that manager and, behind a role gate, to a change-function lead. They are never visible to other team members or to executive readers.

4. One recommendation per workday, Monday through Friday. Reply to adjust it.

The daily plan arrives by email once per workday, Monday through Friday. Choosing Not today moves the same recommendation to the next workday rather than discarding it. The email is replyable, and a reply shapes the next workday's plan. Each recommendation also lives at a link inside the product where the manager can record what they did with it. There is no weekend delivery and no second recommendation pushed later in the day.

Workplace safety

Detecting and escalating workplace safety concerns (harassment, threats, crisis situations) is outside SteadMark's current scope. If you encounter something urgent, follow your organization's established reporting channels.