We use essential cookies so the site works. With your consent, we also use analytics tools (PostHog and Microsoft Clarity) to understand traffic and improve the site. “Accept all” enables analytics; “Necessary only” keeps only essential cookies. See our Privacy Policy for details.
SALESKEYS GOVERNANCE
SalesKeys adds approvals, evidence, replay, policies, and security to the flow of work—and governs Vera’s data judgments, verification, and exports—so teams can scale automation without giving up control.
Most AI systems can do something. Few can prove what happened, explain why it happened, and let you intervene before risk spreads. SalesKeys Governance is designed to make workers usable in real business environments: approvals where it matters, evidence when you need answers, policies when boundaries must hold, and auditability when teams need to prove what changed.
Route high-risk actions into approval flows with escalation, delegation, and SLA control.
Search evidence, replay decisions and runs, and understand why a worker acted the way it did.
Set rules for content, tools, writebacks, external data movement, and risk thresholds.
Use RBAC, SSO, session controls, retention policies, and audit logs to maintain oversight.
VERA GOVERNANCE CASE
From whether data can be treated as fact, to whether results may be exported, to how verification status changes, every step is constrained by policy, evidence, and responsibility.
Case: Northstar Automation buyer research delivery
Task VR-2048 · Company research and decision-maker verification
Vera finds conflicting employee counts on the company website and a third-party database, retaining both sources and lowering confidence.
Applied policy
Conflicting facts cannot overwrite automatically; confidence below 85% is marked for review.
The research contains contact emails and source evidence, so role, purpose, and destination region are checked before export.
Applied policy
Cross-region exports containing personal data require Sales Operations approval.
When email status changes from valid to unknown, the system retains the prior value, reason, and verification service response.
Applied policy
Verification changes require evidence, a timestamp, and the responsible actor.
Uncertain data is labeled, sensitive export is gated, and verification changes retain complete evidence.
Teams can replay what Vera saw, why it reached a judgment, which policy applied, and who approved the final action.
Approvals should not live as isolated prompts inside each feature. SalesKeys provides a unified approval layer across workers, actions, and workflows. Your team can apply one set of rules to high-risk messages, writebacks, field changes, escalations, or sensitive actions without rebuilding the control flow each time.
| Action | Worker | Risk | Status | SLA |
|---|---|---|---|---|
| Send pricing proposal | Clara | High | Pending | 2h |
| CRM identity writeback | Victor | Medium | Approved | now |
| Escalate to manager | Victor | High | Pending | 4h |
| External data export | System | Critical | Blocked | — |
When a team asks "what happened?", basic logs are not enough. SalesKeys provides evidence search, conversation replay, run replay, decision explanation, responsibility tracing, and case-ready audit views.
Decision: Clara chose follow-up template B
Action: Email sent to sarah[at]meridian.sg
Gate: Human approval requested for pricing
Approval: Manager approved pricing proposal
Writeback: CRM stage updated to Proposal
Policies in SalesKeys are not an afterthought. They define what workers are allowed to say, write back, or send out, and under what conditions human review is required. This gives teams a unified way to apply content fences, tool allowlists, writeback policies, data egress controls, and high-risk thresholds across the whole system.
