Published commitment · 3 of 4
AI-prepared, human-approved. Nothing sends without you.
The platform drafts. The human approves. There is no “set it and forget it” autopilot mode and we will not build one. Every chase email, every escalation, every tone change is shown to a named user before it sends.
What this means concretely
- The ML stack organises evidence, highlights what changed, and proposes verification questions, but the decision belongs to an accountable human.
- Batch review is allowed, but each resulting decision keeps a named accountable user attached and can be corrected before it is relied upon.
- The platform does not issue legal demands or make an automated credit decision. It presents source-linked evidence and options for human review.
- Every outbound carries a Right of Reply link routing to our rectification path, see the corrections policy.
Why we’re publishing this
Automated decisions can erode trust when users cannot see the evidence, the uncertainty, or the route to correct an error.
We do not build that. The platform is fast, opinionated, and well-targeted; it is also explicitly not autonomous. The product’s job is to make a human decision-maker more effective by assembling evidence, context, options, and monitoring signals.