Platform
One engine. Four jobs, each one specific.
Not “AI-powered”. Here is the verb and the artifact for each product, and the rules every one of them runs under.
What it does, product by product
- Spine
- Reads the conversation and produces a scored lead card with the reasons written out. Classifies what a reply means. Runs a nightly next-best-action pass across the pipeline, queued for approval.
- Relay
- Reads each incoming message and routes it to an owner, with the customer's history across channels attached.
- Scope
- Reads the backlog and the team's real availability and drafts the sprint, flagging risk with the evidence attached.
- Crew
- Reads the day's consented captures and writes a plain-English activity report — the same report the employee sees.
The rules it runs under
- Every action is logged — a full audit trail, per workspace and per person.
- Every action is reversible.
- Anything outside the rules you set is held for your approval rather than executed.
- A playbook earns autopilot; it does not start with it.
- One workspace-level switch stops every automation at once.
- With no model configured it goes inert rather than guessing.
- It works on your records and never learns from them for anyone else.
Why the guardrails come first
The reason most teams leave automation switched off is that they cannot see what it did, or undo it. That is the problem these rules exist to solve — they are defaults, not settings buried in an admin panel.
Still need a person?
Write to [email protected]. We are small enough that someone reads it.