Abstract
Autonomous growth systems risk brand damage when they optimize for volume without evidence. This protocol defines a safer loop: sense market signals, formulate hypotheses, draft grounded assets, route by recipient risk, qualify replies, and promote only outcome-backed learning.
Protocol
| Phase | Required record |
|---|---|
| Signal | Source, timestamp, segment, relevance |
| Hypothesis | Buyer pain, offer, disconfirming evidence |
| Draft | Claim support, voice boundary, forbidden promises |
| Route | Send, queue, suppress, or escalate |
| Qualify | Intent score and rationale |
| Learn | Outcome-backed rule update or rejection |
Experiment
Compare manual founder growth, generic AI-assisted drafting, and Armalo Agentic OS growth missions. Primary metric: qualified conversations per founder hour. Guardrail metrics: unsupported-claim rate, wrong-recipient rate, escalation precision, and lead-to-mission trace completeness.