ORM Agent
⬜ Proposed — not built. Tier 1 — buildable now
Back to the agent roster · Marketing OS
What it would do
Reads brand mentions and judges which are real reputational issues versus noise, assigns severity, and drafts a response. Never posts.
Why an agent
Sentiment APIs are crude. "IIFL share price falls" and "IIFL mis-sold me a policy" score alike and need completely different handling.
What it needs first
A mention feed — Meltwater or LocoBuzz. That integration is the whole cost; the agent shape is already proven.
Roadmap source
Deliverable ⑥ (page 5); ORM → Meltwater | LocoBuzz (page 3); ORM → own (page 4) — see roadmap_ref_extracted.md.
Design notes and open questions
- Architecturally this is the Compliance Agent in a different domain — findings, severity, evidence, cited rule. Reuse the verdict shape.
- Responding publicly is a human action, same rule as Send to CMS. The agent drafts; it does not reply.
- Needs an escalation matrix and legal sign-off path for public replies.
Subagents
- Classifier — service complaint, factual error, market commentary, or spam
- Severity scorer — what actually needs a human today
- Response drafter — writes a reply; never posts it
- Escalation router — sends the ones that need legal or a branch to the right desk
Stub spec. Expand to the full overview.md + technical.md split (as the
three built agents use) when this moves from proposed to scheduled.