IER-0

Support operations, governed.

Centralize support across every product without handing execution to the model. AI prepares the work; policy, approval, and audit keep control.

Your support, governed.

Inbound email, AI triage, human approval, outbound send, and audit become one durable operating loop.

Inbound control

Verified inbound messages, evidence retention, dedupe, and project routing before AI sees anything.

resend/inbound/launchkit
1verifysigned
2triagerisk
3draftregola
4approvalhuman

Policy-bound agents

AI classifies, summarizes, drafts, and proposes while policy and humans own the outcome.

Thinking under policy...
draft proposal

Approval queue

Every outbound reply waits for a human decision and an audit prewrite.

high riskBlocked
owner reviewqueued
executorlocked

Knowledge provenance

Project-scoped sources, freshness, trust state, and citations shape every draft.

source registrytrusted
source: refund-policy.md
scope: launchkit
trust: approved

Multi-project cockpit

One workspace for support identities, policies, customers, and queues across products.

customer.domain.com
project.domain.com
jane.domain.com

Phone support

Hosted phone numbers route callers into project-scoped voice agents with the same approval and audit model.

hosted numberReady
routeproject voice agent
recordcall history
billvoice minutes
Governed support operations

Scale the queue without letting the model own the outcome.

Tier0 keeps every reply inside a controlled path: route the message, inspect risk, cite approved knowledge, draft under policy, then wait for an operator before anything leaves the workspace.

Inbound control
Policy-bound agents
Approval queue

Protect customer trust across every product, policy, and support identity.

Review the control plane
Governed AI

AI prepares the work without taking ownership of the outcome.

Models classify, summarize, draft, and propose. Policies decide. Humans approve. Executors act only after recorded approval.

1PrepareClassify intent, summarize history, cite approved sources, and draft a reply.
2ReviewApply project policy, risk rules, and human approval before customer contact.
3ExecuteSend only after server-side checks, audit evidence, and idempotency are ready.
Operating signals
1Refund requests23.0%
2Account access18.5%
3Setup help14.9%
4Bug reports9.8%
5Security review5.7%