Guides

GTM & outbound agents

Enrich lead → draft email → CRM upsert → schedule follow-up. agent_id: gtm. Cost routing and cache dominate this workload.

Scenario

Turning off cheap-path routing 3×'d cost per lead overnight. RouteIQ model mix plus LensAI cost-by-intent showed the flip immediately.

What Inferix gives you

RouteIQ cheap path + semantic cache for pricing_intro / hours-like FAQ copy
Budget denies per tenant campaign (R11/R12)
LensAI cost.per_successful_task (meeting booked, not just email sent)
TraceForge for enrich → draft → send tool order

Tools & intents

Tools	Intents
enrich.company, enrich.person, crm.upsert_lead, email.draft, email.send, calendar.book	cold_outbound, warm_followup, event_invite, reengagement, pricing_intro, partner_intro

Setup

routing:
  rules:
    - name: gtm-boilerplate
      when: { agent: gtm, intent: [warm_followup, reengagement] }
      model: owned/slm-support
      cache: { mode: semantic, ttl_s: 3600, key: [tenant_id, intent, prompt_hash] }
    - name: gtm-cold
      when: { agent: gtm, intent: cold_outbound }
      model: owned/general-llm
  budgets:
    tenants:
      sales-east: { weekly_usd: 200, on_deny: cache_only }

Policies: Routing.

Watch & failures

Over-escalation (R06) when classifier dies — everything on strong
Cache stampede (R15) after TTL — singleflight
Cross-tenant cache bleed (R18) — composite keys
force_expensive_model chaos — LensAI cost spike

Business success

Count CRM upsert + valid send, not HTTP 200 on email.draft alone (T12).

STT / LLM / TTS latency and failover.

Voice agents →