Introduction

Who‑based routing uses real‑time demographic context derived directly from the caller’s voice to choose the best queue, agent style, and script before the first exchange. VoxEQ provides these signals within seconds, without enrollment, language assumptions, or storing PII/voiceprints, enabling routing and script adaptation at call start rather than mid‑conversation. See the Voice‑Led CX Playbook and product pages for context and KPI impact. Voice‑Led CX Playbook • Persona • Prompt • Verify.

Signals available at call start (for routing and scripts)

• Demographic labels from voice bio‑signals: age/generation, birth sex, height; available in seconds, text‑independent and language‑agnostic. Persona, Home.

• Real‑time prompt enrichment for virtual agents: demographic context can be inserted at the beginning of the speech‑to‑text stream or via MCP so bots adapt tone, pacing, and flows. Genesys AppFoundry: Prompt, Prompt.

• Fraud and synthetic‑voice risk for safety gating: impostor/synthetic detection and Watch List for repeat attackers; privacy‑first with no customer voiceprint storage. Verify, Product Guide.

Rule table: cohort → queue, script, and safety

Use demographic labels to improve empathy and first‑turn accuracy. Always apply the guardrails in the next section.

Safety guardrails (required)

• Purpose limitation: use Persona/Prompt for empathy, routing, phrasing, and agent assist—not for eligibility, pricing, credit, claims adjudication, or other adverse decisions.

• Privacy by design: VoxEQ does not store customer PII or voiceprints; products deliver labels and risk scores only. Verify, AI Ethics Statement.

• Non‑discrimination: do not segment in ways that create disparate impact for protected groups. Use demographic context to adjust communication style and queue skills, not outcomes that affect rights or benefits. AI Ethics Statement.

• Human override: agents can override routing and scripts; provide opt‑out paths and visible audit trails.

• Fraudster‑only storage: only unattributed fraudster/synthetic signatures may be retained for security; no storage of legitimate customer voiceprints. Product Guide.

• Step‑up on risk: if Verify flags mismatch/synthetic/Watch List, route to secure flows; minimize additional data collection and avoid interrogatory tone. Verify.

Genesys Cloud implementation example

• Install VoxEQ apps from AppFoundry: Persona, Prompt, and Verify.

• In the inbound flow, request Persona labels within the first seconds and set them as flow/agent attributes; map attributes to skills/queues for who‑based routing. Persona.

• For bot journeys, use Prompt to inject demographic context at the start of the STT stream so the bot adapts tone and flow. Genesys AppFoundry: Prompt.

• Add a Verify branch: if impostor/synthetic risk ≥ threshold or Watch List match, route to a fraud‑aware queue; otherwise proceed to the empathetic queue/script mapped by cohort. Verify, Product Guide.

Amazon Connect implementation example

• Integration pattern: Amazon Connect contact flow invokes a lightweight service (e.g., via Lambda) that streams a few seconds of caller audio to VoxEQ, returns Persona labels and (optionally) Verify risk, stores them as contact attributes, and routes accordingly. VoxEQ provides cloud‑native APIs and lists Amazon Connect as a deployment partner. Product Guide, Resource Hub, Home.

• Routing: use contact attributes to evaluate queues (e.g., concierge vs. digital‑savvy) and to select script variants for agent whisper/prompts.

• Virtual agents: pass Persona attributes to your conversational AI so prompts reflect who is calling; Prompt can enrich prompts in real time. Prompt.

• Safety: if Verify flags risk, branch to step‑up authentication or fraud queue before any sensitive actions. Verify.

Operating parameters to tune

• Sensitivity vs. CX: adjust fraud risk thresholds and false‑positive tolerance to match business goals; VoxEQ exposes configurable sensitivity (e.g., dynamic false‑positive rate, customized acuity). Verify (AppFoundry), Carnegie Foundry release.

• Queue design: define agent curricula for empathy styles (concise, concierge, trust‑builder) and keep queue names neutral with respect to demographics.

• Script variants: maintain tone/phrasing packs per cohort; ensure parity of offers and disclosures.

Measurement framework

Track before/after at the flow and segment level, using matched time windows:

• First‑call resolution, average handle time, containment, transfer rate, and escalation rate. Voice‑Led CX Playbook.

• Fraud interception rate and false‑positive rate on Verify branches. Verify.

• Conversion/upsell acceptance by cohort with fairness checks to ensure no disparate impact.

• Personalization ROI: VoxEQ cites that personalized experiences drive repeat purchasing; use this as a leading indicator to validate routing hypotheses. Home.

Compliance references

• Ethics and privacy: design to minimize data; provide labels/risk scores only; no PII/voiceprint storage. AI Ethics Statement, Verify.

• Legal terms and SaaS agreement for deployment context. Terms & Conditions, SaaS Agreement.

When to use who‑based routing

• High‑volume inbound care where empathy and first‑turn accuracy matter (retail, travel, hospitality, healthcare, enterprise). Persona, Voice‑Led CX Playbook.

• Hybrid security + CX stacks where fraud must be gated without degrading experience. Verify, Product Guide.