The Swampfox Zorro Speech Platform supports a full range of Speech Customer Engagement applications within ICX — from constrained directed dialog applications to open-ended natural language (“How may I help you?”) prompts and supports traditional Speech Interactive Voice Response (IVR), conversational Intelligent Virtual Assistant experiences, and can even support Agentic style applications.

Modernizes the caller experience from rigid menus to natural conversation, increasing automation and customer satisfaction across interaction types.

Directed Dialog: Menu-style prompts with constrained responses capture the customer’s chosen intent (“Billing”) or provided information (“$23.15”, “July 5th”)

Open Intent Gathering: Open “How may I help you?” prompts let callers speak freely; Zorro classifies the free-form utterance for ICX to resolve the customer’s stated need or problem.

Intelligent Virtual Assistant: Powers Intelligent Virtual Assistant (IVA) conversational self-service applications to let customers quickly solve their own needs without having to wait for human assistance.

Zorro applies Natural Language Understanding (NLU) to transcribed spoken sentences to determine the customer’s intention, enabling ICX to resolve needs regardless of the kind of application (Directed Dialog Intent Gathering, “How may I help you?”, IVA, IVR or Agentic.)

Scoped Intent Recognition: Zorro’s models can be scoped to Enterprise appropriate, process appropriate and end-user appropriate domains to constrain candidate intents and confidence to improve understanding over large open-ended models.

Context and Session Support: Intent recognition improved further through session, context and language support. 

Zorro extracts key named values from speech, converting free-form utterances into structured data that ICX Interaction Flows can act on.

Captures the specific data a customer provides without rigid prompting — shortening conversations and improving data quality for downstream systems.

System and Customer Entities: AI LLM pre-trained system entities recognize complex concepts (like People’s names, Dates, or currency) and returns those values to the application without additional configuration, while customer and application specific entities can be created within the administrative interface to capture Enterprise or Process oriented values (Pharmaceutical names, Process Steps like Claim Numbers or Account IDs).

Rich Value Types: Extracts values such as numbers, digit strings, currency, proper names, titles, cities, and zip codes.

Menu Value Capture: Recognizes key values within directed menus, so flows resolve requests without requiring exact phrasing, providing an easy bridge from traditional grammar-based applications to Natural language-model based solutions.

Zorro delivers human-like synthesized speech with distinct gendered and personalities for both English and Spanish callers.

Delivers a consistent, branded voice experience while removing the cost and delay of studio voice recording.

Multiple Voices per Language: English: Nick (male), Vixey (female); Spanish: Diego (male), Carmen (female).

SSML Control: SSML support adjusts things like speed (prosody), pronunciation and pacing for natural, intelligible delivery.

Zorro 1.0 fully supports US English and Latin American Spanish within a consistent customer experience, serving the customer base from a single logical application.

Reaches more customers in their preferred language without standing up separate systems — broadening self-service eligibility and reducing language-based transfers.

Multilingual Coverage: English and Spanish transcription, Natural Language Understanding, multilingual entity extraction (entity values can be distinct between languages), and Text-to-Speech.

In-Session Switching: Speech-to-Text and NLU languages can be programmatically switched within a session.

Per-Model Language: Each NLU model is language-assigned, so English and Spanish intent and entity models are maintained independently.

Zorro converts spoken customer audio into accurate text — streaming in real time for live interactions and processing recorded files in batch for analysis and agent summaries — accepting file-based and streamed audio from ICX and integrating natively media servers for in-call recognition.

Turns every spoken interaction into structured, actionable text — the foundation for automation, analytics, and reduced reliance on agents for routine requests.

Streaming and Batch: Real-time transcription of live calls plus file and batch processing for offline workloads.

Barge-In Support: Voice Activity Detection halts prompting when a caller speaks, recognizing the utterance for a more natural conversation.

Domain Hints: Context and domain hints improve recognition accuracy for customer- or industry-specific use cases — e.g., helping transcription resolve phonetic similarity between “fraud,” “frog,” and “broad” in a financial-services context.

Zorro integrates tightly with Swampfox Intelligent Customer Experience (ICX) for management, Interaction Support, and generation of audio files. It also connects to Media Servers (Ex. Avaya Experience Portal, etc.) for in-call recognition and TTS using industry-standard protocols.

Extends existing Media Server and ICX investments with modern speech AI — no rip-and-replace, faster time-to-value, and unified management.

Media Server: MRCP v1/v2 with RTP for speech recognition and TTS using the standard speech resource model.

ICX Native Integration: REST for Speech-to-Text, NLU, and Text-to-Speech, with components initiated as needed.

Managed from Within ICX: Manages model creation, training, deletion, updates, and versioning within the ICX roles-based access control framework.

Zorro speech capabilities may be hosted on customer premises or in the cloud and hosted data centers.

Lets organizations meet data-residency and security mandates while controlling cost — keeping sensitive audio in-house or to the cloud as demand dictates.

On-Premises: Runs entirely within a customer data center (GPU required), keeping audio and logs inside the secure perimeter.

Cloud and Hosted: Swampfox Transparent Hosting and Managed Services provide scalable, redundant and expertly managed services.

Zorro is built for resilient, high-volume operation, scaling out horizontally and using standard media server and SIP mechanisms for high-availability/disaster recovery (HA/DR) within and across data centers.

Protects customer-facing availability during spikes and outages, sustaining service levels without over-provisioning fixed capacity.

Media-Service/SIP-Based HA/Failover: Uses the standard speech resource model; Media Server load-balances across Zorro servers and continues if one fails.

Scale-Out Architecture: A load balancer fronts the service endpoints to target 10,000+ concurrent sessions.

Capacity at Scale: Through ICX, supports an aggregate of up to 2,000 ASR and 2,000 TTS sessions, scaling via uniform resource servers.

Zorro can redact information such as personally identifiable (PII) and payment card (PCI) information from its logs, keeping sensitive customer data out of operational records.

Reduces compliance and breach exposure by ensuring regulated data never lands in logs — simplifying audits and supporting privacy obligations.

Configuration-Driven Redaction: Based on configuration, Zorro hides regulatory protected data so sensitive values are not exposed in logs.

Centralized Logging: Each server publishes NLU and TTS logs to dedicated, server-prefixed queues forwarded to ICX for consistent policy.

Federal Posture: Runs with FIPS 140-3 support and secure protocols across managed, customer-managed, and hybrid environments.

The Zorro Management Administrative Console, accessed from ICX Utilities, lets administrators build and manage intent models, define entities, train, and validate behavior — all within a unified administrative and management interface through ICX.

Puts model lifecycle control in the hands of the team — faster tuning and lower change cost.

Model and Intent Building: Create models and define intents by phrases with bulk support.

Entity Definition: Various common system entities available out of box or author custom entities with an in-console test field

Training and Online Validation: Train and validate with a built-in test interface; update and import/export models.

Zorro exposes its capabilities through a REST API supporting exploration and testing.

Accelerates integration for developers and partners — self-documenting APIs and a built-in test harness shorten build cycles and reduce support overhead.

REST: REST HTTP streaming with session context serves, real-time, and batch use cases alongside MRCP.

Swagger and Test Harness: The Zorro server publishes interactive Swagger (OpenAPI) docs with an in-browser test harness for the REST APIs.

Zorro 1.0 is compatible and integrated with Swampfox ICX 4.2 or newer. All deployments and integrations require Swampfox Professional Services.