Rasa

What Rasa does

Rasa is an enterprise platform for building and operating AI agents across chat and voice channels. It extends LLMs with business logic to create reliable, compliant AI agents across millions of conversations, giving teams full control over behavior and performance. Organizations use Rasa to orchestrate complex, multi-step conversations, integrate AI agents with core business systems, and run them reliably in production environments. Rasa supports long-running interactions that span channels and systems, enabling consistent agent behavior as conversations evolve, including escalation, recovery, and coordination across workflows. Designed for large-scale deployments, Rasa supports secure operation across cloud and on-prem environments and integrates with existing data, APIs, and tooling. Enterprises adopt Rasa for control, scalability, and governance for AI agents supporting critical customer and employee interactions. The platform helps teams build, test, deploy, and analyze AI agents at scale, adapting and growing with their needs.

Products

• Rasa PlatformOne platform for automating conversational experiences, combining pro-code infrastructure and a no-code UI for conversational AI teams. It includes Rasa Pro and Rasa Studio.
• Rasa ProA pro-code generative AI native conversational AI framework for developers to flexibly build, integrate, monitor, and deploy AI Assistants. It includes CALM dialogue understanding and management, language-agnostic NLU, enterprise search, contextual response rephraser, custom actions server (Rasa SDK), channel connectors, Kubernetes deployment support, end-to-end testing, PII data management, LLM fine-tuning recipe, multi-LLM management, dependency vulnerability protection, secrets management, observability (OpenTelemetry), multi-node concurrency (Redis), conversational data pipeline (Kafka), and IVR Connector.
• Rasa StudioA no-code user interface for business users to build, analyze, and optimize AI Assistants. It includes everything in Rasa Pro, plus an AI assistant flow builder, testing panel, content and response management system, language and channel specific answer management, change observation and rollback, UI-based saving, conversation analysis view, open prompt editing, Single-Sign-On (SSO), and Role-based Access Control (RBAC).
• Rasa Developer EditionA free Rasa license for developers and teams starting an AI assistant project, usable locally or in production, with up to 1000 external or 100 internal conversations/month and community support.
• CALM (Conversational AI with Language Models)Rasa's architecture that separates language from execution, providing fluent conversations and predictable results in complex enterprise environments. It is a dialogue understanding and management system.
• Rasa NLUAn open-source natural language understanding component that gives developers control over language understanding.
• Rasa X/EnterprisePrevious commercial offerings that are part of the Rasa Platform.

Key capabilities

• Patented dialogue management system for enterprise AI agents
• Orchestrates autonomous reasoning, guided workflows, and shared conversational memory
• On-prem / private cloud deployment options, including air-gapped deployments
• Full control over data, security, and model providers
• Compliance and governance for regulated industries
• Extends LLMs with business logic for reliable AI agents
• Full control over behavior and performance of AI agents
• Supports complex, multi-step conversations
• Integrates AI agents with core business systems
• Reliable operation in production environments
• Supports long-running interactions across channels and systems
• Consistent agent behavior as conversations evolve (escalation, recovery, coordination)
• Designed for large-scale deployments
• Secure operation across cloud and on-prem environments
• Integrates with existing data, APIs, and tooling
• Deterministic flows that are fully inspectable and easy to debug
• Control over when and how LLMs are used to reduce hallucinations and improve reliability
• Built-in tools for re-asking, clarification, and topic recovery
• Flexible to fit unique workflows and scalable
• Hooks, overrides, and full control to shape every layer of the agent
• Deployment on customer's terms (cloud, on-prem, or fully managed)
• Production-grade automation with reusable logic
• Built-in conversation repair for topic shifts, clarifications, and user backtracking
• Centralized content management with templated responses
• Low-code Studio for building, testing, and optimizing without engineering bottlenecks
• Prebuilt starter packs (Example Flows, responses, integrations)
• Multi-LLM routing to distribute tasks, reduce latency, and balance costs
• Conversation Analytics Pipeline to track flow performance, drop-off points, and success rates
• Voice support out of the box with turn-taking, repetition, and timeout behaviors
• CALM (Conversational AI with Language Models) dialogue understanding and management
• Language-agnostic natural language understanding
• Enterprise search
• Contextual response rephraser
• Custom actions server (Rasa SDK)
• Channel connectors (REST and WebSocket)
• Kubernetes deployment support through Helm
• End-to-end testing
• PII data management
• LLM fine-tuning recipe
• Multi-LLM management
• Dependency vulnerability protection
• Secrets management powered by Vault
• Observability (OpenTelemetry)
• Multi-node concurrency (Redis)
• Conversational data pipeline (Kafka)
• IVR Connector to AudioCodes VoiceAI Connect (additional purchase)
• AI assistant flow builder (Studio)
• Try your assistant testing panel (Studio)
• Content and response management system (Studio)
• Manage language and channel specific answers (Studio)
• Observe and rollback changes (Studio)
• Save changes directly in the UI (Studio)
• Conversation view to analyze conversations (Studio)
• Open Prompt editing (Studio)
• Single-Sign-On (SSO) (Studio)
• Role-based Access Control (RBAC) (Studio)
• Self-Managed (On–prem or Private Cloud) deployment mode
• Managed Service deployment mode
• Connect to existing government systems (case management, benefits, citizen records, internal APIs)
• Orchestrate policy-driven workflows for eligibility checks, applications, scheduling, service requests
• Scale with modular service architecture (reuse across channels, programs, departments)
• Identity verification and secure interactions
• Multilingual & Accessible Service Delivery
• Multi-agent orchestration across services and departments
• Open framework and ownership (access and modify prompts, workflows, integrations)
• Full control, no black boxes (CALM keeps business logic separate from LLM)
• Fast, efficient, and built for scale (up to 4x faster, 80% cheaper than LLM-first frameworks)
• Customizable from top to bottom (connect to any API, database, internal tool, extend behavior)
• Connect to banking systems (core banking platforms, CRMs, payment systems, authentication, fraud tools, knowledge sources)
• Orchestrate governed banking workflows (account support, transactions, authentication, fraud handling, lending journeys)
• Deliver one continuous experience across channels (chat, voice, IVR, mobile apps, web)
• Connect to retail systems (commerce, service, loyalty, fulfillment)
• Orchestrate retail workflows (order management, returns, delivery issues, promotions, account support)
• Reuse what works and expand faster (turn proven workflows into reusable building blocks)

Use cases

• Building trustworthy AI agents for real-world use
• Customer support agent
• Internal helpdesk agent
• Search agent
• Voice agent
• Banking Agent (automate tasks related to personal banking and financial management)
• Monitor transactions to identify and prevent fraudulent activities (fraud_detection skill)
• Evaluate user's financial data to determine loan qualification (loan_eligibility_assessment skill)
• Help user research and analyze investment options (investment_research skill)
• Provide real-time updates and insights on user's investment holdings (investment_portfolio_tracking skill)
• Assist users in creating and managing monthly budgets (budget_planning skill)
• Automating conversational experiences
• Deploying conversational AI at scale
• Automating internal processes
• Orchestrating systems
• Customer experience that’s fast, intelligent, and built to scale
• New customer onboarding
• Feature discovery
• Loyalty program support
• Outage notifications & status
• Flight status & notifications
• Check-in & boarding pass
• Provider & facility information
• Symptom triage
• Automating citizen services
• Benefits processing
• Permitting, Licensing & Regulatory Workflows
• Public program workflows
• High-volume Citizen Service Automation (address changes, document reissuance, status inquiries)
• Program Enrollment and Benefits Processing (eligibility checks, applications, renewals, document submission)
• Proactive Communication and Service Continuity (timely updates, issue reporting)
• Customer support automation (resolve FAQs and repetitive issues)
• Technical support & troubleshooting (guide users through device/software issues)
• Billing & payments (automate questions, send reminders, accept payments)
• Subscription & account changes (handle account updates and plan changes)
• Order tracking & post-purchase support (answer 'where’s my order')
• Appointment rescheduling (patients reschedule/cancel)
• Claims status updates (customers check claim progress)
• First notice of loss (FNOL) (kick off claims process)
• Lead qualification (automate early funnel stages)
• Personalized product recommendations (based on user behavior/preferences)
• Cross-sell opportunities (surface relevant add-ons, upgrades)
• Quote & purchase assistance (guide buyers through complex purchase flows)
• Interactive demos & guided tours (prospects explore product/service)
• Conversational landing pages (turn static marketing pages into interactive experiences)
• Account servicing in banking
• Transactions in banking
• Fraud-related interactions in banking
• Lending journeys in banking
• Customer Account Support & Self-Service (balance checks, card activations, address updates, password resets)
• Secure Transactions & Payments (guide through payments, transfers, transaction inquiries)
• Authentication & Fraud-related workflows (identity verification, fraud scenarios, escalation)
• Proactive Service and Customer Engagement (reminders, alerts, next-best actions)
• Knowledge Search & Answer Resolution (answer questions about fees, policies, product terms)
• Loan & Credit Application Support (eligibility checks, disclosures, document collection, applications)
• Operational efficiency & inventory management (real-time stock visibility and alerts)
• Multi-agent orchestration (coordinate actions across systems and departments)
• Appointment scheduling & routing (healthcare, field service, customer teams)
• Account management (users make updates, check usage, access services)
• Subscription & billing management (automate routine changes)
• Trade desk support (financial analysts and trading teams access market data, models, decisions)
• Order Management & Post-Purchase Support (order status, cancellation, delivery changes, shipment issues)
• Returns, Exchanges and Refund Workflows (start returns, check eligibility, choose exchange options, track outcomes)
• Product Discovery & Shopping Guidance (guided product search, comparison, fit, personalized recommendations)
• Promotions & Checkout Recovery (understand offers, apply promo codes, resolve discount issues)
• Loyalty, Memberships and Repeat Purchase Support (points balances, reward eligibility, subscription changes)
• Delivery Issues and Service Recovery (respond to delivery failures, refund delays, replacement requests)
• Support Triage & Escalation (collect details, summarize issue, route to correct team with context)

AI approach

Rasa provides an enterprise platform for building and operating AI agents across chat and voice channels. It extends LLMs with business logic to create reliable AI agents, offering full control over behavior and performance. The platform uses its proprietary CALM (Conversational AI with Language Models) architecture, which separates language from execution to ensure predictable results. Rasa allows users to run their own LLMs, fine-tune them, and integrate with existing systems and data sources. It emphasizes deterministic logic, built-in recovery patterns, and patented dialogue management for complex, multi-step conversations.

Tech named: natural language processing, dialogue management, NLP, NLU, conversational AI, machine learning, deep learning, Conversational AI Platform, AI Assistants, Virtual Agents, AI Agent, Agentic AI, LLMs, CALM (Conversational AI with Language Models), ReAct, openai-gpt-4o, openai-gpt-5-1, Kubernetes, Helm, OpenTelemetry, Redis, Kafka, Vault

Industries served

• Software Development
• Banking
• Financial Services
• Customer Service
• Retail
• Government
• Public Sector
• Insurance
• Telecom
• Healthcare
• E-commerce
• Travel

What it says sets it apart

• Patented dialogue management system for enterprise AI agents
• Extends LLMs with business logic for compliant and reliable AI agents
• Full control over agent behavior and performance
• Orchestrates complex, multi-step conversations and integrates with core business systems
• Supports long-running interactions across channels and systems with consistent agent behavior
• Secure operation across cloud and on-prem environments, including air-gapped deployments
• Meets strict compliance and governance requirements
• Deterministic flows that are fully inspectable and easy to debug
• Control over LLM usage to reduce hallucinations and improve reliability
• Built-in tools for conversation repair (re-asking, clarification, topic recovery)
• Flexible and scalable to fit unique workflows
• Allows shaping every layer of the agent with hooks, overrides, and full control (no opinionated black box)
• Production-grade automation with reusable logic
• Centralized content management for templated and personalized responses
• Low-code Studio for business users alongside pro-code for developers
• Prebuilt starter packs for faster development
• Multi-LLM routing for efficiency and cost balancing
• Conversation Analytics Pipeline for performance tracking
• Out-of-the-box voice support
• CALM architecture separates language from execution for predictable results
• Open framework and ownership allowing access and modification of prompts, workflows, and integrations
• Agents respond up to 4x faster and are 80% cheaper to operate than LLM-first frameworks
• Customizable from top to bottom, connecting to any API, database, or internal tool
• Does not host customer data, systems, or applications; customers retain full data control
• Analyst-recommended for enterprise complexity, accuracy, customization, and control
• Recognized as a Leader in Voice AI by CMP Research Prism for Conversational IVR and voicebot innovation
• Recognized by Forrester for shaping the emerging market for enterprise customer experience automation
• Included in Gartner's reference architecture for conversational AI, highlighting leadership in orchestrating dialogue, context, and integrations
• Proven ability to scale across enterprise deployments and consistent execution in production environments
• Modern approach to conversational AI architecture with dependable performance
• Strong controls and extensible design
• Focus on high-value, high-stakes conversations in regulated environments