Try Managed Agents. Introducing Managed Agents. Today, Cartesia is launching Cartesia Managed Agents, its end-to-end voice agent platform for enterprise conversations. It pushes the frontier of speech-to-speech naturalness and accuracy, and ranks #1 on independent voice agent benchmarks at launch. Cartesia designed the harness and its voice models together, so listening, turn-taking, reasoning, tool-calling, and speech are tuned as one end-to-end system. Everything enterprise teams need to run agents in production comes on one platform: telephony, knowledge bases, tool integrations, and more. And Cartesia has tested it at scale - Managed Agents has already powered tens of millions of conversations in finance, retail, hospitality, and healthcare. Setting a new Pareto frontier. Cascaded pipelines are the dominant paradigm for enterprise tasks, but without careful co-design they can lose out on naturalness and conversational dynamics like turn-taking and interruptions. Today's speech-to-speech models sound natural, but they're hard to control and less reliable for enterprise-grade tasks. With Managed Agents, Cartesia co-designed its voice models and the harness to deliver the best conversational experience without sacrificing the intelligence and controllability that complex enterprise tasks require. On Artificial Analysis's Speech Agent Arena, where blind human raters compare agents on tasks like booking appointments and ordering takeout, Managed Agents ranks #1 of 30 agents with a preference Elo of 1115. On EVA-Bench, ServiceNow's open benchmark for enterprise voice agents in CSM, ITSM, and HR, Managed Agents posts the highest scores for both accuracy (87%) and experience (91%), setting a new Pareto frontier. When callers interrupt, pause, or change their minds, the agent handles it naturally: in its internal evaluations, callers reported no turn-taking issues in 85% of conversations with Managed Agents, compared with 67% for other agents. Designed for complex calls. Cartesia Managed Agents provides all the tools needed to build frontier voice agents in one end-to-end platform. Claims intake, patient scheduling, retail returns, IT helpdesk: enterprise calls rarely stay simple. One conversation can mean verifying a caller, checking three systems, and quoting policy, all without missing a beat. * Multilingual: Build agents in 44 languages. Agents can switch between English, Spanish, Hindi, French, and Japanese mid-conversation, with all 44 coming soon. * Tool calls: Integrate your backend APIs, CRMs, and systems of record, so your agent can look up records and take action mid-call. * Knowledge bases: Get precise answers from hundreds of documents. The agent's context is prefilled dynamically, so it responds without pausing to search. * Outbound campaigns: Schedule thousands of calls with outbound batches, each personalized to the person you're calling.

Cloudera expands AI ecosystem with Neo4j & Cartesia. Wed, 28th Oct 2026 (Today) Cloudera has expanded its Enterprise AI Ecosystem, adding Neo4j and Cartesia to partners including Mistral AI and VAST Data. The expansion adds technologies spanning knowledge graphs, AI models, voice tools and data infrastructure to Cloudera's Anywhere Cloud platform. The additions are intended to support customers running AI workloads across public cloud, private cloud and on-premises environments. Cloudera said the ecosystem is part of its push to help businesses use governed data for AI systems and software agents. Neo4j joins the programme with its knowledge graph technology. Cloudera said this will help customers connect business data with semantic context, allowing AI systems to query relationships across customers, products, transactions and business processes. The focus on connected data reflects a broader industry effort to improve AI reliability by grounding outputs in structured enterprise information. Cloudera is positioning knowledge graphs as a layer for reasoning over internal data, rather than simply retrieving information. "Our partnership with Cloudera helps organisations create a shared knowledge layer across enterprise data in public and private cloud environments, delivering the critical connection between relational data and deep semantic context. The Cloudera Open Data Lakehouse, in combination with Neo4j's Graph Intelligence Platform, allows organisations to query massive datasets, uncover hidden relationships, and power next-generation enterprise AI safely across any cloud or on-premises environment," said Sudhir Hasbe, President & Chief Product Officer, Neo4j. Cartesia brings speech generation and transcription models to the ecosystem. Cloudera said the tools are aimed at voice agents used in customer-controlled environments, particularly for organisations with strict data-handling and sovereignty requirements. Cartesia reports sub-90-millisecond speech-generation latency, while its Sonic-3.6 model ranked first on two Artificial Analysis speech arena leaderboards at launch, according to Cloudera. The company said those features are designed to support more natural voice interactions while keeping sensitive conversations under customer control. "Voice is becoming the primary interface for AI, and the enterprises that stand to gain the most are often those bound by the strictest data sovereignty and governance requirements," said Karan Goel, Chief Executive Officer and Co-Founder, Cartesia. "Together with Cloudera, we're enabling the world's most regulated organisations to deploy real-time voice agents at scale while retaining complete control over their data." Existing partners The expansion also builds on earlier partnerships with Mistral AI and VAST Data. Cloudera said Mistral remains central to its strategy for sovereign inference, with support for a range of models for different workloads. Mistral's portfolio, as outlined by Cloudera, includes the Large 4 model with 1.05 trillion parameters, Timestral forecasting models and Robostral for embodied navigation. Cloudera said model choice matters as companies try to match AI systems to different latency and compute demands. "As enterprises transition to production-ready agentic workflows, maintaining control over sensitive data and infrastructure is non-negotiable," said Marjorie Janiewicz, Chief Revenue Officer, Mistral. "Integrating Mistral's frontier models and products with Cloudera's hybrid data platform will allow organisations to build, train, and deploy custom intelligence on their own proprietary data, with full governance across any environment." VAST Data remains the infrastructure-focused partner in the group. Cloudera said the relationship is aimed at addressing data-delivery bottlenecks for GPU-based AI workloads, particularly where storage throughput and latency constrain model training, inference and analytics. Cloudera said VAST's storage architecture supports its containerised data services and can help improve GPU utilisation and data delivery across hybrid environments. That reflects a wider concern in AI deployments over the cost of underused accelerator hardware. "Most enterprises already have the data they need for AI. The challenge is unlocking the value in data to create a continuous pipeline of AI inference, fine-tuning, and data analysis to build the next generation of intelligent applications," said Jeff Denworth, Co-Founder, VAST Data. "Together, Cloudera and VAST are helping customers build AI factories that connect data, intelligence, and infrastructure into a single operational platform for AI across hybrid environments." Hybrid focus Cloudera is framing the expanded ecosystem around customer choice in deployment models and suppliers. It said its Anywhere Cloud approach is designed to provide a consistent operating model across public cloud, private cloud and on-premises environments, with attention to cost, performance and data sovereignty. Abhas Ricky, Chief Business Officer and GM, Applied AI, Cloudera, said the company sees competitive advantage shifting towards businesses that can use their own data more effectively in AI systems. "The next competitive advantage will come from how effectively enterprises turn their own knowledge into action," said Ricky. "Our ambition is to make every enterprise an AI factory, with the freedom to choose its models, retain control of its data, and improve the economics of every workload. Anywhere Cloud and our expanding ecosystem will connect the capabilities to make that possible-from the first insight to the millionth agentic action."

Raise Financial Services x Cartesia: voice AI for financial access in India. Raiseholding at Raise Financial Services (Raise), the Indian company behind stock trading and investing platform Dhan and AI-powered financial markets research assistant fuzz, among others, are developing multilingual voice experiences with Cartesia. Raiseholding has been investing heavily in building AI specifically for Indian financial markets. Cartesia provides the speech models that help Raiseholding extend that financial intelligence into natural, multilingual conversation. The partnership between Raise and Cartesia supports an ambition both companies share: making financial information more accessible through conversations in the languages people use naturally. "Cartesia gives us best-in-class voice models that integrate well with the real-time voice infrastructure we've built in-house," said Alok Pandey, co-founder and CTO, Raise Financial Services. "We're looking forward to scaling this partnership across our ecosystem, including Dhan; trader and investor educational platform Upsurge; insurance business Pluto, and other group entities." Building financial intelligence, choosing expertise in speech. Its approach to AI started with the requirements of Indian financial markets. Its team developed artham, a 7-billion-parameter language model trained on Indian financial and capital markets data. Built and hosted in India, artham is designed around the structure, vocabulary, and regulatory context of the domestic Indian market. Artham has been running inside fuzz, stock screener ScanX, and Dhan for over a year, summarizing market news, reading context into price moves, and answering user queries with live data drawn straight from official sources. Voice is the newest addition to that stack, and this is where Cartesia comes in. Cartesia's models currently power a portion of Dhan's voice bot interactions, including conversational KYC, while its human teams continue to lead on servicing, complex queries, and relationship support. The scale of multilingual demand is already clear: over 75% of these interactions happen in languages other than English, with Hindi alone accounting for more than half. Access extends beyond language. Over a third of its voice bot calls are handled outside business hours, entirely without agent involvement, extending access to users who reach out in the evenings and overnight. These early numbers point to where voice is headed across its ecosystem. Expressing financial information through speech introduces its own set of technical requirements. An accurate written response must retain its meaning when spoken aloud. Numerical expressions need an unambiguous pronunciation, and financial terminology must fit naturally into the surrounding language. Audio generation also needs to begin promptly enough to sustain a conversation. Raiseholding is building with Cartesia's models to address the speech component of that work. The partnership gives its team, which has built its own financial AI and production architecture, access to specialized speech models purpose-built for multilingual conversation. Multilingual voice infrastructure for the demands of Indian finance. Sonic, Cartesia's text-to-speech model gives developers a choice of voices and regional accents, with controls for pace and emotional expression. It supports 11 Indic languages including Hindi, Tamil, Telugu, and Marathi, alongside language code-switching within a single generation. That flexibility lets teams adapt delivery to different contexts while preserving a consistent voice as a conversation moves between languages. For financial content, expressiveness must be accompanied by precision. Cartesia's work on transcript fidelity and locale-aware pronunciation addresses the treatment of alphanumeric identifiers, dates, and monetary amounts. These capabilities are particularly relevant when English financial terminology appears within an Indian-language sentence, where pronunciation and rhythm need to remain natural without obscuring the information. Cartesia's models are built for secure, high-volume, real-time workloads, with deployment options in India. With Cartesia, Raiseholding run these models within its own infrastructure, built to meet the data-handling and governance standards that financial applications require. A relationship informed by design collaboration. Alongside using Cartesia's models, Raiseholding has been an early design partner to Cartesia and have contributed to its multilingual model development. Through this collaboration between Raise and Cartesia, Raiseholding bring a financial-domain perspective into speech research. Broad measures of naturalness do not, on their own, establish whether a model preserves numerical detail in a code-switched financial sentence. Its domain expertise helps identify the cases that deserve closer evaluation. For Raiseholding, working with Cartesia means the speech models Raiseholding rely on are shaped by the real needs of financial products built for Indian users. Raiseholding look forward to scaling voice across the Raise ecosystem and continuing the technical collaboration between Raise and Cartesia as the partnership expands.