DBT Bureau
Pune, 25 May 2025
Cognizant announced that it has open-sourced its Neuro® AI Multi-Agent Accelerator for research and academic use. This open-source software enables domain experts, researchers, and developers to immediately start prototyping and building agent networks across virtually any use case. The open-source software will help accelerate AI adoption by promoting collaboration in building and customizing multi-agent systems for adaptive operations and real-time decision-making. Enterprises can leverage Cognizant’s Multi-Agent Services Suite to deploy networks of agents in a commercial setting at scale and to efficiently manage them in production, under a commercial license.
The AI Agents Market is anticipated to rapidly grow over the next five years, from a value of USD 5.1 billion in 2024 to a projected worth of USD 47.1 billion by the year 2030. Today’s news speeds how enterprises can leverage interconnected agents to explore new revenue streams and drive scalable business value. Developed by Cognizant’s AI Lab, the open-sourced software for Neuro® AI Multi-Agent Accelerator demonstrates Cognizant’s leadership in AI innovation and its commitment to advancing the application of AI Agents.
Clients like Telstra, Australia’s leading telecommunications and technology company, are working with Cognizant to test and deploy multi-agent systems.
“The open sourcing of the Neuro AI Multi-Agent Accelerator will further empower our teams to rapidly prototype and integrate existing AI agents, and help accelerate our software development lifecycle,” said Telstra’s Group Executive for Product and Technology, Kim Krogh Andersen. “We’re already starting to see the potential for gains in quality, velocity and efficiency as a result.”
Cognizant has also helped a healthcare company create a Contract Negotiator agent network that speeds up medical appeal processing times, as well as a consumer packaged goods firm for analyzing supply chain management. Cognizant is in the process of more than 65 conversations with clients around agentic AI.
Building a successful multi-agent network requires the ability to orchestrate diverse agents, tools, and knowledge sources — including general-purpose large language models (LLMs) and organization-specific systems like service level management (SLMs) or retrieval-augmented generation (RAG) frameworks. Cognizant’s Neuro AI Multi-Agent Accelerator aims to enable virtually seamless integration with APIs, RAG, and third-party agents like Salesforce’s Agentforce, Google’s Agentspace, or Crew AI—via its native Model Context Protocol (MCP) or standard API calls. An optional inter-agent coordination protocol allows these agents to autonomously organize, delegate tasks, and route processes—boosting efficiency and minimizing errors. Agent2Agent (A2A) protocol is also supported, which expands agent collaboration across clouds, platforms, and organizational boundaries.
“To stay competitive in the era of agentic AI, enterprises must be free to experiment—to explore how agents can transform business processes and drive operational efficiencies,” said Babak Hodjat, Chief Technology Officer of AI at Cognizant. “By open-sourcing Neuro AI Multi-Agent Accelerator, we’re expanding access to our cutting-edge multi-agent technology—empowering developers to innovate faster, and enabling decision-makers, regardless of technical background, to rapidly prototype systems and directly observe their impact on key performance indicators.
“Agentforce is built on Salesforce’s deeply unified platform that is open and extensible, empowering our ecosystem of partners and builders to innovate with AI that’s grounded in trust. Cognizant’s decision to open source its Neuro AI Multi-Agent Accelerator exemplifies the kind of partnership that helps our customers move faster and innovate with confidence,” said Gary Lerhaupt, Vice President of Product Architecture, Salesforce. “Together, we’re enabling enterprises to deploy agents that think, collaborate, and deliver value—across every corner of their business.”
Neuro® AI Multi-Agent Accelerator – Key Features:
- Intelligent Opportunity Discovery: Provide a company name or problem area, and the Agent Network Designer will automatically propose a tailored agentic network aligned to your use case—helping you move from idea to implementation faster.
- Rapid and Streamlined Customization: Quickly build and modify multi-agent systems using natural language or leverage prebuilt templates for domains like loan origination, customer service, retail optimization, and intranet automation—dramatically reducing development cycles and risk.
- Scalable, Distributed Operation: Connectors support homegrown tools, APIs, and third-party agents like Salesforce’s Agentforce and Google Agentspace via Model Context Protocol (MCP) or standard API calls. A coordination layer enables agents to intelligently self-organize, distribute tasks, and route processes—boosting efficiency and reducing errors.
- Secure Private Data: Supports regulated industries such as finance and healthcare by isolating confidential information through private data channels to support compliance and data protection.
- LLM and Cloud Provider Agnostic: Easily switch between open-source and most commercial LLMs, as well as private/public cloud providers, without the need for system rebuilds.
- Extensible Coded Tools: Enhance agent networks with custom-coded tools—vital for grounding agent decisions in real-time data, or for defining logical boundaries that trigger human intervention where needed.
- Multi-server, Distributed Deployment: Run agent sub-networks across multiple servers, enabling scalable architectures that support parallel processing, geographic distribution, or segmented use cases.
- Data-Driven Network Definition: Define and update your agentic systems entirely via data-only configuration files—supporting version control, auditability, and rapid reusability across projects.
- Agent Testing Capability: Use the Agent Network Tester to identify bottlenecks or breakdowns within your network. Get actionable insights into coordination issues, agent logic gaps, or integration errors.
Cognizant Scales Agents to 330,000 people via its Intranet, 1Cognizant
Thousands of employees now use 1Cognizant, an intranet assistant powered by its Neuro AI Multi-Agent Accelerator. This tool consolidates and organizes multiple agents to efficiently help employees with various tasks, from reserving meeting rooms and calling cabs to handling inquiries like moving countries or getting married. 1Cognizant can now provide immediate actionable advice and assistance to employees, increasing efficiency and breaking down internal silos.
