4 min Analytics

Cognizant conducts multi-agent orchestration on Neuro AI 

Cognizant conducts multi-agent orchestration on Neuro AI 

Enterprise technology consultancy cum software engineering tools provider company Cognizant has recently detailed work to enhance its Cognizant Neuro AI platform with multi-agent orchestration capabilities. With the rise of agentic AI functions now becoming ubiquitous across the technology trade, the company claims that “most enterprises” struggle with implementing and scaling cross-enterprise use cases for AI. As such, it wants to help this challenge by addressing these problems with services that identify business problems and generate synthetic data (or work to import an organisation’s own anonymised data) to start creating AI models. 

According to Babak Hodjat, CTO of AI at Cognizant, the Neuro AI platform has been tested, piloted and used for “almost any industry or business challenge” involving data analysis, from inventory management and dynamic pricing to fraud reduction and efficient staff allocation.

The enhancements to the Neuro AI platform began as research projects at the Cognizant AI Research Lab, which launched earlier this year. The lab focuses on researching and developing decision-based AI systems. These enhancements are the first developments from the lab that have been integrated into a commercial offering. 

Multi-agent discovery tool 

Cognizant says it has integrated new features where including a multi-agent powered discovery tool to identify use cases called Opportunity Finder, as well as a suite of large language model (LLM) assistants that form an AI decisioning engine. 

The company explains how AI engineers and other software team members first interact with the platform through Opportunity Finder, which is essentially an LLM-assistant that helps identify potential AI decisioning use cases for a business. A Model Orchestrator function (featuring a drag-and-drop interface) enables users to clean up data and apply a variety of machine learning models to it. 

Data preparation is streamlined through LLMs and then machine learning models are applied to predict outcomes while evolutionary AI models prescribe decisions. Once trained, the best models can be further interrogated through a web interface or queried via an LLM assistant. 

Industry vertical configurations

Cognizant Neuro AI is available with pre-built configurations that provide starting points for use cases across include healthcare (like drug discovery and treatment plans), finance (such as cybersecurity and fraud prevention), agriculture (crop yield optimization and pesticide development) and general templates for supply chain, call centers, customer retention, and price optimization. 

“Businesses are struggling with how and where to apply AI to solve business problems, and that’s why we’ve seen most AI use cases limited to prediction-based outcomes or single LLM chat-based solutions,” said Hodjat. “Multi-agent AI systems hold the key to solving these problems, which is why Cognizant Neuro AI is now built with one at its core. This platform puts business leaders – not just data scientists — in the driver’s seat, so they can tap into their own domain knowledge to quickly test and establish decision-making use cases for AI in minutes and then provide the resulting model code to iterate at scale.” 

According to magical analyst house Gartner, the use of multiple AI agents can work toward a common goal that “goes beyond the ability of individual agents” today. Umm, yes, we did figure that out. Thankfully there is more insight from Neil Ward-Dutton, VP for automation, AI and analytics at IDC who says that as enterprises start to try to approach AI strategically and move beyond experimentation, they are crying out to understand how to identify and prioritise use cases. He thinks that [AI model and tools] providers that can use technology to help accelerate the identification of use cases and then use that technology to test and scale implementations, will be in a strong position.

The real challenge with AI

“Many enterprises struggle to apply AI beyond predicting outcomes and that’s because solving real business problems usually involves thousands of different scenarios often with conflicting priorities,” said Murali Vridhachalam, head of cloud, data and analytics at Gilead Sciences, a Cognizant customer.

This then could be the real challenge with AI implementation today i.e. there are hundreds (okay thousands) of tools, models, functions and fabrications in this space – and that’s a reality that is only going to get more granular and fragmented with the rise of agentic AI and the pressing need to orchestrate within these environments – so being able to use templates for industry use cases, having control of both proprietary and synthetic data streams… and also being able to untangle all the business rules, requests and logic and fitting those requirements to the application of AI might just help. A little at least, right?