‘Neuro AI is our flagship AI platform,’ Hodjat tells CRN. ‘We use it to develop decision-making use cases for our clients. Now we’ve made it agent-based. It’s a multi-agent-based system with humans in the loop to empower the platform and empower our user and our clients.’

Cognizant said it has significantly enhanced its Cognizant Neuro AI platform as a way to help enterprises to discover, prototype, and develop AI use cases.

The aim of the new version of Cognizant Neuro AI, unveiled this week, is to help businesses improve decision-making as a way to increase company performance and find new revenue opportunities, said Babak Hodjat, chief technology officer for AI for the Teaneck, N.J.-based Cognizant, ranked No. 8 on the CRN Solution Provider 500.

The enhanced Cognizant Nero AI platform follows the July introduction of the Cognizant Neuro Edge, which brings AI and GenAI to enterprise businesses, especially manufacturers, working with edge AI from chips and devices to applications. Cognizant Neuro Edge is a generic framework aimed at accelerating the development of enterprise edge services.

[Related: Cognizant Launches New Advanced AI Lab With Focus On Core AI Research]

“Neuro AI is our flagship AI platform,” Hodjat told CRN. “We use it to develop decision-making use cases for our clients. Now we’ve made it agent-based. It’s a multi-agent-based system with humans in the loop to empower the platform and empower our user and our clients.”

For many people, their first gut reaction to AI and GenAI is that it will help find some patterns in their data and tell them what’s going to happen in the future, and maybe provide some insights, Hodjat said.

“That’s a part of our platform,” he said. “But what really differentiates it is that it then goes on to give you suggestions as to what to do. So those actions or decisions you want to make are not trivial. It’s not very easy to make decisions, especially against more than one outcome at the same time, such as if you want to maximize revenue while minimizing risk while being sustainable. Those kinds of things don’t always align. So the AI helps you find the best balance, the best decision strategy, by creating a digital twin, a machine-learning-based digital twin of the subject for decision-making.”

The new Cognizant Neuro AI platform is now agent-based so that its agent can collaborate with other agents to identify use cases to help simplify decision-making, Hodjat said.

“So from that really first touch with the client, the AI agent is helping collaboratively identify the use cases that would give them the best, most impact, and then help them build the use cases, end to end,” he said.

Cognizant has already developed over 500 use cases on its Neuro platform for internal use and for work it has done with its clients, Hodjat said.

“It’s like an accelerator,” he said. “We get in front of a client, and we use it. We help the client identify and build the use cases, and we take it from there. What’s new is that the Neuro AI platform will now be available for clients to actually deploy and use it as sort of a use case generation factory to do GenAI agent-based use case generation. That’s partly because many of our clients have demanded it. We’ve had clients saying, ‘We want to bring this in, run it on our own data in-house, and have our own folks trained to use it.’”

The Cognizant Neuro AI platform agents are all GenAI-based, but it also orchestrates non-GenAI-based AI techniques as well, he said.

As an example, Hodjat said to consider a large retailer looking for ideas for new store locations. The user could ask the platform’s scoping agent to guess what data would be available for a typical major US retailer when it comes to deciding where to open new stores. The platform might start looking for demographic, population density, household income, competition, and other data from public or internal sources. That data may be structured or unstructured, he said.

“At that point, for a specific location to open a store, it could produce recommendations for what the store size, what market mix, what product mix, what kind of store format, investment level, market strategy for the launch, staffing levels, and so forth,” he said. “The strategy is going to be informed by the outcomes. So it’s going to want to obviously maximize our annual sales while maximizing our ROI and reducing our cost. The retailer has KPI (key performance indicators), and you want the AI to be aligned with those KPIs.”

The KPIs could be changed, or new data added, before moving on to the Neuro AI’s data generator which can then autonomously work with other agents to write code to generate a use case, Hodjat said. Once the use case is double-checked by the user, the user can start training the model and the platform’s predictor can predict the various outcomes, he said. Users can also add uncertainty to the model, he said.

The Neuro AI platform can actually come up with predictions, as well as raise flags with it determines there is not enough data, Hodjat said

“We can add things, like an LLM here for the front end, so I have a ChatGPT-like interface to it,” he said. “I could add more LLMs. So if I had not generated the data or the data had missing values, I could have an LLM do that. I can actually go talk to the data.”

The platform lets the AI check the AI’s work, Hodjat said.

“That’s something that we really do a lot,” he said. “I can actually have the system review the design that it made and come back to me and say, ‘OK, this design has these strengths, these weaknesses. Here’s some things that I think you should do to improve it.’”

Hodjat said he recently demonstrated the Cognizant Neuro AI Platform to a group of executives who were clamoring to do their own use cases.

“In fact, one use case had to do with Hurricane Milton coming in and their logistics and so forth,” he said. “And we built it on the fly. However, for us to actually build the use case end-to-end, we have to rely on synthetic data, and then that acts as a template. So then we take that in, and we do the hard work of mapping their real data and maybe third party or public data to that template, and can then make it work.”