Cognizant targets connected, AI-enabled manufacturing with new platform

As manufacturers look to extend artificial intelligence beyond data analysis and into physical operations, the focus is increasingly shifting towards systems that can sense, interpret and respond to conditions on the factory floor.

Cognizant has launched a sovereign Physical AI platform aimed at helping manufacturers move beyond isolated artificial intelligence (AI) pilots and connect sensors, robotics, factory automation and other operational systems through a unified intelligence layer.

The technology services company said its Physical AI Platform-as-a-Service, built on the Cognizant Intelligence Spine, is designed to support AI deployment across physical operations while keeping data, rules and institutional knowledge under enterprise control.

In an exclusive interview with Australian Manufacturing, Cognizant Managing Director for ANZ Carla Ramchand said Physical AI differs from conventional industrial AI by enabling systems to interpret their surroundings, respond to changing conditions and take action in real time.

“Many AI tools today analyse data, predict faults, or recommend what someone should do next,” Ramchand said. “Physical AI takes that step further by enabling machines to understand their environment and context, provide reason for changing conditions, and act in real time.”

She said the shift is being enabled by the development of advanced sensors, robotics, multimodal AI and low-latency connectivity.

For manufacturers, Cognizant said this could mean moving from systems that identify problems and alert operators towards systems that can also determine causes and undertake authorised corrective actions within defined safety boundaries.

Moving beyond AI pilots

Cognizant said scaling AI from controlled pilots across multiple machines, production lines and sites can introduce technical and organisational challenges.

“Factories are complex environments,” Ramchand said. “Legacy equipment operates alongside newer sensors, robots and digital platforms, while data ownership and connectivity can vary across.”

She identified latency, integration, cybersecurity and safety as potential challenges when AI systems move into live operations. Organisational issues, including unclear ownership, inconsistent workflows and workforce readiness, can also emerge during implementation.

Cognizant recommends that manufacturers begin with a clearly defined operational problem and measurable target, while designing for production from the outset. Ramchand said security, governance, workforce training and lifecycle support should be incorporated, with digital twins used to test decisions before they reach live operations.

The Cognizant Intelligence Spine is intended to provide a framework between physical systems and the AI layer that reasons and acts, allowing individual AI applications to be connected rather than operating as standalone deployments.

Connecting existing factory infrastructure

Cognizant said its platform is designed to work across legacy and newer technologies, allowing manufacturers to modernise without necessarily replacing existing equipment.

“Interoperability is essential because most factories run on a mix of old and new technology,” Ramchand said.

She said sensors and integration tools can connect older equipment to newer factory systems, allowing existing assets to generate and share data that can be used by AI.

Rather than replacing machinery across an entire facility, Ramchand advised manufacturers to begin with a high-value asset or workflow, secure and connect its data, measure the return and then apply the approach elsewhere.

“Making existing equipment more connected and intelligent should be the focus of factories and not replacing every machine,” she said.

Cognizant said its Physical AI service supports manufacturing applications including predictive maintenance, autonomous quality control, robotic process integration and yield optimisation. Connecting these applications could allow information from cameras, equipment, maintenance systems and production processes to inform decisions across multiple operational areas.

The company also sees potential for Physical AI to address productivity and workforce pressures by continuously monitoring machinery, automating repetitive inspections and helping employees assess faults using live data.

However, Ramchand said the objective should not be complete automation.

“The goal should not be complete automation,” she said. “Manufacturers need to focus on targeting one clear result, prove it works safely, then roll the same system out to other lines and sites.”

Cognizant expects Physical AI to contribute to a shift from isolated automation towards more adaptive manufacturing operations over the next five years. Ramchand said this would not necessarily mean fully autonomous factories, but greater integration between AI-enabled systems and human workers.

As AI begins influencing physical operations, she said manufacturers should establish clear limits on autonomous actions, retain human oversight for safety-critical or irreversible decisions, and use manual overrides and fail-safe modes.

“Manufacturers must have control over the data, rules, and institutional knowledge shaping their systems,” Ramchand said. “Making every action visible, testable, and accountable is where we see this hybrid industrial workforce become a place of trust.”