Enterprise AI is moving quickly from experimentation to execution, but product development environments show why scaling AI remains difficult.

These environments are complex by design. Product development depends on requirements, engineering data, product lifecycle management (PLM), manufacturing execution, quality, service feedback and compliance controls working together. Introducing AI into one part of that lifecycle may improve a specific task, but it will not transform the enterprise if the wider process remains fragmented.

HCLTech’s The AI Impact Imperatives, 2026 report found that respondents expect 43% of major AI projects starting over the next 24 months to fail. In complex product development, that risk is about model performance and whether AI can work within the process, data and governance realities of the business.

Sudhir Kumar Singh, Head of DDMS Practice, ERS at HCLTech, sees this clearly in client conversations.

“People are in a testing phase right now,” he said. “People want AI solutions everywhere.”

The pressure is understandable. Clients are asking transformation teams to compress timelines that previously took 12 months into 6 or 7 months. AI, GenAI and automation are expected to help deliver that acceleration. But in regulated, engineering-led industries, speed cannot come at the expense of process discipline.

“Nobody wants to compromise on the processes. Nobody wants to compromise on the compliance part of it, quality part of it, safety part of it. All that is non-negotiable,” said Singh.

That is why AI in product development needs to be treated as more than a productivity layer. It needs to become part of a connected lifecycle model.

The problem with isolated AI use cases

Many organizations are still approaching AI through isolated use cases. One team may apply AI to shorten the requirements cycle. Another may use it to improve the flow from engineering bill of materials to manufacturing bill of materials. A manufacturing team may introduce a separate AI solution to improve production execution.

Each use case may deliver value, but the bigger opportunity comes when AI can support the full product lifecycle.

The HCLTech report found that 40% of respondents cite cross-functional coordination as a challenge limiting AI adoption, while 39% cite alignment with business strategy. In product development, those challenges are especially visible because no single function owns the complete lifecycle.

An R&D engineer may understand concept design but not manufacturing execution. A manufacturing team may understand production constraints, but not the original design intent. A service organization may understand field failures, but that insight does not always flow back into future product development.

For Singh, PLM is important because it can help create a closed-loop system.

“We call it a closed-loop system,” he said. “The whole system has to work seamlessly today if we are saying that we want to improve productivity or improve time to market.”

That system needs to connect application lifecycle management, PLM, manufacturing execution and the wider application landscape around them. If AI is introduced into only one phase, the enterprise may improve a local process without improving the full lifecycle.

PLM as the lifecycle intelligence layer

PLM can help move AI from isolated use cases into connected workflows because it sits at the center of product data, process and traceability.

In a connected lifecycle, requirements are captured, product concepts are designed, engineering data is managed, manufacturing data is created, products are released and field performance is fed back into future design decisions. The data should move in a loop, not a line.

This is important because AI depends on reliable data. The HCLTech report found that only 21% of respondents feel their data estate is modernized and operational. It also found that the majority report issues with data visibility and automating data orchestration, intelligent data management, consolidating data platforms and building consumable data products.

Those issues directly affect AI in PLM. If product data is duplicated, inconsistent or disconnected across systems, AI will struggle to identify the right source of truth. Instead of accelerating decisions, it may create new uncertainty.

Singh argued that this is where the next stage of AI-enabled PLM needs to focus.

“AI has to ensure that there is a single source of truth,” he said. “There is no duplication of data, no deviation of data, no corruption of data.”

That is especially important as product development becomes more intelligent. AI can help shorten engineering cycles, automate decision support, improve manufacturing readiness and connect field feedback to new product development. But those benefits depend on clean, accurate and contextual data flowing across the lifecycle.

Business value depends on lifecycle outcomes

The business case for AI in PLM should not be defined by individual tool improvements alone. The stronger measures are lifecycle outcomes: faster time to market, improved productivity, fewer data discrepancies, stronger quality, better manufacturing readiness and more effective service feedback.

To achieve that, organizations need cross-functional ownership. IT teams may implement PLM platforms, but business users shape how those platforms are used. Engineering, manufacturing, quality and service teams all bring different perspectives on where AI can create value. If only one group defines the AI agenda, the result is likely to be incomplete.

Singh noted that service organizations can play an important role because they work across multiple phases of the lifecycle and across multiple client environments. That broader perspective can help organizations understand where AI should be embedded, how systems need to connect and where data governance matters most.

The future of PLM will be shaped by this integration. AI will not sit beside PLM as a separate capability. It will increasingly be embedded into product intelligence, lifecycle decision-making and enterprise transformation.

“All the AI solutions are driven by the data,” said Singh. “It is very important that we have the right set of data available to scale the AI solutions.”

For product organizations, the implication is clear. AI can accelerate parts of the lifecycle, but it can only transform the full lifecycle when the foundations are connected.

The organizations that succeed will be those that build AI into a clean, integrated and governed product data environment, where requirements, engineering, manufacturing, service and field insight reinforce one another.

That is how AI can move from isolated productivity gains to enterprise-scale product intelligence.