AI in Product Development

 

Artificial intelligence dramatically increases the throughput of selected development activities. In complex engineering environments, however, sustainable business value emerges only when product definition, systems architecture, verification, and organizational accountability mature at the same pace. This applies equally to products, machines, industrial equipment, and increasingly automated services. Artificial Intelligence Accelerates Tasks—Competitive Advantage Requires System Maturity!

Leadership must optimize the end-to-end development flow rather than individual activities. Otherwise, organizations risk creating technical debt, regulatory exposure, and legal liabilities that materialize later in the product lifecycle. The time to act is now.

Efficiency Gains Shift the Bottleneck

Generative Artificial Intelligence (Generative AI) creates content with substantially greater efficiency than conventional approaches. Its strengths include ideation, research, software development, document structuring, and repetitive knowledge work. Applications span the entire lifecycle—from concept development through verification and operations. Yet these productivity gains are inherently uneven.

An AI assistant can generate software code or test cases faster than a human. It cannot assume the product manager’s responsibility for defining customer value or the systems architect’s responsibility for ensuring overall system integrity. In industrial and cyber-physical applications (“Physical AI”), the challenge becomes even greater because of complex physical interactions and dependencies.

AI compresses individual engineering tasks while shifting constraints to the activities that connect them: stakeholder needs, system, technology and safety requirements, architectural decisions, interface management, verification, and validation. Work does not disappear—it moves. As a result, the structure of the product development process and the quality of engineering leadership become the primary determinants of successful transformation.

Without Technical Coherence, There Is No Business Value

AI creates value only when Product Development—the business process from market opportunity through product launch and lifecycle evolution—and Systems Engineering—the discipline that ensures technical coherence and lifecycle performance—remain tightly aligned. Accelerating isolated activities inevitably creates pressure elsewhere in the system.

A rapidly generated solution may appear economically attractive while overlooking critical system constraints. It creates value only if it satisfies customer needs, complies with regulatory requirements, and can be manufactured, operated, maintained, and evolved throughout its lifecycle.

In practice, Generative AI often accelerates the creation of plausible but unusable alternatives. This apparent productivity merely shifts cost into engineering rework, certification, operations, and maintenance. In many cases, these effects emerge only later—and with greater impact.

Optimize the Entire Development Flow

Today’s most reliable AI applications are those in which outputs can be validated against objective references. Examples include requirements analysis, code generation, identification of inconsistent requirements, traceability management, and generation of verification procedures. Research on requirements engineering itself, however, still contains significant open challenges.

Measuring isolated productivity improvements misses the point. The relevant business metrics are lifecycle metrics, including:

• Time to a verified development baseline
• Time-to-defect detection
• Rework rate
• Engineering change stability
• Market performance after release

Otherwise, organizations merely shift additional verification and maintenance effort onto experienced engineers. Verification and validation therefore become even more critical:

• Verification: Was the solution implemented correctly according to the technical specification?
• Validation: Does the solution solve the intended problem and fulfill its intended purpose?

Together, these disciplines form the essential quality gate because:

• Generative AI can produce technically correct artifacts based on incorrect engineering assumptions.
• AI-generated software may become unnecessarily complex, include unintended third-party code, introduce cybersecurity risks, and reduce long-term maintainability.
• Uncontrolled prompt usage across engineering teams creates inconsistent design quality. Some prompts produce appropriate solutions, while overly simplistic or overly complex prompts generate unnecessary rework and maintenance cost. Prompt standardization therefore becomes an engineering discipline.
• Local optimization in industrial products may unintentionally degrade thermal behavior, electromagnetic compatibility (EMC), tolerance chains, usability, or service accessibility when prompts fail to represent system physics adequately. Recent Chinese research on industrial AI (“Physical AI”) consistently identifies the integration of data, physics-based models, and rigorous verification as a critical success factor. Organizations therefore require cross-functional teams with clearly defined competencies and accountability.

Organizational Capability Determines Success

Engineering professionals increasingly assume supervisory rather than purely execution-oriented responsibilities. They define tasks, evaluate AI outputs, correct deficiencies, and remain accountable for engineering decisions. This transition can significantly increase the value of engineering work.

However, organizations also face the risk of capability erosion if junior engineers begin accepting AI-generated outputs before developing sufficient understanding of engineering principles and failure mechanisms. Structured learning paths therefore become essential because effective human oversight requires deep technical competence.

Accountability must remain personal and explicit. AI is a tool—not an accountable decision maker. Responsibility remains with people:

• The AI system owner is accountable for the AI capability.
• The Product Manager remains accountable for customer value.
• The Systems Architect remains accountable for system integrity.
• The Process Owner remains accountable for development execution.

What This Means for Your Business

Organizations should structure their AI portfolio according to the development object and expected engineering outcome: What is being developed, and what artifact is being produced? Every AI use case requires:

• A clearly defined business purpose
• A measurable value hypothesis
• An appropriate risk classification

Approved engineering baselines, end-to-end traceability, and formal release procedures must apply equally to AI-generated artifacts.

Managing the overall development flow is more important than maximizing local productivity. An AI pilot is successful only if it reduces the time required to reach a validated product baseline while demonstrably lowering total lifecycle cost. Anything less represents localized optimization without business impact.

Every AI deployment also requires an organizational capability model that protects systems thinking, verification competence, and accountable decision-making. Keep humans responsible for engineering decisions through a Human-in-the-Loop governance model.

Only then does AI evolve from an isolated productivity tool into a controlled component of the product development system—delivering measurable and sustainable business performance.

THERON supports organizations in transforming their product development processes to realize this potential.

Back to Publications