Summary

• Researchers led by Nyenrode Business University frame generative AI as a “synthetic teammate” in product development.
• The approach keeps humans responsible for objectives, inputs, process quality, and outputs.
• The research gives managers a more useful adoption model than replacing staff or blindly automating creative work.

Nyenrode Business University-led research is urging companies to treat generative AI as a “synthetic teammate” in new product development, rather than as a replacement for human judgement or a managerial shortcut.

The research, developed with academics from IE University, Royal Holloway, Pennsylvania State University, and Dickinson College, argues that generative AI can support customer understanding, idea generation, concept development, prototyping, launch planning, and commercialisation. Human teams remain responsible for objectives, inputs, processes, and outputs.

Enterprise AI adoption is often trapped between two weak positions. One treats AI as a productivity engine that can replace expertise, while the other treats it as a risky tool that should be confined to experimentation. A synthetic teammate model is more operational. It recognises that AI can contribute useful work, but only inside a managed process with supervision, quality control, and clear accountability.

In product development, that distinction is practical rather than philosophical. Generative AI can analyse customer data, surface patterns, produce concept variations, simulate messaging, or accelerate early design. It can also hallucinate, reinforce bias, misread context, leak sensitive information, generate unusable ideas, or create false confidence. The technology’s value depends on how well teams assign tasks, check outputs, and decide when human expertise must override the machine.

Treating AI like a junior colleague is not a perfect metaphor, although it is more useful than treating it as an oracle. Junior colleagues can produce drafts, research, and ideas, but they need direction, review, and boundaries. They should not set strategy alone, approve compliance-sensitive outputs, or turn weak assumptions into final decisions.

The research gives product leaders a way to place AI inside the workflow. During customer understanding, it may support market intelligence and segmentation, while humans assess data quality and relevance. During ideation, it may produce volume and variation, while managers screen for feasibility and strategic fit. During prototyping, it can speed up iteration, while specialists check quality and specifications. During launch, it may support marketing and supply chain planning, while humans interpret cultural, commercial, and operational realities.

That model fits the current enterprise adoption problem. Many organisations have given staff access to AI tools before redesigning processes or governance. The result can be scattered experimentation, duplicated work, and uncertain accountability. A teammate framing encourages teams to define roles, handoffs, review gates, and evidence standards.

Human oversight can also become a slogan if people lack time, expertise, authority, or incentives to challenge AI output. Employees expected to approve more work faster, without understanding how outputs were produced, may become rubber stamps rather than safeguards.

The research points to a more mature phase of AI adoption. Productivity gains are more likely to come from redesigning work so that AI handles bounded tasks while humans retain judgement, responsibility, and organisational memory. The boss should still be human.