The technology sector is caught in an AI gold rush. Manufacturers are embedding artificial intelligence into everything from toothbrushes that claim to analyze brushing habits to refrigerators that supposedly track grocery inventory. Yet the ability to run a model on a chip does not automatically justify doing so. Savannah Kunovsky, who leads industrial product design work at IDEO's Emerging Tech Lab, recently explained to IEEE Spectrum that many product teams face pressure to deliver an "AI story" for investors and executives, even when the technology creates more friction than value. The central question is not whether AI can be added to a device. It is whether AI genuinely helps the person using it.

What Happened

Kunovsky described a common pattern: organizations rush to integrate large language models or computer vision because competitors are doing the same, only to discover the new feature obstructs the user. She cited the Humane AI Pin as a prominent example. The screenless wearable asked people to abandon a familiar smartphone interface for a slower, less dependable AI assistant. In another instance, a client requested a cooktop with no knobs, relying entirely on voice commands to regulate heat. During testing, a pan began smoking while the user waited for the AI to respond. The design team ultimately retained the physical knobs. Kunovsky also noted that automakers have shifted climate and volume controls onto touchscreens and voice menus, forcing drivers to look away from the road. The issue is significant enough that Europe's Euro NCAP safety program now rewards vehicles that retain physical controls.

Why This Matters

Weak AI integration does more than irritate customers. It can damage trust, introduce safety hazards, and transform practical products into costly gimmicks. When artificial intelligence replaces a control that already works by feel—such as a stove burner or a light switch—it demands attention at the worst possible moment. Kunovsky emphasized that tasks requiring instant reactions, tactile feedback, or performance under stress are usually better served by mechanical controls. At the same time, AI excels when it operates in the background or expands human capability. Accessibility tools illustrate this clearly: hearing aids that isolate one voice in a noisy restaurant, apps that describe the visual world for people with low vision, and earbuds that translate spoken conversation in real time. The deciding factor remains whether the technology helps someone accomplish their goal more easily.

Key Takeaways

  • Start with the user, not the technology. Teams should define who the product is for and what they are trying to achieve before deciding whether AI has a role.
  • AI is not always the right tool. The appropriate capability might be computer vision, sensor fusion, an on-device model, or no AI at all.
  • Physical controls still matter. Volume knobs, light switches, and stove burners often outperform voice or touchscreen interfaces for fast, tactile tasks.
  • Effective AI can be invisible. The strongest integrations either power the core experience or work quietly in the background, learning preferences without demanding attention.
  • Human override is essential. Users need obvious, fast ways to undo actions, edit suggestions, or turn features off, particularly with generative AI that can produce unpredictable results.

Conclusion

Adding AI to a product should never be an objective in itself. As Kunovsky emphasized, the measure of success is straightforward: does it help someone do what they set out to do? Companies that prioritize user needs, evaluate technical feasibility second, and weigh business viability third are more likely to build AI features that endure. Whether the final design incorporates a sophisticated model or a simple mechanical switch, the goal stays the same—create products that respect the person using them.