“Physical AI” becomes mainstream: robots leave the lab

Coverage is increasingly framing CES 2026 around robots and physical AI, not just chatbots.

Article link: https://open.substack.com/pub/johanosteyn/p/physical-ai-becomes-mainstream-robots?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Audio summary: https://youtu.be/BWAx-kHV8t0

Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/

If you want a clean takeaway from CES 2026, it is this: AI is moving off the screen. For a while, the public story of AI has been chatbots and assistants. At CES, robotics and what many are calling physical AI moved into the spotlight, from humanoid demonstrations to industrial machines designed for real work. That matters because software changes how we think; robots change what gets done, where it gets done, and who gets paid.

CONTEXT AND BACKGROUND

Robotics has had many hype cycles. We have been promised household helpers for decades and mostly got narrow successes: robot vacuums, factory arms, and clever sensors. What is changing is the intelligence layer. Modern AI can interpret messy environments and follow instructions more naturally, which is essential when a machine must navigate the physical world.

The second shift is platformisation. Major vendors are trying to standardise how robots are trained, simulated, and deployed, so companies can build and adapt robots faster than before.

INSIGHT AND ANALYSIS

It is worth staying grounded. Many humanoid demos still look impressive and fragile at the same time: tasks can be slow, the environment must be prepared, and the systems remain expensive. Analysts watching CES have stressed that general-purpose humanoid robots are not yet a mass commercial reality.

But the early value does not require a perfect human substitute. The first wins will be narrow, high-impact work in semi-controlled settings: inspection, inventory movement, sorting, remote operation in unsafe zones, and routine monitoring. A robot that patrols and escalates to a human is already useful, even if it cannot do household chores reliably.

South Africa’s lens is sharper. We need productivity and safety gains, but we also have high unemployment and skills gaps. If robots arrive as imported black boxes, we risk displacing labour without growing the local technician and integrator roles that should come with adoption. The smart approach is capability-building alongside automation.

IMPLICATIONS

For business leaders, treat physical AI as a safety-and-operations programme, not a gadget purchase. Start with dangerous or repetitive work, define human oversight and incident procedures, and measure performance before scaling beyond pilots.

For policymakers and educators, focus on skills and accountability. Robotics will raise demand for mechatronics, maintenance, integration, and governance. We should build pathways into these roles now, while clarifying who is responsible when autonomous systems cause harm.

CLOSING TAKEAWAY

CES 2026 did not prove that robots are ready to run our homes tomorrow. It did show that robotics is reorganising around AI platforms, and that the pace is accelerating. The real test for South Africa is whether we use physical AI to improve safety and productivity while strengthening skills, dignity, and trust. If we get that balance wrong, the backlash will not be about technology. It will be about who society feels was protected, and who was left behind.

Author Bio: Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net