AI Video Assistants Could Take Blue-Collar Work Without Touching a Tool

Gemini is currently the only frontier base model — among the families that include Gemini, GPT, and Claude — with native video input. That single technical gap is the hinge of an argument by Komninos Chatzipapas, a founder writing in Inc., that blue-collar jobs are far less protected from AI than the common view assumes.

The standard argument for safety runs like this: a model can write code, but someone still has to assemble the furniture, repair the cabinet, or paint the room. Chatzipapas thinks that reasoning misses the real product being sold. "I think that argument underestimates how much people pay for knowing what to do," he writes. "For some jobs, the customer already has the hands. What's missing is the expertise."

Who supplies the labor?

His central example is flat-pack furniture assembly. The physical actions are manageable for most customers. The instructions are confusing, the pieces look similar, and one mistake can force a restart. People pay to remove that friction, not because they lack muscle.

Now imagine an AI assistant watching through a phone camera. It could identify each part, explain which tool to use, and check the assembly after each step. When the customer gets stuck, it could respond to that exact problem. The customer may still prefer to hire someone. But if enough people decide the guidance is good enough, an assembly business loses paid jobs even though, as Chatzipapas puts it, AI never touches a screwdriver.

The same mechanism applies inside firms. Better guidance could help less experienced workers complete routine tasks that previously required more supervision from senior staff.

What does video understanding change?

A tutorial shows how a job should go. An assistant with reliable video understanding could respond to the process as it unfolds, notice mistakes, and adjust instructions when conditions change. Chatzipapas calls that distinction the larger opportunity, and says today's models are only a starting point on the path to it.

Google already points the way. Gemini Live provides camera-based guidance, including highlighting a tool on the user's screen, and Google describes using its camera feature for step-by-step repair help.

How should service businesses test their exposure?

Chatzipapas does not claim every trade is replaceable. Some work demands equipment, physical skill, and judgment that instructions cannot supply. Customers also pay for convenience; many people will hire someone even when they know how to do the job.

The risk arrives earlier, at the margins. Losing some routine bookings, or competing against workers who can do more with AI, could squeeze profit margins well before an entire service becomes replaceable. He offers owners a concrete audit:

• Review the last 50 jobs. Which required specialist judgment, and which mainly required explaining a repeatable process?
• Test camera assistance on a few low-risk tasks under experienced supervision.
• Measure completion time, mistakes, and callbacks — finishing faster offers little advantage if it causes rework.
• Redesign the service itself: a repair business could sell remote guidance for simple maintenance and reserve site visits for problems that need an expert.

The employment effects cut both ways. Higher productivity could let a business serve more customers and grow with fewer additional hires. Whether jobs expand or shrink depends on whether demand grows fast enough to absorb the extra capacity.

The bottom line

Calling physical work "AI-proof," Chatzipapas writes, encourages owners to overlook these possibilities. The person holding the tools may remain essential while AI changes who gets hired, how much customers pay, and how many workers a business needs — which is why he frames the coming pressure not as replacement, but as a shift in the economics of expertise.