First Humans as Robots, Now Robots as Humans
We trained humans to work like machines. Now our standardized work processes are the training manual for AI
"I've spent my career training humans to act like robots," he said, pausing to take a sip of water. "Now I'm training robots to act like humans."
The statement hung in the air between us, its implications unfurling like smoke. Across the table sat the founder of an AI sales technology company, a man who had built his fortune by hiring hundreds of college graduates and molding them into perfectly calibrated sales development representatives. His previous company had been a marvel of human standardization – young people trained to follow scripts, overcome objections with predetermined responses, and convert leads with machine-like precision.
His career arc was a perfect circle from mechanizing humans to humanizing machines.
This pattern—training humans to follow rigid processes, then teaching machines those same processes—reflects a much broader trend in how we've organized work.
What began in factories with time-motion studies has infiltrated our offices, call centers, and virtual workspaces. The assembly line mentality didn't disappear—it evolved, donning business casual and hiding behind terms like "best practices" and "scalable processes." We've spent decades breaking down complex work into its component parts, documenting every step, and training people to execute with consistency.
The assembly line didn't just transform manufacturing—it became the invisible blueprint for knowledge work. Just as McDonald's revolutionized food service by breaking down burger-making into discrete steps (toast bun, add patty, squirt sauce, layer lettuce), we've applied the same thinking to white-collar work. The marketing team follows their own recipe: research keywords, draft outline, create content, optimize for SEO, schedule distribution. The sales process: identify prospect, send initial email, follow up on day three, overcome objection A with response B. We've turned professional work into a series of standardized stations on an invisible assembly line.
In our pursuit of efficiency and scale, we've created comprehensive training manuals for our work. These documented processes—originally designed to onboard new employees quickly—now serve as perfect instruction sets for AI systems.
The founder across from me (and countless others across the industry) had simply connected these dots. After years of standardizing human sales behavior and documenting successful techniques, he was now applying those same patterns to his AI systems. The training manuals created for humans transferred seamlessly to machines.
Many workers have invested their careers in mastering these standardized processes. Their value has been measured by how perfectly they execute established protocols. As AI systems learn to follow these same protocols, the nature of human contribution inevitably shifts. The question becomes not whether the process can be automated, but what aspects of work remain.
So what work remains uniquely human in this new landscape? One way to know you're beyond the reach of standardization is if you struggle to answer the question "What exactly do you do?" It's the question that made Peter Gibbons squirm in Office Space, trying to explain his nebulous role to the efficiency consultants. If you can't distill your work into a clear process flowchart, you might be safe—for now.
These unexplored, undefined spaces—where problems haven't been fully articulated and solutions haven't been standardized—remain the terrain of human work - for now.
AI automation presumes that a manual version already exists and can be used as a benchmark or standard. It needs the recipe before it can cook the meal. As Wired founder Kevin Kelly exhorts us, the future belongs to those who "work on things without a name"—the undefined territories where standardization hasn't yet reached, where the map is still being drawn.
Look at your daily work. How much of it follows a script someone else created yesterday? And ask yourself: why wouldn't that script become an algorithm tomorrow?
If you like this post, you should check out:
