I've been thinking about why so many discussions about AI in the enterprise feel oddly familiar, and I keep returning to an observation that initially seems tangential: we're essentially re-running the management theory debates of the 20th century, but at software speed.
This isn't immediately obvious, but bear with me—because understanding this parallel helps explain both the genuine transformation AI represents and the specific anxieties it's triggering in organizations today.
The Scientific Management Moment
Let's start with Frederick Taylor. In the early 1900s, Taylor's "Scientific Management" promised to revolutionize productivity through systematic analysis of work processes. His famous time-and-motion studies at Bethlehem Steel—determining the optimal size of a shovel, the ideal arc of a worker's swing—were the original "there's a science to this" moment. Taylor's insight was that knowledge about how to do work efficiently could be extracted from experienced workers, systematized, and then redistributed as standardized procedures.
The Taylorist promise was compelling: management could observe work, codify best practices, and then scale those practices across the organization independent of individual worker expertise. This was fundamentally about knowledge transfer—taking tacit expertise and making it explicit and replicable.
What strikes me is how precisely this maps onto what we're attempting with AI today. Large language models are, in effect, performing a Taylorist analysis at unprecedented scale: observing how knowledge work gets done (through training on vast corpora of text), identifying patterns, and then offering to execute those patterns on demand. When we prompt an AI to "write a marketing email in the style of a SaaS company" or "analyze this data like a management consultant would," we're essentially asking it to apply the systematized knowledge it extracted from observing millions of examples.
The parallel goes deeper. Taylor's critics—and there were many, including labor unions who correctly saw deskilling threats—argued that Scientific Management reduced workers to interchangeable parts, stripped away autonomy, and ignored the tacit knowledge that couldn't be captured in a stopwatch study. Sound familiar?
The Knowledge Work Inflection
But here's where the history gets more interesting, because management theory didn't end with Taylor. Peter Drucker's articulation of the "knowledge worker" in the 1950s represented a fundamental challenge to Taylorism. Drucker observed that an increasing share of workers were doing jobs where the work itself was non-routine and required judgment, creativity, and expertise that couldn't easily be reduced to standardized procedures.
The management challenge shifted from "extract and standardize knowledge" to "how do we enable knowledge workers to be productive when we can't fully specify what they're doing?" This led to decades of theory about autonomy, intrinsic motivation (Dan Pink), learning organizations (Peter Senge), and flatter hierarchies. The implicit assumption was that meaningful knowledge work was, by its nature, resistant to the kind of systematization Taylor championed.
This is the assumption AI is now stress-testing.
What's critical to understand is that AI isn't just a more efficient way to do Taylorism—it's revealing which knowledge work was always more systematizable than we wanted to admit, and which truly requires the human judgment Drucker emphasized.
When an AI can draft a competent legal memo, write serviceable code, or analyze a financial statement, it's not that the AI has achieved human-level reasoning—it's that we're discovering these tasks involve more pattern-matching and less irreducible expertise than we'd convinced ourselves. The tasks that seemed to require years of training and judgment turn out to have been, in retrospect, more algorithmic than we realized.