Is AI Hollowing Out Your Job Without You Knowing?

Your calendar is full. Your manager just gave you a solid performance review. The work is getting done. By every visible metric, you're fine.

That might be exactly the problem.

Nate B. Jones, who runs the AI News & Strategy Daily channel and a companion Substack, recently put out a 34-minute video that cuts through the most common framing around AI and jobs—the "will it replace me?" question—and replaces it with something considerably more unsettling: how much of what you did last week actually required you?

It's a sharper question. And if you sit with it honestly, it's a harder one to answer than it looks.

The Slow Hollow

The argument Jones is making isn't about sudden displacement. It's about a structural lag that most organizations haven't processed yet—and that most performance systems are structurally incapable of seeing.

His anchor metaphor is the travel agent. Expedia didn't vaporize that profession overnight. Online booking quietly removed the economic rationale for the routine layer of the job—the simple itinerary, the straightforward hotel booking—while the rest of the role appeared, on the surface, to continue functioning. The data didn't show a cliff. Then a downturn came. Then the reckoning happened fast.

"The tech did not have to eliminate every part of the role to change the role," Jones says. "It only had to make the routine booking layer less defensible. And then when the industry hit pressure, the adjustment happened very quickly."

The agents who survived moved toward complexity: corporate travel, emergency problem-solving, luxury itineraries, the kinds of trips where a wrong call has real consequences and a bot's confident-sounding answer isn't good enough. The ones who didn't survive were, in Jones's framing, defending routine booking as a professional identity.

The question he's posing for knowledge workers in 2025 is: which of those are you?

The data he cites gives the question some weight. OpenAI and University of Pennsylvania researchers estimate that roughly 80% of U.S. workers could have at least 10% of their tasks affected by language models. Anthropic's Economic Index puts it more concretely: about 49% of jobs have already seen at least a quarter of their tasks performed using Claude. Microsoft researchers, analyzing 200,000 Copilot conversations, found that the most common work people bring to AI is information gathering and writing—not exotic edge cases, just the daily connective tissue of most office jobs.

None of this says "your job is gone." All of it says: pieces of your job are being quietly absorbed, and most organizations haven't restructured around that yet.

The Four-Letter Audit

Jones's practical contribution is an audit framework he calls T, C, L, D. The exercise: open the last ten business days of your calendar, your sent mail, your Slack, your docs. Tag every discrete item—not the project, not the role, the item—with one of four letters.

T is Theater. Work that exists because the organization needs to perform coordination, not because it produces examined value. The status meeting where nothing actually changed. The deck that exists because a senior person needs something to flip through. The alignment call that produced no alignment but allowed everyone to say alignment was attempted. Jones is pointed about why this layer exists: "Large organizations create theater because theater is legible. It gives people something to point at. It creates the appearance of coordination. It lowers social risk." The uncomfortable implication is that AI doesn't need to make theater good—it only needs to make it adequate, which is what theater already was.

C is Commodity. Real work, real value, just not work that requires you specifically. Summarizing, routing, applying known rules to known situations, writing the status update that someone does read but anyone could have written. Jones is careful not to be dismissive here—commodity work is often genuinely hard-won skill. "The problem is that a skill can be real and still become less scarce. Markets don't really care how hard something was to learn." The test he offers: could you write a spec and have someone else in your organization produce an output that's roughly as useful? If yes, it's probably a C.

L is On the Line. The uncomfortable middle. Pattern recognition over structured inputs. Relationship management that depends on institutional history you carry. Editorial calibration in formats you know well. Work that "used to feel hard and now feels a little too repeatable." Jones suggests tagging these quickly and moving on—the point isn't precision, it's seeing where the line is moving.

D is Durable. Work where the output depends on something you can't fully specify in advance. You saw that the stated problem wasn't the real problem. You read a room. Your presence changed the outcome in a way that goes beyond being faster or more organized or more available. Jones's definition is deliberately hard to operationalize: "D is the work where you are carrying context and taste and courage and calibration or pattern recognition that can't be cleanly specified before the work starts." The honest admission buried here is that durable work often looks invisible from the outside—which is its own problem in organizations that measure visible throughput.

What the Count Reveals

Jones is candid about what he expects most people to find: more T and C than they wanted, less D than their self-image suggested. That last part is where the framework gets genuinely uncomfortable.

"Your professional identity is often built around the durable part of the work," he says. "You think of yourself as the person who can read the room, who can diagnose the customer, who can spot the bad deal. And maybe you are that person. But the audit asks a colder question. How much of the last two weeks did you spend on that? Not how much of your self-image depends on it. How many hours?"

This is the gap the audit is trying to surface: the distance between the story you tell about your job and the actual composition of your week. For people whose D number is genuinely high, the audit is reassuring. For people who have been coasting on the narrative that their judgment is irreplaceable while spending most of their time on work that a well-prompted model could do at three in the morning—it's a different conversation.

There's a real tension here that Jones doesn't fully resolve, though it's worth naming: not all organizations use durable work even when it's available. The workers who do the genuinely non-replicable thinking sometimes have the least visible output—and performance systems, as Jones correctly notes, are optimized for visible output. The audit might tell you that you're doing more D-level work than you thought, and your organization might still not be structured to recognize or reward it. The framework is individual; the problem is structural.

There's also the question of what "durable" looks like in five years, not today. The L-category work—things that feel like yours, but you'd struggle to articulate exactly why—is probably where the most interesting competitive pressure is building. The dividing line between "pattern recognition that can't be cleanly specified" and "pattern recognition that an improved model handles easily" has been moving consistently in one direction.

The Legibility Trap

One of the more interesting threads in Jones's argument is what he calls the legibility paradox. Organizations reward visible throughput. Performance reviews measure what they can measure. Which means there's a period—possibly a long one—where someone can be doing genuinely important work that doesn't register, while someone else is generating high-volume commodity output that looks great in a sprint review.

The trap for knowledge workers right now, Jones suggests, is that AI tools let you become dramatically more productive at exactly the kind of work whose value is collapsing. You can write more status updates faster. You can produce more first drafts of documents. You can synthesize more meeting notes. The metrics go up. The economics of the role quietly deteriorate.

"Leaders who pour recovered AI time into more commodity work are becoming twice as productive at the part of their job whose value is collapsing," Jones observes, "and it feels like progress because old systems still reward visible throughput."

The audit doesn't solve this. But it makes the shape of the problem visible before the organization's next moment of pressure forces it into the open—the budget freeze, the reorg, the recession that finally prompts the question that's been quietly forming: why is this role bundled this way?

Travel agents who were still defending routine booking when that question arrived didn't have a great answer. The ones who had already moved didn't need to.

---

Marcus Chen-Ramirez is a senior technology correspondent for Buzzrag. He spent eight years as a software engineer before moving into journalism.