Now, the robots want white collar jobs

AI will do to office work what automation did to factories.

I didn’t want to use social media.

I created a Facebook account for a college class and then forgot about it for years. I skipped the early days of Twitter. I don’t remember when I signed up for LinkedIn. I finally joined MySpace just in time to watch it die.

I had zero interest in these platforms. But what I did want was a job. I spent the Great Recession unemployed in New York. Somewhere in between binging DVDs of “The Office” and “Lost” while eating frozen pizza from Trader Joe’s, I had a realization: If I ever wanted to work in media (or maybe anywhere) again, I needed to get active on social media.

So, I did.

Yep, I was a total noob. But I was trying. As annoying as it seemed to me at the time, my late adoption of social media helped put me on a long road from journalism washout to landing my dream job as a columnist. All it cost me was brain rot.

My adapt-or-die alarm bells are ringing once again with the stunning advances of AI. Except, this disruption is unfathomably greater than social media. We no longer have the luxury of debating AI’s virtues or nursing our feelings. It’s here. The framework for adoption is this: Do you want to have a future? If yes, then you need to use AI.

Axios’ Jim VandeHei put it like this:

Stop downplaying the tectonic shifts that could hit every job, starting next year. Employees need the hard truth that entire classes of jobs could be wiped away, especially if people don't quickly adapt. I recently told the Axios staff that we're done sugar-coating it, and see an urgent need for every employee to turn AI into a force multiplier for their specific work. We then gave them tools to test. My exact words to a small group of our finance, legal and talent colleagues last week: "You are committing career suicide if you're not aggressively experimenting with AI."

I encourage you to read the entire column, because VandeHei has the urgency level exactly right. Notice that his comments aren’t specific to media. VandeHei and Mike Allen followed up with another column Wednesday reporting that “AI could wipe out half of all entry-level white-collar jobs — and spike unemployment to 10-20% in the next one to five years.”

Predictions are hard, as they say. But it’s undeniably, directionally true that AI platforms create efficiencies right now on a scale that eliminates the need to pay for labor or services you might otherwise use. And they’re only getting better. Just last week, Anthropic released Claude 4, and I found the advances so immediately noticeable that I retraced old problems to find new insights.

I’ve been seriously experimenting with AI for about five months now, mostly Claude, which is a large language model similar to better-known ChatGPT. I might write a separate post on how I apply AI to journalism, if there’s a demand for such a thing, but for now I want to share some broader applications:

• I created the logo for this newsletter with AI. I’m very happy with it! I should acknowledge there are ethical concerns about using AI for creative work. My wife, a designer by trade (and someone who has spent even more time with AI than I have) did not share those concerns, since hers was the (unpaid) labor being replaced. She did provide substantial input and guidance.

• I removed a wasp nest from an area where my children play. I wanted no part of this project and otherwise might have called a pest control service. I wasn’t even sure what the nest was. I uploaded a photo to Claude and it identified the nest as belonging to paper wasps, explaining the nest’s state of development (early), how to remove it and how many wasps would be there in a few weeks if I didn’t (50-plus). I hate killing anything, even insects. When I expressed mild uncertainty about removing it, hoping for a way out, Claude responded: “With young children playing in the area, this really isn't a situation where you can safely ‘live and let live.’” Notable: AI has no sympathy for lower life forms.

• I created a detailed trip itinerary. My wife and I are surprising my mom with a trip for her birthday next month. AI created a detailed, well-paced itinerary that could have been the product of a travel agent. I realize people have already been talking about the death of travel agencies for 20 years. But we actually used one in 2018 to plan a trip to Australia. In that case, we had a long list of things we wanted to do but felt overwhelmed tying them together into a practical order. We outsourced the work to a travel agent for the sake of our sanity. Now, I’m confident AI could do that for us.

• I created a home renovation plan. A tree limb went through our garage roof, which has us considering whether this is the right time to take on a larger roof/gutter/siding project that we were hoping to put off for at least a couple more years. I took photos of our house and garage, and uploaded them to Claude, which generated a project including materials, colors and price estimates. The pricing has been remarkably accurate, especially considering contractors are walking around our roof to put together numbers while all I did was take some photos. As estimates have come in, I’ve uploaded the documents to Claude, which has told me which numbers are reasonable and which ones reflect over-charging.

• Additionally, Claude created a multiphase exterior remodeling plan, suggesting architectural details that would enhance the character of our 1955 home. I don’t know if we’ll ever get to any of these things, but the AI-generated plan amounts to design work worth thousands of dollars.

These project-based examples don’t include all the other random times when I just want to learn about the evolution of Major League Baseball rules from the 1990s to present or Calvinism or the English Revolution. AI is replacing my use of search engines, which comes with a whole host of other implications for tech and media.

Yes, AI gets things wrong sometimes (like a reading list published in the Chicago Sun-Times), but I think that gets overblown. I’m not asking AI to do my work for me¹, and you shouldn’t either.

I forget who said this, but I read someone describe AI as a really smart personal assistant who also likes to bullshit. That sums up my experience. We all know humans who are like that and we learn to appreciate their value without dwelling on their faults.

If you want to hate on AI, I’m not going to try to talk you out of it. We need to be clear-eyed about both the risks and rewards of AI. Before we can get there, though, we also need to understand it. We can only understand it by using it.

From my career to my personal life, AI has made me faster and more efficient at a wide range of tasks and analysis. I’m confident that AI won’t replace the specific work I do, but am I 100% confident? Not really. That’s a tension we all need to wrestle with.

Robots have been taking over factory work for decades. Now, they’re coming for the white collar jobs. Some at-risk career paths are predictable (data entry, customer service, administrative roles), but others will surprise us.

We all need to be thinking about whether, or how, we create value that can’t be automated. If you’re not sure whether AI can do the work you do, start using it and find out. Pick an LLM and start feeding it prompts to help with projects.

Those of us with office jobs have a choice: You can either make AI your assistant or your replacement.

1. I do ask AI to edit my work for me. Claude’s take on my headline and subhed: “It will get clicks and accurately represents your core message.” Thanks, Claude, that’s what I’m going for here.