Wisdom that outlasts the algorithm.
THE CURVE [insight]
The bottom rungs of a career were never just grunt work. They were the reps that built judgment — and AI is very good at exactly those rungs.
This week I gave the lunch keynote at the Kentucky Chamber's Young Professionals Summit. I chose to talk about building a career you cannot predict. Then came the Q&A, and the room told me what the talk was really about.
Three questions. None of them was "will AI take my job?"
The first came from someone early in their career: The starter work, the tasks you learn on, is being condensed by AI and pushed further down the line. How will people continue to grow?
The second came from someone who manages interns and co-ops. The interns lean on AI so heavily they can't explain the code it wrote for them. Ask for clarification and the answer is that the chatbot said so. People have said to them, word for word: "Well, it's not wrong because I told it to not be wrong."
The third came from someone a few years into a client-facing role. Their company requires an AI tool that records the meeting, writes the notes, and drafts the follow-up. They described themselves as "very relationship-driven," yet they’re being forced to use a tool that makes the conversation feel hollow.
These are ground truth insights that landed live. Three people, three vantage points, one worry: how do I get good when the machine does the work I was supposed to get good on?
Here is the picture I keep coming back to: A career used to be a ladder, and the bottom rungs were the grunt work. The first draft your manager tore apart. The data pull you got wrong twice. The meeting notes where you had to decide what mattered. No one designed those tasks as training. But they were training regardless. The work was the reps, and the reps built the judgment that moved you up.
That is exactly where AI is best. Therein lies the trouble. The job at the top of the ladder is still there. It’s the rung that got you there that is disappearing.
I told the room that answers got cheap, and your judgment didn't. I still believe it. But judgment has a price, and for most of us that price was paid in reps.
Two pieces of research sharpen the picture. In a 2024 study in Science Advances, Anil Doshi and Oliver Hauser gave some writers story ideas from an AI model. Readers rated those stories more creative and better written, especially the ones from writers who started out less creative. The stories written with AI’s help also came out more alike. Better…and more alike. That is the entry-level trade in one line: your work looks better sooner; it also looks more like everyone else's. 😕
Researchers at the MIT Media Lab have a name for the other half of the bill: cognitive debt. The effort you skip now comes due later, as thinking you never practiced.
A lot of my teaching runs on one line: Offload the toil. Reserve the call. Toil is the effort that takes little judgment. The call is the judgment itself — what you send, what you sign, what you say to the client. Hand the machine the first. Keep the second.
Apply this lens to every task: is this toil, or is this a call?
That line is right for someone 20 years in. For someone two years in, it needs a warning, and I shared the warning in this talk:
Some of the toil is your reps. Hand off every first draft while you're still learning the job, and you skip the practice that builds the judgment.
So offload the toil, but keep enough reps to earn the call.
The fix for the missing rung is choosing, on purpose, which reps you keep. Early in a career, the first draft is where you learn what good looks like. Later, the same draft is toil. A task changes categories as you grow. You just have to notice when it moves.
Offloading used to be a question of efficiency. For anyone still climbing, it is a question of what you are building yourself into.
The harder version of this question is whether the rung continues to exist at all. That one is on the people doing the hiring, and it is the subject of another issue.
The machine can do the reps. It cannot give you the experience.
──────────────────────────────
THE SIGNALS [three reads: unfinished, sharp, owed]
→ Half-Baked.
"It's not wrong because…I told it to not be wrong" is a new kind of mistake. It treats the instruction as the check: the prompt said be accurate, so the output must be. My unfinished question is whether people grow out of this the first time it burns them, or whether it sets as a habit because the first few years never burn them. I don't know yet. The answer decides whether this is a training problem or a hiring problem.
→ Hot Take.
The required-tool question matters most and gets discussed least. Most advice about AI at work assumes you choose what to offload. Plenty of early-career people don't get that choice. But the line between toil and judgment holds even when the tool is required. The notes and the follow-up draft are toil, so let the tool carry them. The first meeting is where you make the call. Your company decides what the machine touches. You still decide what you practice.
→ Confession.
That reps line was in my notes. I cut it on stage for time, and then two of the three questions asked for it. My live answer to the first question was to build a personal brand and be known for something. The person asked how you get the experience, and I answered that you need to to chart your own course and get noticed. This issue is the answer I owed that room.
──────────────────────────────
THE NEXUS [your turn]
What was the grunt work that taught you the most? Who does that work now?
If the answer is "a model," tell me which part you would keep by hand.
Reply if you see this differently. I read every response.
──────────────────────────────
THE MOVE
THE MONDAY MOVE [this week]
Pick one first draft you would normally hand to AI this week. A summary, a memo, the first pass at an analysis.
Write it yourself first. Give it 20 minutes and let it be rough.
Then ask AI for its version and put the two side by side. Write down one thing it caught that you missed. This is your study guide.
Write down one thing you caught that it missed. This is your judgment, on paper.
If you manage people early in their careers, run the same move from the other side. Before one AI-assisted piece of work ships this week, ask its author one question out loud: what would make this wrong? The answer tells you whether they did the rep or skipped it.
That is Responsible Cognitive Offloading for someone still learning the job: offload the toil, keep the reps, you make the call.
THE ASYMMETRIC MOVE [this quarter]
If you lead people, name at least three judgment calls in your function. Not the tasks — the calls. The ones where being wrong costs something and no template tells you the answer.
Then require a human first pass on those three, and only those three. Everything else, hand to the machine with my blessing.
It costs you one meeting and a short list. Skip it, and you find out in three years, when you have a team that can generate anything and verify nothing. That bill comes due quietly, and it comes due on you.
THE DECADE MOVE [this decade]
Start keeping a record of your calls. The decision, what you believed at the time and what actually happened. A line each. Boring for a year, valuable after ten.
Judgment has always been hard to prove, so we've used proxies: the degree, the title, the years served. Those proxies are getting noisy, because more of the work behind them runs through a machine.
A record of decisions and outcomes is the one asset that gets more valuable as everything around it gets cheaper to produce. This approach has been integral to the success of Ray Dalio and Bridgewater Associates. It's also the thing a sponsor can repeat about you, and the thing an agent checking your track record can actually read.
──────────────────────────────
THE COMPOUNDING ASSET [the part you keep]
Three lines from this week's keynote that I'm keeping:
→ "Answers just got cheap. Your judgment didn't."
The whole talk in seven words.
→ "Humans trust ahead. AI agents trust in reverse."
We extend trust and expect it back. An agent checks the record first. Two different orientations, now sharing the same workflows.
→ "Don't get stuck at a local maximum."
A milestone that feels like arrival becomes a rut. Expect to climb down from one peak to reach a higher one.
The talk was "Careers on the Curve," the lunch keynote at the Kentucky Chamber's Young Professionals Summit on September 21. No recording yet. When there is one, I’ll share it here.
──────────────────────────────
THE GROUNDING [return trip]
I write this newsletter on a simple bet: that most of what matters about this technology is happening on a curve, and the people who read the curve early get to make different decisions than the people who wait. I told the room one more thing: when the picture isn’t clear, position yourself for “no regrets moves.” Keeping your reps is one.
This one is for the people on the bottom rungs, and for the people who hired them.
If something here changed how you're thinking, hit reply. I'll respond.
Thank you for reading.
— Chris
Columbia Faculty · University of Louisville Exec Ed · Chief of AI & Strategy at Essential Innovations · Founder, Attainable AI