The dominant mode of professional engagement with AI is reading about it. Articles, podcasts, newsletters, conference talks. A year of consumption produces a generic worldview about “AI’s impact on work” — a story that may or may not apply to your work.

There’s a faster, sharper, more useful approach: pressure-test your actual career capital against current AI in three structured experiments. Six weeks. Real evidence. The output is not a worldview; it’s a map of your specific situation.

“You don’t future-proof a career. You pressure-test it. Three experiments produce more strategic signal about AI than a year of articles.”

This piece is the framework for those three experiments — what to run, how to run them, what to do with the results.

The Pressure-Test Frame (vs. “Future-Proofing”)

“Future-proofing” is the wrong frame for the AI question, in two specific ways.

It assumes a known future. Future-proofing implies you can identify which skills will be safe and stockpile them. AI is moving too fast and unevenly for this; what’s “safe” today may not be in 18 months, and what looks exposed today may not be either.

It implies a defensive posture. Future-proofing positions you as protecting against decline. It doesn’t ask which parts of your career capital might compound more under AI — and for many professionals, those compounding parts are where the largest opportunity lives.

Pressure-testing inverts both moves. It assumes the future is uncertain and tests your actual capital against current AI to learn how each part behaves under pressure. The output isn’t predictions; it’s a posture of structured curiosity and a map of where your career capital is durable, where it’s exposed, and where AI accelerates it.

This is what every other engineering discipline does: pressure tests don’t predict failure modes; they reveal them. Career strategy benefits from the same approach.

The Three Experiments

Each is two weeks, time-boxed, with a specific hypothesis. Run them sequentially over six weeks; each one’s findings inform the next.

Experiment 1: The Replacement Test

Hypothesis to test: “A meaningful slice of my work — one I’d previously assumed required my expertise — can now be done well by current AI in a fraction of the time.”

The setup: identify one specific output you produce regularly that you’d classify as “probably exposed” if pushed. Examples: a structured analysis you do weekly, a kind of report or summary, a category of first-draft writing, a data-massaging task, a research compilation.

The execution: spend two weeks producing this output in two parallel modes — once with your standard process (no AI), once with AI doing as much as possible (you reviewing and finalizing). Track time spent and qualitatively assess output quality on both.

The decision criterion: at the end of two weeks, ask: did AI produce an 80%+ quality version in less than 30% of the time? If yes, that slice of your work is genuinely exposed — your value here came from execution speed on a well-specified task, and the speed differential is collapsing. If no, the slice is more durable than you thought; understand specifically why AI fell short, because that gap is your actual capital here.

What you learn: which parts of your professional value depend on speed-on-clear-spec work — the most exposed category in the career capital model.

Experiment 2: The Augmentation Test

Hypothesis to test: “A part of my work I’d classified as ‘judgment-shaped’ produces meaningfully better output when AI is integrated as a co-pilot — and the bottleneck is now my judgment rather than my execution.”

The setup: identify a different output, one where you’d say your value comes from judgment, taste, or domain expertise rather than execution speed. Examples: strategic analysis, framework design, written argument, complex synthesis, design decisions, customer insight extraction.

The execution: spend two weeks producing this output with AI integrated as a working co-pilot — not for first drafts, but for argument-stress-testing, alternative framing, gap-finding, evidence assembly. Compare the AI-augmented output against your typical unassisted output.

The decision criterion: is the output meaningfully better — sharper, more thorough, more durable, more correct? If yes, that part of your career capital is compounding under AI. The work gets better when you can direct AI well, and the value-bottleneck is now your judgment, which is exactly where you want it. If no, examine why — usually it’s a sign that your prompting hasn’t matured, or that this slice is actually less judgment-shaped than you thought.

What you learn: which parts of your work get more leveraged under AI — and where the strategic move is to invest in becoming the person who directs AI well in your specific domain.

Experiment 3: The New-Surface Test

Hypothesis to test: “There’s a problem in my domain that AI alone cannot solve well — and articulating why AI fails reveals the durable surface of my career capital.”

The setup: pick a hard problem in your domain — the kind you’d ask a senior colleague to weigh in on, not a clear-spec task. Examples: a strategic question with no obvious right answer, a customer situation requiring contextual judgment, an organizational dynamic, a product decision with conflicting evidence, a relationship issue.

The execution: spend an afternoon (not the full two weeks — this one resolves faster) trying to solve the problem with AI. Then articulate, specifically, why AI’s output falls short. What’s the missing element? Context the AI can’t access? Judgment based on years of pattern-matching? A relationship requirement? A taste call where the criteria can’t be specified?

The remainder of the two weeks: identify three other examples of similar AI-resistant problems in your work. The pattern across them is the signal.

The decision criterion: the gap AI couldn’t bridge is a description of your durable capital. If the gap is substantive and recurring, that capital is what your career should compound around. If the gap is small or intermittent, the slice may be more exposed than you thought.

What you learn: what your actual moat is — the part of your professional value that compounds more, not less, as AI gets better.

How to Run Each in Two Weeks

Time budget per experiment: 4-6 hours of focused work, spread across two weeks.

Week 1, days 1-3: define the slice, set up the parallel process, do the first round of work. Week 1, days 4-7: complete one full cycle of output in the experimental mode. Week 2, days 1-3: complete a second cycle (eliminates first-time-through artifacts). Week 2, days 4-7: evaluate, write the findings as 200 words, decide.

Don’t skip the writing-up step. The 200 words of findings — what worked, what didn’t, what you now know about your career capital — are the durable artifact. Without writing them down, the experiment dissolves into ambient feeling within a week.

Capture each set of findings in your Strategic Narrative or weekly review. The cumulative effect is a real map of your AI exposure across three dimensions: replacement, augmentation, and new-surface.

What the Data Tells You

After all three experiments, you’ll have one of three rough patterns:

Pattern A: the work is mostly exposed. Replacement test confirms a large slice is genuinely automatable; augmentation test shows modest gains; new-surface test reveals a small moat. The strategic implication: redirect. Use the decision matrix and design a 90-day migration toward a direction where your durable capital is more central.

Pattern B: the work is mostly durable. Replacement test reveals limited automatability; augmentation test shows real gains; new-surface test reveals substantial moat. The strategic implication: deepen. Invest in becoming the person who directs AI well in your specific domain, repositioning from operator to operator-of-AI. Most of your career capital is on the right side of the disruption.

Pattern C: mixed signal. Some slices exposed, some durable, some unclear. The strategic implication: portfolio reshape. Drop the exposed work, deepen the durable work, and run a focused investigation on the unclear slices. This is the most common pattern, and the work is to make a clean split rather than continuing to invest equally across all of it.

The patterns aren’t recipes. They’re frames the data points you toward. The actual move depends on your Strategic Narrative, your direction, and the specific shape of the experiments’ results.

After the Three: Designing the Strategic Response

The three experiments are not the end of the work. They produce evidence that updates your career strategy — and the strategy update is what compounds.

After running the three:

  • Update your Strategic Narrative. What’s true about your direction now that the experiments have produced real data? What season are you actually in? What’s the next move?
  • Design one follow-on experiment. Based on what the three revealed, what’s the highest-leverage next two-week investment? Often this is a focused deepening of a compounding capital surface, or a focused migration test toward a more durable direction.
  • Schedule the next pressure-test cycle. AI is moving. Your map will go stale. Run a fresh three-experiment cycle every 9-12 months, or sooner if you sense a major capability shift in AI that affects your work.

The professional who runs this every 9-12 months has a current map of their AI exposure and a continuously updating strategy. The professional who reads articles for the equivalent time has a generic worldview and uncertain action.

This is not extra work on top of your career. It is your career strategy in a world where the disruption is real and the response has to be evidence-based. The strategic version of “what should I do about AI” is “run the three pressure tests, then update the narrative.”

Frequently Asked Questions

What’s an AI career pressure test?

An AI career pressure test is a structured experiment that produces real evidence about how AI affects a specific part of your career capital — replacement risk, augmentation potential, or new-surface durability. Each test is two weeks, time-boxed, with a specific hypothesis. Three tests together produce a working map of your AI exposure that’s grounded in your actual work, not in articles about other people’s work.

How long does each experiment take?

Two weeks per experiment, with about 4-6 hours of focused work spread across the period. Less than two weeks doesn’t produce real signal; more than two weeks lets the experiment dissolve into ambient curiosity. Time-boxing is what makes this useful — the goal is evidence, not exhaustive coverage.

Can I run all three experiments at once?

Possible but not recommended. Running them sequentially — one experiment per two-week block — lets each one’s findings inform the next. The replacement test often surfaces unexpected patterns that change how you’d design the augmentation test. Running in parallel produces less coherent data and more confusion.

What does “AI-resistant” actually mean?

It means a part of your career capital where AI improves output rather than replacing the operator, where the value-bottleneck is your judgment or relationships rather than your execution speed, and where the work would degrade in quality if the human were removed. AI-resistant is not a fixed list of skills — it’s a property of how a specific person’s work interacts with current AI capability. The pressure tests measure this for your situation.

What if all three experiments say my career is exposed?

Then you have valuable strategic information. Three options. One: redirect — design a 90-day migration toward a direction where your judgment, taste, or relationships are more central. Two: augment — invest in becoming the person who directs AI well in your current domain, repositioning from operator to operator-of-AI. Three: repackage — find the parts of your existing capital that the experiments revealed as compounding (there are almost always some) and rebuild your career story around them. Exposure is information; the strategic move comes after.