The Five-Year Version of Your Job

What this is for

The question underneath most of the AI anxiety among people with real careers is not "how do I use this". It is "is the thing I spent twenty-five years getting good at about to stop being worth anything".

Nobody answers that honestly. The optimists say your experience is more valuable than ever, which is a reassurance, not an argument. The doomsayers say your job is going away, which sells newsletters. Neither has looked at what you actually do all day.

This does. You break your role into its real tasks, sort each one into four buckets, and read the pattern. It takes about forty minutes and it is uncomfortable in the middle. What you get is a specific read on your own role rather than a general feeling, and a short list of things that just became worth more.

Before you start

• Forty minutes and somewhere you will not be interrupted.
• Your calendar for the last month, open. You are going to need it, because nobody can list their own job accurately from memory.
• A willingness to put something you are proud of into the "gone" column. If everything you value lands in "protected", you have not done the exercise, you have performed it.

One framing note that matters more than any prompt below. Tasks disappear. Jobs mostly do not. Roles are bundles of tasks, and bundles get rewoven rather than deleted. The useful question is not whether your job survives, it is which parts of it are still load-bearing in five years and whether those parts are enough to build on. That is answerable. "Will AI take my job" is not.

Step 1: List the tasks, properly

Open your calendar for the past month and write down what you actually did. Aim for fifteen to twenty items. Two rules:

Be specific enough to be checkable. "Strategy" is not a task. "Wrote the quarterly plan and defended it to the exec team" is two tasks and both can be examined.

Include the invisible ones. The things not on your calendar are often the ones that matter most: the corridor conversation that stopped a bad decision, the judgment call you made in four seconds because you had seen it before, the person you calmed down. These are usually where the protected work hides, and they are always missing from the first draft of the list.

If you finish with fewer than twelve items you have described your job title rather than your job. Go back to the calendar.

Step 2: The four buckets

Sort each task into exactly one. The definitions are deliberately narrow, because a loose definition lets everything drift into the comfortable bucket.

Gone. A capable person using current tools does this acceptably in a fraction of the time, and the quality difference no longer justifies the cost difference. Note the test is not "AI does this as well as I do". It is "as well as this task actually needs", which is a much lower bar and the reason this bucket is bigger than people expect.

Changed. The task remains yours, but the work inside it moves from producing to directing and checking. You stop writing the first draft and start deciding whether the draft is right. Same responsibility, different daily activity, and usually less of your time.

Protected. Someone would object if a machine did this, or the value is specifically that you did it. Three grounds count: accountability, because you are the one who answers for it; relationship, because it had to come from you; and judgment that runs on context nobody has written down. Be strict: "I enjoy it" and "I am good at it" do not qualify.

Newly worth more. The parts of your job that get more valuable because everyone else now has these tools. This bucket is the reason to do the exercise and it is the one people leave empty on the first pass. Prompts below will push on it.

Step 3: Read the pattern

Count the buckets, then ask three questions.

How many of your tasks landed in "gone"? If it is more than half of them, your role as currently defined is under real pressure and the honest response is to move deliberately rather than wait. That is not a catastrophe; it is a timeline. A count is a rough proxy for time, so if the gone tasks are all ten-minute jobs and the protected ones eat your afternoons, weight it accordingly before you read anything into the number.

What is actually in "protected"? Look at the reason, not the task. Most protected work is protected for one of three reasons: accountability, relationship, or judgment that depends on context nobody has written down. Knowing which of the three is yours tells you what to deepen.

Is "newly worth more" empty? If so, you have almost certainly missed something rather than having nothing. When a capability becomes cheap and universal, the scarce complement to it becomes more valuable. When everyone can generate a competent draft, knowing which draft is right is worth more. When everyone can produce analysis, having the relationships to get it acted on is worth more. Taste, judgment, accountability and access do not get cheaper when production does. They get scarcer in relative terms.

The prompts

Prompt 1: Sort the list without flattery

Use either ChatGPT or Claude.

ROLE: You are a labor economist who studies task-level automation.
You are precise and you do not reassure people. You have no stake in
whether the news is good.

CONTEXT: My role: [TITLE AND FIELD, e.g. director of clinical
operations at a 300-bed hospital]. Years in it: [N]. Below are the
tasks that fill my actual week.

TASKS: [PASTE YOUR LIST]

CONSTRAINTS:
- Sort each task into exactly one of: GONE, CHANGED, PROTECTED,
  NEWLY WORTH MORE. Use the harder judgment where it is borderline.
- GONE means current tools do it to the standard the task actually
  requires, not to my personal standard.
- Do not put something in PROTECTED because it needs experience.
  Experience alone is not protection. Name the specific mechanism:
  accountability, relationship, or undocumented context.
- Do not soften the result and do not add an encouraging summary.
- Where you lack information to judge, say so and ask the question
  you would need answered.

OUTPUT FORMAT: A table with Task, Bucket, and a one-line reason
naming the mechanism. Then a count of how many tasks landed in
each bucket. No closing encouragement.

Prompt 2: Argue the other side

Run this in the same conversation, on the output of Prompt 1.

ROLE: Same economist, now arguing against your own sorting.

CONSTRAINTS:
- Take the three tasks you marked PROTECTED that you are least sure
  about and make the strongest case that they belong in CHANGED or
  GONE within five years. Be specific about what would have to be
  true.
- Then take the three you marked GONE and make the strongest case
  that a version of each survives, and name precisely which part
  survives and why.
- Do not split the difference. Argue each position properly.

OUTPUT FORMAT: Six short arguments under two headings. Then one
paragraph naming which of your original calls you now think was
weakest, and why.

Prompt 3: Find what got more valuable

ROLE: Same economist. You are now looking only for scarcity.

CONTEXT: Use my task list and your sorting from above.

CONSTRAINTS: When a capability becomes cheap and widely available,
the things that complement it become relatively scarcer and more
valuable. Identify which parts of my role are complements to the
capabilities that just got cheap in my field. Be concrete: name the
capability that got cheap, and the specific thing I do that is worth
more as a result. Do not list generic answers like "critical
thinking" or "creativity". If my list genuinely contains nothing in
this category, say that plainly rather than inventing something.

OUTPUT FORMAT: Up to five items. Each one: the capability that got
cheap, the complement in my work that gained value, and one sentence
on how I would deliberately do more of that complement in the next
six months.

Prompt 2 is the one people skip and the one that makes the exercise honest. A single pass gives you a sorting shaped by how you feel about your job that day.

What to do with the answer

Three moves, in order of how much they matter.

Stop defending the "gone" column. Time spent protecting work that no longer needs you is time not spent on the rest. This is the hardest part, because the gone column usually contains something you are known for.

Deepen whatever is in "newly worth more". This is the whole return on the exercise. If the answer is that your read on which analysis matters is now scarcer than the analysis, then get closer to the decisions and further from the production.

Name your protection mechanism and invest in it. If your protection is relationships, the answer is more contact and fewer memos. If it is accountability, it is being the person who signs, visibly. If it is undocumented context, write some of it down and become the person who holds the map.

This is not a five-year plan. It is a read on where your role's weight is shifting, which is the input to a plan. Run it again in a year; the buckets move, and they move faster than people expect.

Where this goes wrong

• You list your job description instead of your week. The job description is aspirational and general. The calendar is what you actually did, and only one of them can be sorted honestly.
• You put everything you enjoy in "protected". Enjoyment is not a moat. The sorting only tells you something if it can tell you something you do not want to hear.
• You skip the counterargument prompt. One pass produces a sorting shaped by your mood. The argument against it is where the real information is.
• You conclude your job is fine because no single task is fully gone. Roles rarely vanish in one move; they thin out. Half your week moving to "changed" is a significant result even if nothing is fully gone.
• You do this once and file it. The line moves. A sorting from two years ago is a historical document, not a read on where you stand.

This is the strategic version of a question that also has a tactical version. What to Delegate to AI vs. What to Keep is the same judgment applied to this week's work rather than to your role's next five years.

The 2-minute version

Open your calendar, list fifteen to twenty things you actually did last month including the invisible ones, and sort each into gone, changed, protected, or newly worth more. Be strict about protected: name the mechanism, which is accountability, relationship, or context nobody wrote down. Then argue the opposite case on your least certain calls. Read the "newly worth more" column, and if it is empty, look again, because it almost never is.

---

More where this came from. The resource library has 23 free guides like this one, written for people with real careers rather than developers.

Weekly, on Tuesdays. The GenXcelerate newsletter: practical AI intelligence for people who already know how to do the job. Subscribe free.

Experience Is the API. GenXcelerate