How to prepare your career for AI: a practical 30-day plan

If you searched for how to AI-proof your career, the concern makes sense. But no honest list can guarantee which jobs will survive unchanged or disappear by 2030. A role contains different tasks, and a tool’s ability to assist one of them does not tell you what your employer will do.

The ILO’s 2025 research identifies task transformation as the most likely overall effect of generative AI, while recognizing that exposure varies. Anthropic’s March 2026 study also distinguishes theoretical capability from observed use. Neither gives you a personal job-loss date.

A practical goal is to understand your changing work and demonstrate good judgment within it. This four-week plan is a suggested learning exercise. Adjust the pace to your responsibilities; it is not a promise of career protection.

Write down five recurring tasks from a normal week. For each, note the input, the output, who relies on it and what a mistake would cost. Use generic descriptions without customer names, company secrets or confidential documents.

Preparation

What do you gather, summarize, format or draft?

Judgment

Which assumptions, exceptions or trade-offs need to be checked?

Responsibility

Who approves the outcome, explains it or deals with the consequences?

These are lenses for looking at a task, not permanent categories of “safe” and “unsafe” work. The same activity may include all three. Choose one preparation task where you can inspect the result and retain the necessary review.

Keep: a five-task inventory and one task to investigate.

Use the free assessment to find a starting point if you need prompts for this reflection. Its score is a self-reflection index, not a replacement probability.

Use an employer-approved tool with public or invented inputs. Write the quality standard before you start: for example, every action must match the source, each number must reconcile, or every policy statement must point to the relevant passage.

Compare total effort with the usual method, including setup, checking and corrections. Record what went wrong as carefully as what got faster. Keep the test small enough that a failed result gives you information rather than creating a problem for somebody else.

Keep: one before-and-after example, a review checklist and an effort record.

Try our fictional meeting-notes exercise before creating your own.

Choose a skill from a difficulty you actually encountered. That gives your learning a target. The OECD’s research notes that most workers exposed to AI will not need specialist skills in developing AI models. Understanding your own work can be a more relevant starting point.

• The answer used the wrong assumptions.
  Practice problem framing: write a definition, an example and an exception before asking for a draft.
• You could not tell whether the answer was correct.
  Strengthen domain knowledge or data checking. Work through one source-backed example you can independently verify.
• The output was accurate but unhelpful.
  Practice communication: rewrite it for a specific reader and ask what decision they can now make.

If appropriate, ask a colleague to assess an approved, non-sensitive example against your checklist. Ask where it fails and what is missing. Praise alone will not tell you what to practice.

Keep: one improved artifact and one piece of specific feedback.

Review the task, the corrections and the feedback together. Continue if the workflow meets the quality standard at an acceptable cost. Adjust it if one fix is clear. Stop if checking remains unreliable, permission is missing or the work is not useful.

Turn that decision into a concrete learning priority: a task to repeat, a responsibility to discuss with your manager or a skill to practice next. Before buying a course, identify the work sample it should help you produce and check whether the syllabus teaches that skill.

Keep: a one-page evidence note and a next action you can explain.

Should I change careers because of AI?

An exposure score alone is too little evidence. Compare the actual tasks of a possible role, entry requirements, local opportunities and the cost of retraining. Talk to people doing that work and test an adjacent task where possible. This guide does not account for your finances, health, responsibilities or local hiring conditions.

Which jobs can AI never replace?

We cannot give a permanent guarantee. Hands-on work, accountability and relationships affect how a role changes, but do not make every task in it immune. Use concrete role examples, such as our data analyst guide and financial analyst guide, to compare the work rather than treating a title as a safety label.

If you want to understand how our assessment weighs these factors, read the methodology and its limitations.

These sources support the ideas about AI, work and responsible use. The exercises are our teaching examples. They have not been validated as a way to prevent job loss or improve earnings.

• ILO: Generative AI and Jobs — A Refined Global Index of Occupational Exposure

  20 May 2025

  Studies task-level occupational exposure across countries. Exposure describes potential change; it is not an individual worker’s probability of losing their job.

• OECD: Artificial intelligence and the changing demand for skills in the labour market

  10 April 2024

  Research on changing skill demand in AI-exposed occupations. It supports looking beyond specialist AI-development skills, not guaranteeing any skill will preserve a job.

• Anthropic: Labor market impacts of AI — A new measure and early evidence

  5 March 2026

  Combines theoretical capability with observed Claude usage and US labor data. One provider’s activity cannot describe every employer, country or future outcome.