A Practical Framework for Auditing How Automatable Your Job Really Is

You don't need a research report to find out -- you need about twenty minutes and an honest list of what you actually do.

Most people think about job security at the level of a job title -- "is my role safe." That's the wrong resolution. Almost no job is uniformly safe or exposed; it's a bundle of individual tasks, and each one sits somewhere different on the automation spectrum. The useful audit happens at the task level, not the title level.

Start by listing what you actually did in the last two weeks, as specific verbs: not "manage reports" but "pull last month's numbers into a template," "write the summary paragraph," "decide what to flag to leadership." For each one, ask a single question: does doing this well mainly require finding or arranging information that already exists somewhere, or does it require judgment, context about your specific situation, or a relationship with another person? The first kind is what today's AI tools are already good at. The second kind is what they're consistently bad at, and where the actual research on employer priorities keeps landing too.

The point isn't to panic about the first list or feel falsely safe about the second -- it's to know which is which, on purpose, so you can make a deliberate choice: get faster at the exposed tasks using the same AI tools that are automating them, and get more visible and more skilled at the durable ones. That combination is what shows up, again and again, as what actually holds up.

Also worth reading: AI vulnerability score by job title · how we score AI resilience