Outside Engineering, AI Skill Is Judgment About Your Own Work

What the Posting Actually Wants

Read the verbs in the posting before assuming the bar involves code. "Uses AI tools to draft campaign copy," "builds models with AI assistance," and "analyzes data using AI" all point at the same underlying test regardless of function: can you tell when the output is wrong for your specific situation. None of them require you to write software.

AI-skill language has moved well beyond engineering postings. Employers already price it into what they advertise outside technical roles: in Lightcast's analysis of its own posting database, postings that mention AI skills run 28% higher in advertised salary than postings that do not, and more than half of postings requesting AI skills in 2024 sat outside IT and computer science. ¹ That is a raw comparison of posting groups, not a controlled estimate of what anyone is paid. The demand is real and it is not confined to the roles you might assume.

The posting language is also the first place to check for whether the tools sit inside your daily work or on its edges. A line asking you to review AI-drafted material before it ships describes a checking role. A line asking you to configure or maintain an AI system describes something closer to engineering, whatever the department. Reading that distinction before you apply saves you from either overselling or underselling what you can already do.

Show Domain Judgment, Not a Tool List

The test for a non-technical role is domain-specific in a way a generic AI tour cannot answer. How do you screen for AI judgment in finance or marketing? describes the version employers are increasingly being told to run: an hour on a task built from the function's own material, with one number in the packet contradicted by a source the candidate has to open.

That is the shape to prepare for and the shape to bring evidence of. In marketing, a segment the model built from a stale definition of "active user." In finance, a reconciliation that balances only because a transposed figure canceled out another error. In operations, a staffing model that assumes a shift pattern that no longer exists. Each example needs the same three parts: what was claimed, what did not survive contact with your own data, and what you changed.

Write the example down before the interview, not during it. Name the claim, the specific detail in your own data that contradicted it, and what you changed as a result. A reviewer who has run this kind of screen before is listening for exactly that shape, and a candidate who can produce it without hesitation stands out from the far larger group who can only describe using AI in general terms.

Why This Reaches Marketing and HR Postings Now

The language has spread faster than most non-technical candidates expected. As of December 2025, in Indeed's US posting data, about 15% of marketing postings and 9% of human resources postings contained AI-related terms, up sharply from a near-zero base a few years earlier. ² A mention in a posting is not proof the role tests for it heavily, but it means the vocabulary is now ordinary in your function, not borrowed from engineering.

In a June 2026 ZipRecruiter survey of more than 1,000 hiring decision-makers, 74% called AI skills a strong advantage or a flat requirement, and half said they expect candidates to already be practical or advanced users. ³ That figure describes what hiring people say they expect, not a guaranteed outcome tied to any single application, and a vague "familiar with AI tools" line answers none of it. Specificity is what the number is actually asking for: which tool, on which task, what you caught.

That spread matters for how you read a posting with no AI language in it at all. Absence is not evidence the function is exempt, only that this particular employer has not written the requirement down yet. The underlying test, whether you can catch a wrong output in your own field, applies whether or not the posting names AI directly. Treat a posting with no AI language as a normal posting, not a safer one, and prepare the same domain example either way.

Bring the Domain, Not the Software

Lead with your function, not your tool list. Describe the task you did in your own field, the AI-produced draft or output involved, and the specific thing about it that only someone who knows the field would catch. That is a stronger credential than a certificate, because it cannot be earned by someone without your domain background.

A discrete choice study of 543 hiring managers and HR specialists found that raising a candidate's relevant work experience from zero to two years moved selection odds within the choice exercise by more than 21 percentage points, well past anything attributable to how a credential was delivered. ⁴ The study measured degrees rather than AI certificates specifically, but the direction supports the same advice here: put the domain example first, and let a tool name sit in the background where it belongs. If you barely use AI, say so and bring the judgment covers the same principle for the interview itself, where the temptation to lead with a tool list is strongest.

This also answers the career-changer's real worry, which is usually not whether they know enough about AI but whether their prior field still counts for anything. It does. The domain knowledge that lets you spot a wrong reconciliation, a bad segment, or an unrealistic staffing model is exactly the credential a generic tool tour cannot fake, and it is the one thing a career-changer already has more of than a recent graduate. Say it in those terms if an interviewer asks why you are changing fields: the subject matter is what makes the check possible, and that does not reset when the job title does.

See what gets scored >