Define it first

When CoBlack scores an opening against a member’s Career Capability Map, the listing’s requirements are read against what the person has actually done. One line gives that match nothing to work with: ‘AI fluency required’, with no word on what the phrase means. Indeed asked employers and job seekers about that line this summer. The answers show how rarely it has been written down.

Ten people, ten answers

On October 6, 2026, Hannah Calhoon, VP of AI at Indeed, published a newsroom release titled “What Is AI Fluency? Closing the Gap Between Hiring Intent and Job Reality”. Its opening observation is the whole problem in one sentence: “if you ask ten people what ‘AI fluent’ actually means, you’ll probably get ten different answers.”

The numbers behind it come from a YouGov survey that Indeed commissioned, run from July 23 to August 13, 2026 across the US, UK, Canada and France. The published findings are the US results: 1,007 active and passive job seekers, and 500 employers described as hiring decision-makers at mid-market and enterprise organizations with 500 or more employees. Indeed publishes no margin of error. It is an attitude survey, and we read it as one.

The attitudes diverge. In the release’s words, “74% of employers say using AI effectively is a fundamental skill. But job seekers haven’t caught up, with only 34% considering it important to their current roles.”

A requirement most firms have not written down

Here is the finding we keep returning to. About a quarter of employers have a documented, consistent definition of AI fluency across roles, the release reports. The other 73 percent have no written definition that holds across their roles.

Candidates notice. Nearly one-third of job seekers, the release says, see AI requirements in job listings without any explanation of what is actually expected. Calhoon does not soften what follows: “When expectations are fuzzy, everyone loses. Companies end up evaluating candidates inconsistently across teams, and great applicants walk away because they aren’t sure if they have the ‘right’ kind of AI experience.”

One caution about what this does and does not show. This survey does not measure whether anyone was rejected for lacking AI fluency. It measures what people believe, not what happened to applicants, and we will not stretch it further. What it does establish is enough: the line appears on listings, and most employers have not documented what it means.

That is the belief this piece exists to state. A requirement nobody has defined is not a fair thing to screen a person on. If you list it, define it.

The word has definable parts

The release itself shows the definition is within reach. Asked what AI fluency consists of, employers named concrete tasks: “While 68% cited using generative AI for tasks such as writing, research and summarization as a component of AI fluency, 58% cited reviewing AI output for accuracy, quality or bias.” Calhoon draws the line that matters: “Strong AI fluency isn’t just knowing how to produce an answer, but whether that answer is any good.”

The release also offers a definition drawn from the research, and reports that it earned agreement from 78 percent of employers and 69 percent of job seekers. Its advice to employers follows directly: “replacing vague requests for ‘AI expertise’ with the specific tasks.”

Until listings catch up, the evidence has to do the defining. When a posting says AI fluency and nothing more, the strongest answer is a specific thing done: drafting with a generative tool, reviewing a model’s output for accuracy, automating a workflow. That does not get anyone hired. It does replace an undefined word with a defined act, and it leaves the employer less to guess at.

The bar we write down

We cannot make an employer define a requirement. Employers write their own listings, and an undefined line stays undefined until they define it.

What we can do is refuse to screen anyone on a bar we have not written down. CoBlack sources openings only from validated employer career pages and ATS feeds. We match on capability, not titles or keywords: each member has a Career Capability Map built from what they have actually done. We score every opening against that map and apply only at 60 percent or higher. The bar is written down and it is the same for every opening. The member sees every opening we matched with its score beside it, and the score explains itself. Auto Resume writes a fresh resume for each opening from the member’s own evidence; it never invents a skill the member has not shown. The interview stays human. We walk beside the job seeker, never in their place.

A line nobody has defined cannot be fairly passed, and it cannot be fairly failed. A requirement earns its place on a listing the moment someone writes down what it means.