Why every AI skill on your CV should be checkable

Every AI skill on your CV should be checkable, because AI skills are now so widely claimed that unsupported claims carry almost no weight. An employer reading "experienced with AI agents" cannot tell you from the next candidate. A skill linked to a certificate number, a work sample or a named referee gives them something to confirm.

Why have AI skill claims lost their value?

Because they cost nothing to make. Anyone can write "AI-savvy", "skilled in prompt engineering" or "experienced with AI agents" on a CV. Many people do, with very different levels of ability behind the words. Hiring managers learn quickly that the phrase predicts little.

The problem is worse for AI than for older skills. Established fields have recognised qualifications and shared language. AI skills do not yet, so the same claim can mean an afternoon's experimentation or years of operating agents in a regulated business. Without a way to check, a careful employer discounts all claims equally, which punishes the people with real skill most.

What makes a skill checkable?

Something outside your own say-so that an employer can confirm without relying on you. Four kinds of evidence work.

A certificate checked at source. The employer goes to the issuer, by certificate number or QR code, and sees the credential's current status. A lapsed or revoked certificate shows as such. This is the quickest check for an employer and the hardest for anyone to fake.

A work sample. Something you produced that demonstrates the skill: a set of review decisions with reasons, written up without confidential details. An employer can read it and question you on it.

A referee who can speak to it. A former manager who can confirm that you reviewed held items, ran a queue or led an agent rollout. Name the specific skill when you ask them to act as referee.

A test the employer sets. Some employers run a work sample test during hiring. Offer to do one if they do not.

The distinction between skills you can prove and skills you can only claim is close to the distinction between hard skills and soft ones. Most AI agent skills are hard skills in that sense: they have specific content you can be tested on.

What does a checkable CV look like?

Each AI-related line ties a skill to its evidence. For example:

The third row shows that even knowledge can be made checkable: name the specific thing, and invite a question. The OWASP Top 10 for LLM applications, which includes prompt injection, excessive agency and sensitive information disclosure, gives you precise terms that an interviewer can probe.

Why does this matter more for agent work?

Because agent work involves authority. A reviewer approves held items, and some businesses make approval rights depend on a current certificate. Where that practice applies, the employer is not just hiring a skill; they are deciding who may authorise refunds, claims or payments. They need to check, not trust.

It also matters because of the gap between knowing and doing. Many people know about AI agents. Fewer can show they have made decisions on an agent's work to a standard. The argument that this, not technical skill, is where the gap lies is made in the skill gap is not technical; it is knowing what an agent may do.

What is the trade-off?

Making everything checkable has costs.

Not everything can be checked. Some skills, such as staying calm under pressure, are hard to evidence. Do not drop them; describe them with a specific example and accept that they will be weighed as claims.

Confidentiality limits work samples. You cannot share an employer's customer records. Strip out identifying details or describe decisions in general terms.

It shortens your CV. Removing unsupported claims can leave fewer lines. That is fine. A short list of checkable skills persuades more than a long list of assertions.

Some checks take effort. A referee may be slow to reply, and a work sample takes time to read. Order your evidence so the quickest check comes first: a certificate number an employer can look up in seconds should lead, with slower forms of evidence behind it.

It exposes lapses. A certificate checked at source shows if it has lapsed. Keep it current, or describe it plainly as past training.

What should you do now?

Go through your CV line by line. For each AI-related skill, ask: how would an employer check this without asking me? If there is no answer, either find evidence, rewrite the line as a specific example, or remove it.

Then fill the biggest gap. For most people moving into AI operations, that is evidence of review judgment, which a certificate with a practicum or a work sample provides. The page on what a job-ready skill is helps you decide which skills to evidence first, and the most valuable AI skill for non-technical workers suggests where to focus. The overview of AI skills for job seekers sets out the rest.