How to write AI skills into a job description

A useful AI requirement in a job description names a task, the standard the output must meet, and how the person checks it. "Proficient with AI tools" fails that test. Nobody can score a candidate against it, and two interviewers will read it two different ways. This guide shows a four-part pattern, with examples for six roles.

Why job descriptions now carry AI lines

AI requirements are no longer limited to technical jobs. Lightcast analyzed more than 1.3 billion job postings. It found that 51% of postings asking for AI skills in 2024 were for roles outside IT and computer science (Lightcast). Human resources postings showed the fastest growth of any sector, at 66%.

Employers say they weigh these skills. In Microsoft and LinkedIn's 2024 Work Trend Index, 66% of leaders said they would not hire someone without AI skills. The survey covered 31,000 workers in 31 markets (Microsoft). The same report found that only 39% of employees who use AI at work had received training from their company.

Those two numbers explain many weak postings. Employers want the skill, but many have not defined it. The posting then repeats a phrase from another posting.

What a vague AI requirement costs you

• It cannot be tested. "Familiarity with generative AI" has no pass mark, so screeners fall back on keywords in a resume.
• It screens on tool names. A candidate who drafts reports in Claude may skip a posting that says "ChatGPT expert", though the skill transfers.
• It leaves out judgment. Checking AI output is a separate skill from producing it, and a vague line says nothing about checking.
• It makes the interview inconsistent. Each interviewer invents a meaning, so candidates face different bars.

The four-part pattern

Write each AI requirement as one sentence with four parts. Keep it to the work the person will do in the first six months.

1. The task: the piece of work AI helps with, such as first drafts of release notes or summaries of customer calls.
2. The input: what the person gives the AI, such as the brief, the data, the policy or the earlier thread.
3. The standard: what good output looks like, such as accurate figures, the house style, or no customer data in the prompt.
4. The check: how the person confirms the output before it leaves their hands, such as comparing totals to the source file.

The four parts line up with skills an assessor can observe: asking well, giving context, checking the work and keeping data safe. Emeri's standard defines six such skills, and the AI skills matrix shows them at four levels.

Examples by role

Each example shows a common line, then a rewrite that follows the pattern. Adapt the details to your own work.

• Marketing. Common: "Experience with AI content tools." Rewrite: "Draft campaign copy variants with AI from a brief. Edit them to brand voice and check each product claim against the spec sheet."
• Customer support. Common: "Comfortable with AI chatbots." Rewrite: "Review AI-suggested replies before sending, correct policy errors, and flag answers the knowledge base does not support."
• Finance and operations. Common: "Advanced AI and Excel skills." Rewrite: "Summarize monthly variance reports with AI. Reconcile each figure to the source ledger before sharing."
• Recruiting. Common: "Leverage AI to source candidates." Rewrite: "Draft outreach and interview guides with AI. Keep candidate data out of unapproved tools, and review drafts for biased wording."
• Legal and compliance. Common: "AI-savvy." Rewrite: "Use approved AI tools to produce first-pass contract summaries, and verify every cited clause and authority against the original document."
• Data and analytics. Common: "Prompt engineering skills." Rewrite: "Write and explain SQL with AI. Test each query against a known result before its output goes in a report."

For more role detail, see the role guides under AI skills by role, such as AI skills for marketing and AI skills for support.

Essential or preferred

Decide whether each AI line is essential or preferred before you advertise. Put it under "required" only if the person cannot do the job without it from day one. Many AI skills can be taught on the job, so "preferred" is often the honest label.

The label carries weight in the United States. Under the Americans with Disabilities Act regulations, a job description written before advertising or interviewing is evidence of which functions are essential (29 CFR 1630.2). An AI line marked essential should describe work the role does. It should not describe a habit the team happens to like. This is general information, not legal advice.

Tool names: list them, but do not screen on them

Naming the tools your team uses helps candidates. It tells them what they will work with. Write it as context, not as a gate: "We use Microsoft Copilot and Claude. Experience with any comparable assistant is fine."

Screen on a tool only when the role administers or builds on it. An analyst who will maintain your team's AI workflows may need a named platform. A sales representative who drafts follow-up emails does not.

Say how you will assess it

Each AI line should map to one step in your process. Candidates prepare better when they know the step. Interviewers score more consistently when the step is fixed in advance.

• A structured interview question about a real piece of AI-assisted work. See AI interview questions by role.
• A short work sample, done with the tools the role uses. See how to design an AI skills test for hiring.
• A credential, if you accept one. Name what it measures, and accept equivalent evidence such as a portfolio of work.

If you accept credentials, check them the same way for every candidate. How to verify a candidate's AI certification covers the steps.

Where Emeri fits

Emeri sells an AI proficiency certification. Candidates are assessed on real work tasks and receive a band from E1 to E4, described on the bands page. A posting can name a level, such as "E2 or equivalent", if the employer accepts other evidence too. The employers page explains how an employer reads a result. A candidate can take the free 15-minute check first.

A checklist before you post

1. Each AI line names a task, an input, a standard and a check.
2. Each line describes work the role does in its first six months.
3. Required and preferred lines are labeled on purpose.
4. Tool names appear as context, with comparable tools accepted.
5. At least one line covers checking output or protecting data.
6. Each line maps to an interview question, a work sample or an accepted credential.
7. Two interviewers agree on what a strong answer to each line looks like.

Job seekers read the same postings from the other side. AI skills in job postings explains what candidates look for in these lines.