A job description needs to explain two things clearly: what the person will do and what skills they need to do it. When those two lists are disconnected, qualified candidates can be screened out, and the role becomes harder to understand.
A modern job description should connect each required skill to the work it supports. Responsibilities should describe the actual work, while skills should explain what someone needs to perform that work successfully.
Why Skills Belong Up Front
Skills are what a candidate can show they do. Most fall into three groups:
- Technical: SQL, Figma, Python, operating a forklift
- Interpersonal: clear writing, managing stakeholders, negotiation, coaching
- Domain: B2B SaaS pricing, supply chain forecasting, healthcare compliance
The mix a role needs changes faster than its title does. A description built on a fixed list of tools goes stale within a hiring cycle or two. One built on durable skills lasts longer and attracts people whose resumes look unconventional but who can do the work.
Drop the degree unless the work requires it
Skills-first hiring judges candidates on evidence that they can do the job, ahead of where they studied or which titles they held.
The Burning Glass Institute’s research on the emerging degree reset found employers across many occupations removing four-year degree requirements, which widened the pool of qualified applicants and made it more diverse. IBM publishes its skills-first approach, and its postings are a useful reference for how a large employer words requirements without leaning on credentials.
Where formal training does matter, “BS or equivalent practical experience” keeps the standard and still lets the operations analyst from the opening apply.
Wording That Shrinks the Applicant Pool
Some qualified people rule themselves out before they reach the requirements. Words like “dominant,” “competitive,” and “rockstar” do a lot of that work. The job title can also shape how people read the role before they get to the requirements, so it should match the actual work and level of responsibility. Use a tool that shows that gender-coded wording in job ads changes who applies, and it flags those words in any draft pasted into it.
Two swaps fix most postings. Replace “ninja” and “guru” with the actual job title. Replace “fast-paced environment” with a description of the pace, such as “most projects ship within two weeks.”
Why Responsibilities Still Define the Role
Responsibilities tell a candidate what the job is: the outcomes it owns, the decisions it makes, and what an ordinary week involves. A good set lets someone picture their first 30, 90, and 365 days and decide whether they want them.
The effect lasts well past the hire. Gallup’s work on clear expectations finds that knowing what is expected at work is one of the strongest drivers of engagement and performance, and only about half of employees strongly agree that they do.
Vague responsibilities produce the opposite result. Compare two lines written for the same role:
- “Handle various projects as needed”
- “Run the monthly vendor review for the three largest logistics partners and report savings to finance”
The first line can mean anything, including a second full-time job nobody mentioned in the interview. New hires who find that out in month two tend to be gone by month six. “Wear many hats” usually signals the same thing in friendlier wording.
Jeffrey Zhou, CEO and Founder of Fig Loans, a lender that offers small-dollar installment loans to people with limited or no credit history.
He says, “In a regulated business, a vague job description is a risk. If a role touches lending decisions or customer data, the description has to say which decisions that person owns and which ones they escalate. Candidates can read that and judge whether they’re comfortable with the responsibility, and the hiring team knows what to test for in interviews. Specific responsibilities also shorten onboarding, because the new hire’s first month is already written down.”
Tie Each Skill to an Outcome
Skills and responsibilities work best in the same sentence. Before writing those sentences, cut both lists down.
Essential skills: four to six
List only the skills that predict success in the first year. For most roles, that comes to four to six. Everything else goes in a short “nice to have” list. A long wall of must-haves makes qualified people decide they fall short, and the hiring team ends up screening a smaller, less varied pool.
Core responsibilities: three to five
These are the outcomes the person must own. Occasional tasks and shared duties can sit below them in a line or two. They shouldn’t influence who gets hired.
Merged statements
A merged statement names the outcome, the skill or tool used to reach it, and a measure where one exists. A few examples from different functions:
- Own quarterly pipeline forecasting for the SMB segment using Salesforce and SQL, improving forecast accuracy by 10% within six months.
- Design and ship accessible web components in React and TypeScript, working with design to meet WCAG 2.2 AA.
- Plan seasonal paid search campaigns for the brand’s cordless robotic pool cleaner, using Google Ads and GA4 to reach a 4:1 return on ad spend between April and August.
- Lead cross-functional incident reviews and turn the findings into two process changes per quarter.
Each line tells a candidate what they would do and what they need to be good at to do it. It also writes the interview questions. For the forecasting role, the obvious one is to ask for a forecast the candidate built in SQL and how close it came to the actual numbers.
Spell Out the AI Access
Tech candidates now read a posting for what the company will give them to work with, and AI access sits near the top of that list. Microsoft and LinkedIn’s 2024 Work Trend Index found that 75% of knowledge workers use AI at work, and 78% of those users bring their own tools. A candidate who has been paying for a personal coding assistant wants to know whether that changes on day one.
In engineering, the question has moved to how much compute comes with the job. At GTC in March 2026, Nvidia CEO Jensen Huang proposed giving engineers token budgets worth roughly half their base salary. Microsoft EVP Charles Lamanna said a candidate had asked for a token allocation in their offer, and he suggested thinking in terms of $100 to several hundred dollars of usage per day. Engineers at Meta and OpenAI now compete on internal leaderboards that track token consumption.
What to state in the posting
Most descriptions mention AI only as a required skill. Candidates also want to know what they’ll have to work with. A short “Tools and AI access” block can cover:
- Approved tools: which assistants and coding agents the team uses, such as Claude Code, GitHub Copilot, or ChatGPT Enterprise
- Usage limits: whether each person gets a monthly token or spend budget, and who approves increases
- Agent permissions: which actions AI can take on the person’s behalf, such as opening pull requests, running tests, reading production data, or drafting customer emails that a person sends
- Data rules: what can go into a model and what can’t, which matters most in regulated roles like the lending example above
- Training: whether the company provides it or expects candidates to arrive already skilled
Match the skill to the access
Asking for “experience building with AI agents” when the team has no approved agent tooling creates the same mismatch as a vague responsibility list, and the new hire finds out in month two. A merged statement fixes it:
“Automate regression testing for the payments service using Claude Code and the team’s shared agent budget, cutting manual QA time by half within two quarters.”
That line names the outcome, the tool, the budget, and a measure. It also gives the interview panel a question to ask: have the candidate walk through an agent workflow they built and what it cost to run.
Writing the Description, Start to Finish
The order matters here. Outcomes come first, skills second, and wording last.
- Map the outcomes. Talk to the hiring manager, someone performing well in a similar role, and one person from a team the role works with closely. Agree on what the hire must deliver in the next 6 to 12 months. Those outcomes become the core responsibilities.
- Work backward to the skills. For each outcome, ask which skills produced similar results before. List standard tasks, skills, and work activities by occupation, and look at which skills employers in the market are asking for right now.
- Draft merged statements. Write at least half of the responsibilities in the outcome-plus-skill format, in everyday words.
- Review with the interview panel. Each interviewer should be able to write a structured question for every essential skill and core responsibility. A line nobody can write a question for is too vague. Cut any skill a new hire could pick up in their first two weeks.
- Revisit every 6 to 12 months. Track how many applicants move through each stage and read candidate feedback. When strong candidates drop out right after reading the posting, check the description before anything else.
Checking the Draft Before It Goes Live
A hiring manager reading their own draft rarely spots the words that put people off, because those words sound normal to them. A text check catches what a reread misses. Ongig’s Text Analyzer scans a job description for biased language around gender, age, and disability, as well as jargon and reading level, and suggests replacements before the posting goes out.
