Job descriptions and structured interview kits
Turn a role's real requirements into a clean job description and a matching structured interview kit (questions plus a scoring rubric) that your team owns, reuses, and runs the same way for every candidate.
Tools you'll use
A structured interview kit is a fixed set of job-related questions, asked of every candidate in the same order and scored against a written rubric, paired with a job description that states the role's real requirements in plain language. Two documents sit at the front of every hire: the job description that pulls people in, and the interview kit your panel uses to decide.
Most hiring teams draft from scratch, recycle outdated postings, or run unstructured interviews where each panelist asks different questions. That creates delays, job language that quietly narrows the applicant pool, and decisions based on gut feel rather than consistent evaluation.
Structure is what makes interviews actually work. A 2022 meta-analysis by Sackett and colleagues, summarized by the Society for Industrial and Organizational Psychology, found structured interviews to be the single strongest predictor of job performance among common hiring methods, with a mean operational validity of r = .42 — ahead of cognitive ability tests at r = .31. Job descriptions written in plain language, with clear "must have" versus "nice to have" lines, widen the pool and speed up screening. AI tools can now draft strong first versions of both documents in minutes, tied directly to the actual role, so your team owns, reuses, and runs the same process for every candidate.
Moriva's take
Gate 1, real work: yes, this is a weekly task for any team that is hiring, and it sits right at the top of the funnel. Gate 2, owned: strongly yes. The output is just documents and templates your team holds, edits, and reuses with no platform lock-in. Gate 3, measured: easy to point to. Track drafting hours per req and whether panels actually use the rubric. This is a clear "start here" for People teams. Keep a human in the loop on final wording and on any compliance-sensitive language, and you have a safe, high-value first project.
How do you job descriptions and structured interview kits?
- 1
Start from the real role, not a template
Open Claude Cowork and give it the raw inputs you already have: the hiring manager's notes, an old posting for the role, the team's goals for the next year, and three or four tasks the person will actually do in week one. Ask it to draft a short competency list first, the four to six things this person must be able to do, before any prose. Grounding the work in observable job behaviors is what makes everything downstream defensible and accurate.
- 2
Draft the job description with a must-have / nice-to-have split
Ask the tool to turn the competencies into a job description with separate 'Required' and 'Preferred' sections, plain 'you will' language, and a short note on the work setup and accommodations. Tell it to flag and remove gendered or coded terms ('rockstar', 'ninja', 'digital native', 'culture fit') and to define what proficient looks like for each skill. This clarity widens the pool and cuts screening time, and it is a deliberate review pass rather than a vibe check.
- 3
Build the matching structured interview kit
Have the tool generate one behavioral or situational question per competency, anchored to past behavior, so the same questions run for every candidate in the same order. Ask for a behaviorally anchored 1-to-5 rubric for each question, where each score level describes what a weak, solid, and strong answer actually sounds like. This is the part most teams skip, and it is the part that drives consistency and reduces bias.
- 4
Split the kit across the panel and add follow-ups
Ask the tool to assign each competency to a specific interviewer so two people are not testing the same thing, and to add two or three probing follow-ups per question. Have it produce a one-page scorecard each interviewer fills in independently before the debrief. This keeps the panel covering the whole role and keeps scores honest before people influence each other.
- 5
Turn the winning format into a reusable template you own
Once you like the shape, point Claude Code at a folder of your past job descriptions and kits and ask it to extract your house style and structure into a template plus a short checklist. Now any new role starts from your standard, not a blank page. The template is plain files your team holds and edits; there is no vendor to renew and no lock-in.
- 6
Automate the repetitive version for high-volume roles
If you hire the same roles often, ask Claude Code to build a small script that reads a simple spreadsheet of role inputs (title, team, key tasks, level) and outputs a draft description and interview kit per row as formatted documents. Your team runs it, reads the drafts, and edits, no engineer on retainer. This is where a focused week of setup pays back every time a new req opens.
- 7
Run a quick fairness and accuracy review before publishing
Before anything goes live, have the tool re-read the final description and kit against your competency list and a plain checklist: every required item maps to a real job task, no inflated 'years of experience' bars, no coded language, salary range present where required. A human signs off on wording and any legal-sensitive phrasing. The tool drafts and checks; a person decides.
What could go wrong (and how to handle it)
Inflated or irrelevant requirements creep into the description and quietly screen out good candidates.
Force every 'Required' item to trace back to a real week-one task in the competency list. Have the tool justify each requirement and cut anything it cannot tie to the work.
Coded or biased language (gendered terms, 'digital native', 'culture fit') slips through and narrows the applicant pool.
Make a bias-language pass an explicit step, not an afterthought, and keep a human reviewer who knows the team. Maintain a banned-words list in your template that the tool checks against every time.
Over-automation: a script generates dozens of postings that nobody reads closely before they go live.
Treat every generated draft as a draft. Require a named person to read and edit each one. The automation removes the blank page, not the judgment.
The interview rubric becomes a checkbox exercise and interviewers ignore it.
Assign competencies per interviewer, require independent scorecards before the debrief, and review whether scores were actually filled in. Train the panel once on how to use anchored ratings.
Compliance exposure where job-posting and hiring rules apply (pay transparency, accommodation language, and, in some jurisdictions, rules on automated decision tools).
Use the AI to draft documents that humans read and choose between, not to score or auto-reject candidates. Keep counsel or an informed HR lead on final language, and know your local rules before publishing.
Accuracy drift: the tool invents a responsibility or a benefit the role does not actually have.
Ground every draft in the inputs you provided and have the tool cite which input each claim came from. The final review step exists to catch invented specifics before they reach a candidate.
Prompts to get started
FAQ
Isn't this just letting AI write our job postings? How is that different from copying a generic template?
The difference is grounding. You start from this role's real tasks and a competency list, and every line in the description has to trace back to that. A generic template describes a generic job; this produces a description and a matching interview kit for the actual role, then a human edits it. The tool removes the blank page, not the judgment.
Does using AI here create legal risk under hiring laws like the EEOC guidance or NYC's bias-audit rule?
Those rules are mainly aimed at automated tools that score, rank, or auto-reject candidates, which is a different and higher-risk use. Drafting documents that humans read, edit, and choose between is far lower risk. Keep a person on final wording and on any score, never let the tool make the hiring decision, and check your local pay-transparency and accommodation rules before publishing.
Do we need engineers to set this up?
No. The drafting itself runs in Claude Cowork, which is built for knowledge work and needs no coding. If you later want a script that turns a spreadsheet of roles into draft kits, Claude Code builds and runs that for you in plain English, and your team owns the result and can change it. There is no vendor seat to renew.
Why bother with a scoring rubric? Our interviewers are experienced.
Experience does not fix the core problem: unstructured interviews predict job performance much worse than structured ones, and they widen score gaps between groups. A behaviorally anchored rubric means every candidate is judged on the same evidence, and your debriefs argue about answers, not impressions. It is the single highest-value part of the kit.
How do we know it actually saved time?
Measure it directly. Track hours from 'req opened' to 'description and kit approved' before and after, and whether panels actually filled in their scorecards. Teams commonly cut description drafting from hours to minutes per role. If the rubric is being used and drafting time dropped, you have your answer.
Sources
- Structured interviews had the highest mean operational validity (r = .42) of common selection methods, ahead of cognitive ability at r = .31 (per Sackett et al., 2022). — Society for Industrial and Organizational Psychology (SIOP), 2023
More from HR & People
Onboarding docs and a policy Q&A bot
Turn your scattered onboarding and policy documents into a Q&A assistant that answers new-hire and employee questions with quoted sources, and escalates anything sensitive to a human.
Resume-screening assist and structured scorecards
An AI assistant that reads each resume against a job-specific rubric and returns a structured scorecard with evidence, so your recruiters spend their time on real candidates instead of the first pass. A human still makes every decision.
Prep renewals and spot upsell openings
Build a repeatable agent that pulls usage, support, and account data into a per-customer renewal brief flagging churn risk and concrete upsell openings, 90 days ahead of each renewal date.
Want help shipping this?
We'll build it with your team on your real work — and leave you owning it, not renting it.