Using AI in Recruiting and Talent Sourcing
Deliberate Academy Editorial Team
Reviewed for accuracy and professional relevance
Enjoying the course?
Sign up free to track your progress and earn a verified certificate when you pass.
- Apply the intent-led job description drafting approach using AI, including a second inclusive language review pass
- Explain why AI CV screening tools trained on historical hiring data carry a specific bias risk and describe the safer alternative use of AI in screening
- Recognize the GDPR Article 22 obligations for automated decision-making in employment contexts and identify when legal review is required
- Apply AI to build structured interview frameworks and assess how structured interviewing improves hiring decision quality
Recruiting is where AI adoption in HR is most advanced — and where the risks are most consequential. A poorly written job description drives away qualified candidates before they apply. An AI screening tool trained on biased historical data reproduces that bias at scale. A well-structured AI workflow, on the other hand, can help a two-person talent team compete with a function three times its size.
Writing Job Descriptions That Actually Work
Most job descriptions are written by copying an old one and editing around the edges. They accumulate jargon, exclusionary language, and requirements that bear no relationship to what the job actually demands. AI is genuinely useful here because it can help you start from intent rather than copy.
A strong prompt for job description drafting looks like this: describe the role's outcomes (what does success look like in 90 days?), the team context, the three or four non-negotiable skills, and the seniority level. Ask the AI to produce a draft, then review it critically — not just for accuracy but for tone, inclusion, and whether the requirements match the real job.
Inclusive language review is one of the most immediately valuable applications. AI tools can flag gendered phrasing, unnecessarily exclusionary requirements ("must have X years of experience" when you mean "must be able to demonstrate X"), and jargon that makes roles seem more specialized than they are.
After drafting a job description, run a second AI pass with this prompt: "Review this job description for language that might discourage women, people from non-traditional backgrounds, or candidates from underrepresented groups from applying. Suggest specific revisions." You will almost always get at least two or three improvements worth making.
Improving candidate pool diversity by rewriting from outcomes
Context
A senior recruiter at a professional services firm noticed that a persistent software engineering role had attracted fewer than 10% applications from women over three consecutive hiring cycles, despite the team's stated diversity goal. The job description had been carried forward from 2021 with minor edits.
Action
The recruiter rebuilt the job description using an AI-assisted outcomes-first approach: describing what success looked like in the first 90 days, specifying three non-negotiable technical capabilities, and then running an inclusive language review prompt that flagged six phrases — including 'rockstar developer' and 'competitive environment' — as likely to suppress applications from underrepresented candidates.
Outcome
The revised description produced a 34% increase in applications from women in the following cycle and a broader overall applicant pool. The recruiter documented the prompt workflow and shared it with the wider talent team as a standard process for any role that had not been rewritten from scratch in the past 18 months.
AI-Assisted Candidate Sourcing
AI-powered sourcing tools — LinkedIn Recruiter's AI features, Gem, Beamery, and others — can identify passive candidates who match a brief you define. They are useful for expanding the top of your pipeline, particularly for roles where qualified candidates are unlikely to be actively job-searching.
The key skill is prompt and filter quality. Garbage in, garbage out applies directly. If your sourcing brief over-specifies credentials (top-tier university, brand-name employer, specific tools) rather than capabilities, the AI will surface a narrow, homogeneous pool. Build briefs around demonstrated outcomes and transferable skills where possible.
A recruiter uses an AI sourcing tool to identify passive candidates for a senior data analyst role. The brief specifies: 'Big Four consulting experience, top-10 university degree, Python and SQL.' The tool returns a narrow pool of 12 candidates, all with similar backgrounds. What does the lesson suggest went wrong?
Select one answer.
CV Screening and Application Scoring
AI-assisted CV screening — where an ATS scores and ranks applications — is the area of highest regulatory and ethical scrutiny in HR AI. The core problem: most models are trained on historical hiring data that reflects who was hired in the past, not who performed best or who would perform best in the future.
If your company previously hired mostly from a particular profile — graduates from certain universities, people from specific industries — an AI screening tool trained on your data will reproduce that pattern and amplify it, often invisibly.
What to do instead. Use AI to screen for stated minimum requirements (years of relevant experience, specific licenses or qualifications, right-to-work where applicable) while keeping humans in the loop for any subjective assessment. Treat AI scoring as a shortlist-assist, not a decision-maker.
In the EU, automated decision-making that significantly affects individuals — including employment screening — is subject to GDPR Article 22. Candidates may have the right to opt out of automated processing and request human review. Speak to your legal team before deploying AI screening at scale.
In the United States, the equivalent exposure runs through a different but equally consequential framework. Title VII of the Civil Rights Act of 1964 prohibits both intentional discrimination and practices that produce an unjustified disparate impact on a protected group — and disparate impact liability attaches even where the AI tool never uses a protected characteristic directly: if a screening tool produces a statistically significant adverse effect on candidates of a particular race, sex, or other protected group, the employer bears the burden of showing the tool is job-related and consistent with business necessity. The Americans with Disabilities Act (ADA) adds a further constraint specific to AI screening: résumé parsers, video-interview analysis tools, and gamified assessments can inadvertently screen out qualified candidates with disabilities — for example, by penalizing atypical speech patterns, unconventional eye contact, or employment gaps connected to a disability — and employers remain obligated to offer an accessible alternative process on request.
Employers hiring for roles based in New York City face a further, more specific obligation. NYC Local Law 144 requires employers using an "automated employment decision tool" to substantially assist or replace hiring or promotion decisions to commission an independent bias audit of the tool before use, publish a summary of the results, and notify candidates that such a tool is in use.
Specifically, the bias audit must be conducted by an independent auditor no more than one year before the tool is used — in practice, this means a fresh audit at least annually for any tool still in use. A summary of the results must be posted publicly on the employer's website before that use begins, and candidates or employees must be notified at least 10 business days beforehand that an AEDT will be used, along with how to request an alternative selection process or a reasonable accommodation. The law is enforced by the NYC Department of Consumer and Worker Protection (DCWP); as with any relatively new law, specific enforcement patterns and interpretive guidance continue to develop, so treat the audit-and-notice mechanics above as the stable core obligation and confirm current DCWP guidance for anything beyond it.
Several other US states and cities are considering comparable legislation, so treat Local Law 144 as an illustration of a fast-moving compliance area rather than the only applicable rule.
Interview Preparation and Structured Interviewing
AI is useful for building structured interview frameworks. Given a role description and competency model, AI tools can generate a bank of behavioral interview questions mapped to each competency, with example strong and weak responses for calibration.
Structured interviewing — where every candidate answers the same questions, evaluated against the same criteria — is consistently shown to produce more accurate and fairer hiring decisions than unstructured interviews. AI can do the scaffolding work quickly, freeing the interview panel to focus on actual assessment quality.
Candidate Communications
Drafting personalized rejection messages, interview invitation emails, and feedback summaries are all legitimate, low-risk AI drafting tasks in recruiting. The volume of these communications in high-throughput recruiting is significant. AI can produce a first draft in seconds; a recruiter reviews and personalizes it in under a minute. That is a real efficiency gain at scale.
Why does an AI screening tool trained on your historical hiring data carry a specific bias risk that goes beyond general AI accuracy concerns?
Select one answer.
Exercise
Your Task
Take a live or recently closed job description from your organization. Paste it into an AI tool and run the inclusive language review prompt from this lesson — asking it to flag language that might discourage women, people from non-traditional backgrounds, or underrepresented candidates from applying. Review the suggestions and identify which two or three changes would most meaningfully broaden the candidate pool. Then rewrite the requirements section of that same job description starting from the role's intended outcomes rather than a credentials checklist, and compare the two versions.
Your reflection
Did you complete this exercise? What did you find? (Saved locally in your browser)
Try It: AI-Graded Practice
The exercise above is self-assessed. The exercise below is graded automatically, so you can get direct feedback on whether your rewrite actually applies the outcomes-first approach from this lesson.
- AI is genuinely useful for job description drafting — start from the role's intended outcomes rather than copying old descriptions, then run an inclusive language review pass to catch gendered phrasing and unnecessarily exclusionary requirements.
- AI-assisted sourcing tools expand the top of your pipeline by identifying passive candidates — the key skill is building briefs around demonstrated outcomes and transferable skills rather than over-specifying credentials that narrow the pool homogeneously.
- AI CV screening carries real bias risk because models learn from historical hiring data, which reflects who was hired in the past — not who performed best — and can reproduce and amplify demographic patterns invisibly at scale.
- Use AI to screen for stated minimum requirements only and keep humans in the loop for any subjective assessment — treat AI scoring as a shortlist assist, not a decision-maker.
- In the EU, automated decision-making that significantly affects individuals in employment is subject to GDPR Article 22 — speak to your legal team before deploying AI screening at scale.
- In the US, Title VII disparate impact liability and ADA accommodation obligations both apply directly to AI screening tools, and NYC Local Law 144 requires a bias audit, public results summary, and candidate notice for automated employment decision tools used to hire into New York City roles.