How to Use Claude for Talent Acquisition

Using Claude in Talent Acquisition

The panel presented Claude as a practical tool for reducing repetitive recruiting work—not as a replacement for recruiter judgment. Its value depends on how well teams define tasks, provide context, connect tools, and review outputs. Panelists particularly praised its text handling, visual-design capabilities, and ability to work with connected tools, while noting that its writing can be verbose or robotic.

Recruiting workflows and examples

• Hiring-manager calibration: Move co-founder Adriano Gerdman described a workflow that uses a job description and hiring-meeting transcript to generate six candidate profiles across three levels. It explains why each profile fits and identifies assumptions to test.

  • The panel said this can reduce alignment time from around 24 hours to as little as 10 minutes, depending on hiring-manager feedback.
  • The initial profiles can be generated in about two minutes using Claude and a Clay connection. The recruiter can then share the rationale with the hiring manager.
  • The goal is not only to find obvious matches, but also to challenge the brief and broaden the talent pool.

• Candidate sourcing: Alla Pavlova demonstrated using Claude with the Exa connector to search for profiles against a job description. Results can include location, current role, match rationale, discrepancies, source links, a shortlist, and draft outreach.

  • The panel said Exa’s free tier may be sufficient for roughly 50–100 profiles a day, depending on usage.
  • Keyword-based searches can surface people who meet formal criteria but may be difficult to engage. Recruiter expertise is still needed to find less-obvious candidates, such as colleagues whose profiles do not explicitly list relevant skills.

• Content and design: Pavlova described using Claude Design to create websites, one-pagers, animations, and resumes. She also showed a scheduled task that uses a saved writing-style skill and Exa to generate three LinkedIn post drafts daily. She reviews the drafts rather than automating publication.

• Automation and reporting: Screen-recording a routine task can help Claude turn repeated clicks or data handling into a reusable workflow, such as updating spreadsheet text fields. Panelists recommended keeping work in a designated folder or project rather than granting unnecessary access to other files.

Playbooks and frameworks

Reusable skills

• A skill captures a repeatable task so it can be invoked without rebuilding the instructions each time.
• Recommended rule: one skill, one task. A skill may contain several sequential steps, and skills can be combined into larger workflows.
• Teams can adapt skills from the library or create their own. Adriano’s recruiting marketplace contained 24 skills and was described as being updated weekly.
• Examples include extracting role requirements from an intake transcript, applying a writing style, and cleaning spreadsheet data.

AI adoption maturity model

Adriano described four stages:

1. Situational: Individual use, often driven by a single superuser.
2. Experimental: Tools are being tested and some knowledge is shared.
3. Operational: Shared workflows exist, but data may still be transferred manually.
4. Orchestrated: Connected workflows operate with human supervision.

A survey of 162 mostly in-house TA practitioners found that 89% were at the first or second stage. The panel’s takeaway was that many organizations have access to AI tools but lack shared best practices, consistent knowledge-sharing, and plans for measuring impact.

Team rollout

Start with individual experiments, share effective approaches, then standardize and add technical safeguards before scaling workflows across the function. Shared projects and documentation can also help new hires ask questions about company guidance conversationally.

Agents

The panel distinguished a fixed, reusable skill from an agent that carries out a sequence of tasks with some decision-making. They advised defining the problem, required judgment, and necessary system access before automating a process. Human review remains important, especially for candidate outreach and publishing content.

Metrics and business implications

• Poll of viewers: About 40% wanted Claude to support candidate search, and 30% selected reporting and analytics. Assessment received about 13%, administration 8%, while planning and coordination drew less interest.
• Faster intake and calibration can free recruiters to spend more time on search quality and stakeholder alignment.
• For recruitment agencies, automating manual work may allow recruiters to manage more vacancies and increase revenue opportunities. For in-house teams, increasing requisition volume per recruiter may not be the right objective; teams may instead prioritize better work on each vacancy and candidate.
• The panel noted that recruiter workflows have changed faster than the candidate experience, which often still consists of applications, recruiter calls, and status emails. Improving the candidate experience was identified as an opportunity.

Risks and recommendations

• Protect candidate and company data. The panel warned about privacy and GDPR implications in sourcing and recommended disclosing the source of candidate information when contacting people.
• Limit access. Use a dedicated folder or project containing only the documents the workflow needs.
• Check third-party skills and connectors. Use trusted sources; the panel warned that unofficial marketplace items could behave undesirably or access data inappropriately.
• Understand provider policies. The discussion raised concerns about storing sensitive intellectual property in cloud AI services, including data handling, safeguards, and possible changes to model routing. The panel encouraged teams to learn about local, self-hosted, or air-gapped AI for sensitive use cases.
• Measure outcomes. Decide what impact AI is expected to have and how it will be assessed, rather than adopting tools without success criteria.
• Keep human oversight. Review candidate matches, outreach, and generated content before taking action.

Presenters and sources

• Hung Lee — host of Brain Food Live on Air
• Alla Pavlova — recruiter and technology trainer
• Adriano Gerdman — co-founder of Move, a recruitment consultancy

Sources and tools discussed included Adriano’s LinkedIn survey of 162 TA practitioners; Claude and Claude Design; Clay; Exa; the Claude skills library; and Greenhouse, which was mentioned as a sponsor and as a source of a software-vendor evaluation guide.