Unmasking AI in Recruiting: Demystifying the Spooky Side

A conversation about artificial intelligence in recruiting cuts through the hype to focus on what large language models can actually do reliably today, and where teams get themselves into trouble by moving too fast. The comparison drawn throughout is to early, unrefined technology: powerful but poorly understood, with rules and norms still being written in real time.

Key points covered include:

• Why matching candidates or running AI-led interviews looks simple in a demo but carries legal, ethical, and practical risk that scheduling or workforce planning does not.
• A recommended sequence for adoption: start with job descriptions, outreach emails, and interview prep guides, where mistakes are low-stakes, before moving toward higher-risk uses like automated screening.
• Why treating generative AI as a plain autocomplete for emails wastes its potential, and how prompting it toward a specific voice, format, or creative constraint produces outreach that stands out rather than reading like a form letter.
• Candidate attitudes toward chatbots and automated messaging, and why the assumption that applicants demand a human on the other end doesn't hold up against the reality of unanswered applications and long silences in the hiring process.
• How bias enters AI output through training data pulled from the open internet, and why careful prompting can reduce gendered or otherwise skewed language but cannot eliminate the underlying problem on its own.
• What a bias audit looks like in practice: comparing diversity metrics across the top and bottom of the hiring funnel against the makeup of the original talent pool, so that any drop-off has a measurable baseline rather than a guess.
• The argument that creativity itself, human or machine-generated, is mostly recombination of familiar patterns, and why that reframes what "original" work from an AI system really means.

The overall guidance favors caution paired with experimentation: test tools on low-risk tasks, build internal policy as comfort grows, and put measurement in place before scaling any AI-driven process into decisions that affect candidates directly.