Advanced AI Recruiting Workshop: Streamline Sourcing and Outreach with Your Own AI Assistant

Generative AI writing tends to sound flat and generic, and the fix isn't a better prompt here or there but a shift in how recruiters relate to the tool itself. The content walks through why AI output reads the way it does, how tokens work, and why treating a chatbot like software instead of like a new team member produces weak results.

The core exercise centers on brand voice cloning: taking a boring, generic job description and running it through a custom tool trained on a well-known company's talent brand video. The same job description comes out sounding distinctly like Apple in one pass and distinctly like Boeing in another, proving that tone, vocabulary, and sentence structure can be reverse-engineered from a transcript and applied to anything written afterward.

Steps covered include:

• Feeding a transcript, email set, or brand video script into a writing-style analysis prompt that outputs a style guide covering tone, vocabulary, sentence structure, and grammar patterns
• Pasting that style guide into a ChatGPT Project (or a Gemini Gem, or a Claude Project) under custom instructions so every future message inherits the brand voice
• Naming the AI assistant and giving it context and background before asking it to do work, the same way a new hire would be briefed rather than issued a one-line command
• Using a free tokenizer tool to see how prompts get broken down, useful for explaining AI mechanics to skeptical colleagues or leadership
• Cloning a personal writing style from a LinkedIn profile so outbound messages sound like the recruiter wrote them, not a chatbot

There's also a pointed warning about tool choice: paid versions of ChatGPT, Gemini, or Claude are treated as the minimum standard, while Microsoft's Copilot is flagged as inadequate because it restricts access to custom instructions. The broader argument ties back to layoffs already happening at large companies tied to AI adoption, framed as a signal that recruiting teams who aren't building custom instructions and prompting fluently are on borrowed time.

Everything can be reverse engineered by a prompt.

The takeaways apply beyond sourcing emails, extending to intake notes, resume matching, and any recurring piece of writing a talent team produces at scale.