ChatGPT agent mode: The “little helper” that transformed recruiting, crafted user personas, and solved parking nightmares | Michal Peled (Honeybook)

What It Covers

Michal Peled from HoneyBook demonstrates three ChatGPT workflows: using agent mode to automate LinkedIn recruiting searches, converting customer research into interactive AI personas, and creating customized calendars to avoid parking price surges near Oracle Park.

Key Questions Answered

• •Agent Mode Recruiting: ChatGPT agent mode logs into LinkedIn, searches profiles using specific criteria (Israel-based, active within three months, one year minimum tenure), and delivers five qualified candidates in ten minutes—four were new discoveries the hiring team hadn't found manually, one was already interviewing.
• •Persona Creation Pipeline: Upload customer research files to NotebookLM, prompt it to generate detailed persona instructions with citations, refine output to under 8,000 characters, then deploy as custom GPTs. This transforms hundreds of pages of unused research into conversational personas teams actually consult dozens of times.
• •Prompt Engineering Framework: Start with role definition (you are an IT recruiter), add task description, include specific restrictions as bullet points, and always add guardrails like "don't add or modify text not written or implied in sources" to reduce hallucinations and maintain accuracy in outputs.
• •Prompt Debugging Technique: When outputs fail, feed the broken prompt back to ChatGPT with three components: what's wrong with current output, numbered list of desired improvements, and explicit permission to delete, rewrite, or add anything. This typically fixes issues in one iteration without manual rewriting.