"I Have an Idea and I Just Go Do It." She Built 20 AI Agents to Automate Her PhD.

About this episode

What happens when a leadership development consultant who's never written code decides to automate her entire PhD? Nathalie Salles spent her career in human development at Facebook and Google, coaching leaders and scaling talent programs. Now she's doing a PhD on AI, workplace, and gender, and somewhere along the way, she became a builder. She taught herself n8n, built a pipeline to process over 1,500 research papers, and started creating AI agents for her coaching clients. In this conversation, she shares what human-AI collaboration actually looks like when you're learning it from scratch, why the software engineers were impressed, and how she's using AI to give power back to the people who need it most.

Key takeaways

1. You don't need a technical background to build powerful AI systems. Nathalie chose the hardest automation tool (n8n) on purpose and built a research pipeline that outperformed what software engineers had built.
2. You have to lead the AI collaboration. AI won't structure the interaction for you. Be explicit, set boundaries, and tell it what you need.
3. Academic research on AI is accelerating faster than humans can read it. 24,000 papers were published in two years, more than the previous seven decades combined.
4. AI agents can extend your expertise to areas you don't know. Nathalie built HR specialist agents for legal, compensation, performance management, even areas she had no expertise in.
5. AI can be a mirror for personal development. For coaching clients, AI assistants accelerate learning and extend the relationship between sessions.
6. We naturally trust AI too much. Understanding that tendency is the first step to collaborating with it more effectively.
7. Start with an assistant, not with agents. Don't jump to agentic workflows. Build a focused AI assistant first, learn its quirks, then build up.

Chapters

• 00:00Introduction to Nathalie Salles
• 00:33Three Careers in One: Human Development to AI Research
• 01:33When AI First Caught Her Attention at Facebook
• 03:09Finding Her Tribe in Responsible AI
• 03:42Automating a PhD: 1,500 Papers and Counting
• 05:23Cognitive Load and the AI Highway
• 05:36Building the Literature Review Pipeline with n8n
• 07:34A Non-Techie All the Way Through
• 08:15When Software Engineers Said Her Results Were Better
• 09:05What Good AI Collaboration Looks Like
• 10:43Leading the Collaboration: Setting Boundaries with AI
• 12:25Building AI Agents for the HR Space
• 13:00Creating a Team of 15-20 AI HR Specialists
• 15:06Coaching Leaders with AI Mirrors
• 17:33Giving Power to People Through AI
• 18:10Advice for Getting Started with AI
• 19:59Identity in the Age of AI
• 21:27What She Would Automate: Finding Her Tribe
• 22:52Outro: Diving In Instead of Running Away

Notable quotes

• "I've never felt as free as in the past two years. I have an idea and I go do it."
• "When I don't understand something, I kind of dive in instead of running away from it."
• "I have to lead the collaboration much more than I thought I needed to."
• "When I presented it to software engineers, they were like, wow, your results are better than ours."
• "I got the AI bug, I'm never going back."
• "How do I give power to people who might not have had a chance to use those tools?"
• "I would want to actually build the ability to find my tribe."