Klara shares how she got started with AI, why communities helped her build confidence, and how she now works with an AI-first mindset as a solo founder. The conversation zooms in on one practical starting point for L&D teams: building custom GPTs that solve narrow, real problems, from onboarding support to manager coaching for difficult conversations. You will also hear how testing, iteration, and a simple product mindset can make these tools more useful (and less overwhelming).

What You’ll Learn

1. How to create a custom GPT using the conversational “Create” flow versus the more structured “Configure” flow
2. How to design a manager coaching GPT that role-plays a difficult conversation and gives constructive feedback
3. How to improve GPT quality through tighter instructions, model choices, knowledge bases, and iteration
4. How to use low-stakes personal projects to build confidence before deploying GPTs for workplace use
5. How a product mindset can help L&D teams build and test AI tools that people actually use

Tools Referenced

• ChatGPT
• Perplexity AI
• Google Workspace
• Gemini
• Lovable

Key Takeaways

1. Start with a “try things out” mindset, not perfect questions. Clara’s early curiosity turned into a habit of experimenting with new tools and learning what AI is actually good for (through use, not theory).
2. Community creates permission and pace. Being in AI-focused groups gave Clara exposure to what is possible, early access to tools, and peer support that made it easier to keep learning.
3. Use an AI-first approach when you need to learn fast. As a solo founder, Clara leaned on AI to speed up learning business basics (tax, sales, website creation) and to reduce reliance on corporate infrastructure.
4. Custom GPTs work best when the problem is narrow. The more specific the use case, the easier it is to define the role, context, and output, and the more useful the GPT becomes.
5. Build for yourself first, then build for others. Clara described personal “low-stakes” builds (like a movie recommender) as an easy entry point before creating GPTs meant for managers or teams.
6. Treat GPTs like products, test with users, and iterate. Clara emphasized trying the GPT, noticing what is not working (too much text, wrong model choice, unclear output), and refining instructions or restarting when needed.
7. Overwhelm is real; focus on what you can control. You cannot control how fast AI changes, but you can control your learning habits, practice time, and the communities you learn with.

To explore

• Klara created this CustomGPT Builder to help you better create CustomGPTs
• Use this GPT by describing what you are hoping to build, give it feedback, have a conversation, then copy-paste the outcome (a Markdown file) into a new GPT's system instructions

Follow Klara:

• LinkedIn: https://www.linkedin.com/in/klarahermesz/
• AI Enablement Academy: https://aienablement.academy/