Tina Huang: 72 hours of AI Agent Workshops in 18 Minutes

After attending over 72 hours of talks and workshops on AI agents during conference season, I've compiled the essential insights on opportunities, skills, career advice, and practical tips for AI agents. Let's dive in.

Big companies like Google are integrating AI agents into their products, giving them autonomous capabilities. By 2025, agentic AI products will become mainstream, opening vast opportunities for startups and existing businesses alike.

For startups, the sweet spot lies in vertical AI agents—specialized agents tailored for specific industries or functions, similar to how vertical SaaS companies thrived by focusing on niche markets with deep domain knowledge.

Overall, vertical AI agents represent a transformative opportunity, potentially surpassing traditional SaaS in value by automating not just software but also human workflows, making this an exciting time to build, learn, and innovate in the AI agent space.

To build a successful vertical AI agent, first choose a niche you understand well, then identify repetitive, administrative tasks to automate. AI agents excel at 24/7 availability, cost efficiency, and personalization, which can replace entire teams in some cases.

Two critical skills repeatedly emphasized are prompt engineering—crafting precise, clear instructions for AI—and writing evals, which are structured tests to measure an agent's performance across tasks and edge cases, ensuring reliability in production.

If you're interested in deepening your AI agent skills, consider joining hands-on boot camps and workshops that cover building multiple AI agent systems and deploying them in production, helping you stay ahead in this rapidly evolving field.

Agent engineering is emerging as a key career path, combining prompting, traditional software engineering, product domain knowledge, and machine learning. This role bridges the gap between impressive demos and reliable, scalable AI agent products.

A practical framework for automating workflows with AI agents includes: 1) Observe current processes, 2) Decompose into tasks, 3) Map task flows, 4) Prototype agentic workflows, and 5) Evaluate and iterate using evals to refine performance.

Looking ahead, focus on mastering prompt engineering and evals, exploring voice-enabled AI agents to enhance user experience, and learning coding skills. While coding is important, domain experts can leverage no-code tools and collaborate with engineers to build AI solutions.