Beyond Prompt Engineering: How AI Really Gets Work Done – An APEGS Webinar

Event Summary

As large language models (LLMs) become more capable, the challenge shifts from prompting individual tasks to building reliable, real-world systems around them. This presentation introduces harness engineering – the practice of designing control layers that transform raw models into structured, dependable AI agents. Starting with simple prompt engineering in Excel, we demonstrate the limitations of single, long-context skills and motivate a more modular approach. We then walk through a practical example: a lightweight data analytics agent built entirely in Excel using Copilot. The solution decomposes work into reusable skills—data cleaning, analysis, comparison, insight generation, and validation – coordinated through a custom skill router using slash commands. A simple memory system tracks state across steps, enabling multi-stage workflows. This session provides a hands-on, accessible introduction to agent design patterns, showing how familiar tools like Excel can be used to prototype and understand modern AI agent architectures.

Presenter

Bin Lu, P.Eng., is a Senior Analytics Engineer at SaskPower, specializing in applying artificial intelligence and machine learning to distribution asset management. He holds a background in Electrical Engineering and Computer Science, along with a Master of Science in Applied Artificial Intelligence. With 17 years of experience, Bin works at the intersection of operational technology (OT) and information technology (IT). He collaborates with engineering teams to identify AI/ML use cases and translates them into data science solutions. He partners with IT and platform teams, using tools such as Databricks, to develop, package, and deploy production-ready AI solutions across the organization.

Who Should Attend

This presentation is intended for engineers, geoscientists, and technical professionals who are interested in understanding how emerging AI technologies can be applied in engineering and operational contexts.

The session is designed for participants with a strong problem-solving and system-thinking background, but no prior experience with artificial intelligence or large language models is required.

It will be particularly valuable for individuals involved in:

• Data analysis and decision support
• Operational efficiency and process improvement
• Digital transformation and automation initiatives

The content focuses on practical, real-world applications, demonstrating how tools like Copilot in Excel can be used to build structured, reliable AI workflows. Emphasis is placed on usability, validation, and system-level design, making the material relevant to engineers seeking to apply AI in a controlled and responsible manner.