I finished the AI foundations course on Coursera a couple of weeks ago. Honestly, I went in expecting a lot of hype and surface-level content, but it was more useful than I expected, especially when I filtered everything through my day job in FP&A. The biggest takeaway was the mental model for when AI adds value vs. when it doesn't. The course framed it as a question of structure: if the task has clear inputs, clear outputs, and lots of historical examples, AI can probably help. If the task requires judgment, context, or navigating ambiguity, AI is a tool but not a replacement. Most of FP&A lives in that second bucket, which is actually reassuring. The value isn't in automating the whole job: it's in automating the prep work so you can spend more time on the judgment calls. A few specific things that stuck: **Prompt engineering matters more than model selection.** For most business use cases, the difference between a good prompt and a bad prompt is bigger than the difference between models. I've seen this firsthand: a well-structured prompt with examples and constraints gets me 80% of the way there on variance commentary, data cleaning, and even first-draft presentations. **RAG is underrated for enterprise use.** Retrieval-augmented generation, basically giving the model access to your own documents, solves the "it doesn't know our business" problem. I want to experiment with this for internal reporting. **Start with the workflow, not the technology.** The course kept coming back to this and I agree completely. Map the process first, find the bottlenecks, then see where AI fits. Don't start with "how do I use AI?" Start with "what's slow and painful?" I've already been applying some of this thinking day to day, and the [[signal/AI Tools for Analysts|AI tools for analysts]] post captures the broader landscape. But this course gave me better language for explaining *why* certain tools work in certain contexts. Worth the time. Would recommend to any finance person curious about AI but skeptical of the hype. ← [[Notebook|Back to /notebook]]