In the last couple of years, “AI” went from a curiosity to something that shows up in every toolchain—and it’s getting harder to tell what’s genuinely useful versus what’s just a new label. In this session, Thomas Betts will recap the practical foundations of AI beyond the buzzwords, then fast-forward to the state of AI in 2026—with an emphasis on what matters to developers building solutions on (and around) Blackbaud products. We’ll break down what’s changed in the past year: AI agents that can act (not just chat), the rise (and risks) of vibe coding, and why spec-driven development is becoming the antidote to “prompt roulette” for anything you intend to maintain. You’ll learn where these approaches shine (rapid prototypes, internal tools, workflow automation) and where they can hurt you (security, correctness, maintainability, and hidden costs). We’ll also cover the essential do’s and don’ts: treating generative models as prediction engines, not sources of truth, designing for human oversight, and setting lightweight guardrails that help citizen developers ship safely and confidently. You’ll leave with a practical mental model, a vocabulary you can trust, and a clear checklist for evaluating “AI-powered” features—so you can cut through the marketing and use modern AI tools to save time without breaking things.