Every software company on Earth is selling you an “agent” right now. Most of those things aren’t agents. The word has been stretched so thin it covers everything from a chatbot with a new logo to a genuinely autonomous system that can reason, plan, and execute multi-step workflows. If you can’t tell the difference, you can’t evaluate what you’re buying, building, or betting on. This session fixes that.

We’ll build from first principles: what a large language model actually does under the hood, what “orchestration” means in practice, and where the real boundaries are between rule-based automation, workflows, copilots, true agents, and multi-agent systems. You’ll get a clear taxonomy with concrete examples — including assessments of where Blackbaud’s own products fall on the spectrum.

We’ll dig into the orchestration patterns that make agentic behavior possible (tool calling, planning loops, state management) and the “human in the loop” design spectrum from full-approval to fully autonomous.

Along the way, we’ll arm you with a practical scorecard for evaluating any product’s agent claims, so the next time you read a press release or sit through a vendor demo, you’ll know exactly what’s real. No jargon-shaming, no dunking — just clarity, precision, and a vocabulary that makes you dangerous in the best possible way.