Business Central Integrates Copilot and AI Agents. Business Central embeds Copilot and first-class AI agents to automate ERP workflows, using Azure OpenAI models, permissioned data, and audit-ready metadata. Outcome engineers must treat agents as first-class workflow components with traceable context and audit trails — Principle 09 & 11.

Anthropic’s Code with Claude showed off coding’s future—whether you like it or not. Anthropic’s Claude Code automates end-to-end coding, letting agents self-test, self-correct, and “dream” to learn codebases without constant human oversight. If you build delivery pipelines, design agents to own testing and verification loops so they can deliver executable artifacts — Principle 03 & 09.

A 0.12% parameter add-on gives AI agents the working memory RAG can’t. Delta-mem compresses long interaction history into a tiny, dynamic matrix, giving agents persistent working memory with 0.12% parameter overhead. Adding efficient, explicit working memory changes how you architect stateful agents and context engineering — Principle 06 & 11.

Datasette Agent. Datasette Agent brings a conversational, extensible AI assistant to Datasette, letting you query databases, generate charts, and extend behavior with plugins. That pattern—small, composable assistants with plugin sandboxes—gives you a reproducible way to expose governed data and orchestration hooks in production — Principle 03, 06, 07.

LLM Guidance Does Not Transfer Across Providers. Researchers show that LLM guidance and prompt strategies rarely transfer across providers, forcing cross-model testing and validation. Outcome engineers must bake model-portability tests and multi-provider validation into CI and evaluation pipelines rather than rely on single-vendor heuristics — Principle 11 & 16.