Compound AI and multi-agent systems: choose the architecture the work needs
Distinguish compound systems, controlled workflows and multi-agent coordination. A worked timing example and practical acceptance tests for enterprise architecture.
The writing
AI Articles
Arguments, architectures, and practical questions about AI in organizations. Explore the writing by topic, or search for a specific question.
AI articles
83 articles
Version drift: when a well-sourced AI answer uses the wrong policy
Keep policy versions, effective dates, permissions and approvals aligned from retrieval to action. A worked example and test plan for enterprise AI version drift.
Decision-centric AI: improve the path from evidence to action
Design AI around an owned decision and a confirmed outcome. Use a practical decision canvas, workflow example and measured expansion gates.
MCP for enterprise AI: a practical guide to integration, authority, and operations
Understand MCP's current architecture, version compatibility, discovery and authorization, then design a controlled enterprise integration with clear action boundaries.
Neuro-Symbolic AI for Multimodal Reasoning: Foundations, Advances, and Emerging Applications
Neuro-symbolic AI connects learned representations with explicit structures such as rules, graphs, or planners, but the term covers multiple designs. This guide compares representative approaches and applications while examining what…
Red-teaming AI systems: a practical playbook for testing real actions
Scope an AI red-team exercise, test prompts and tool permissions, inspect destination outcomes, protect evidence, and verify fixes without mistaking a refusal for a security boundary.
Persistent AI memory: what to keep, trust, and forget
Separate context, retrieval, and model adaptation, then test persistent AI memory for provenance, corrections, permissions, deletion, and useful continuity.
AI coding assistants: an enterprise adoption guide built around verified changes
Evaluate AI coding assistants through bounded tasks, repository permissions, independent tests, review effort, and delivery outcomes rather than generated-code volume.
LLM observability: tracing failures from request to outcome
A practical guide to AI telemetry, evaluation, privacy and incident response, with a worked investigation and checks for selecting an observability platform.
Small language models: how to choose a model that fits the job
Evaluate a small language model against the task, deployment constraints and cost of an accepted result, with current examples and a practical selection process.
Liquid Neural Networks & Edge‑Optimized Foundation Models: Sustainable On-Device AI for the Future
Liquid Neural Networks (LNNs) are transforming the landscape of edge AI, offering lightweight, adaptive alternatives to traditional deep learning models. Inspired by biological neural dynamics, LNNs operate with continuous-time updates,…
Living intelligence: what AI, biotechnology, and sensors can build together
Examine the evidence behind living intelligence, and evaluate where sensing, AI, and biology belong in the same system.