This article analyzes how AI assistants' long-term memory can misapply past preferences to new tasks, distinguishing stable preferences, task constraints, and factual judgments, and argues for memory systems that explain their sources and applicability, referencing NIST, OWASP, and China's data protection law.
The article argues that increasingly precise risk scores fail to provide the contextual evidence business users need to act confidently, advocating for explainable interfaces that surface reasoning, uncertainties, and handoff clarity alongside scores, referencing NIST's trustworthy AI principles.
The article examines how AI agents are shifting from answering questions to proposing workflow actions, blurring the line between suggestion and execution, and argues that organizations must preserve clear decision boundaries, provide adequate confirmation materials, and maintain traceable handoffs to ensure human accountability and sustainable collaboration.
This article explores why updated knowledge bases often still surface outdated answers, explaining how older content's completeness and familiarity outweigh newer, conditional rules, and argues for embedding version context, applicability conditions, and source traceability into AI-generated answers to maintain trustworthiness.
The article argues that increasing system notifications and AI-generated summaries create an illusion of awareness but lack clear responsibility assignment, impact context, deadlines, and consequences, causing critical decisions to stall; it proposes transforming notifications into accountable action items with defined owners, time bounds, and escalation paths, separating broadcasts, todos, and escalations into distinct channels.
AI-generated meeting minutes create an illusion of completeness that obscures unresolved decisions, pending conditions, and disagreements, turning records into polished artifacts that hinder real progress by masking what still needs clarification before action can begin.
This article explains why the same AI model succeeds in one job role but fails in another, arguing that usability depends on role-specific risk, responsibility, and judgment interfaces rather than model accuracy, and proposes a framework of three role-based questions and three handover checkpoints for effective AI integration.
Integrating large language models into business systems often only accelerates existing steps without transforming workflows because organizations fail to define judgment, evidence, and responsibility boundaries, leaving AI outputs as unactionable suggestions that require manual re-review and coordination.
This article explores how highly automated workflows isolate exceptions, stripping context needed for human judgment, and argues for designing exception handoffs as structured re-judgment tasks with complete context packages, proper human placement at decision forks, and metrics focused on recovery quality rather than failure rates.
This article explains why AI systems produce different answers over time despite no apparent model updates, identifying five key variables—model configuration, knowledge retrieval, external tools, permissions, and human operations—and argues for lightweight 'explanation cards' to make answer changes traceable and governable.
This article analyzes why traditional one-time authorization fails when AI agents dynamically select and chain tools, proposing a context-aware framework of connection, delegation, and confirmation grounded in MCP specifications and NIST AI risk management to ensure accountable, scoped, and auditable agent actions.
The article argues that as AI systems integrate into real workflows, evaluation must shift from checking answer correctness to assessing process reliability—handling incomplete inputs, evidence conflicts, tool-use boundaries, and post-error traceability—citing Chinese regulations, NIST, and OWASP frameworks, and proposes four key questions for scenario-based evaluation.
The article argues that integrating AI agents into business systems shifts the core challenge from capability to accountability, proposing a four-layer identity framework (initiator, decision-maker, executor, affected) and a four-question checklist to govern agent actions, referencing OWASP and NIST frameworks.
The article argues that long-term memory in AI agents introduces governance challenges beyond storage, requiring three boundaries—retention, invocation, and exit—to prevent outdated or temporary information from incorrectly influencing actions, illustrated by a workflow change scenario and aligned with emerging AI regulations.
This article argues that as AI applications rapidly integrate models, the real risks shift to the overlooked middle layer—gateways, plugins, vector databases, and orchestration components—that control data flow, tool access, and audit trails, and proposes a three-chain framework (capability, data, responsibility) for governance.
This article argues that unreliable enterprise AI Q&A systems stem from poor knowledge governance — not model limitations — and outlines five essential questions a trustworthy knowledge base must answer: source traceability, version validity, applicability scope, ownership, and error correction loops.
