{ "schemaVersion": "1.0", "language": "en", "id": "ai-systems", "name": "AI Systems", "description": "Models, retrieval, agents, and input-output pipelines.", "alternateNames": [ "AIシステム", "AI agents", "Business automation", "Workflow Automation", "RAG", "AI Agents", "Software Development", "Machine learning", "Machine Learning", "Agent", "Multimodal AI", "Document Understanding", "Tables/Graphs", "OCR", "Model Routing", "Small Model", "System Design", "AI Agent", "Business Automation", "Tool Utilization", "Software development" ], "canonicalUrl": "https://catalysphere.co.jp/en/topics/ai-systems/", "machineReadableUrl": "https://catalysphere.co.jp/en/topics/ai-systems/index.json", "updatedAt": "2026-10-01", "articles": [ { "slug": "ai-agent-audit-log-minimization", "title": "More audit logs do not make agent actions reconstructable", "summary": "Design evidence for AI-agent actions around a chain of approval, execution, and state change—not full-content retention. Capture disputed content selectively under protection, then test reconstructability and minimization as separate release gates.", "quickAnswer": { "what": "Decide first whether the target is approval, state change, or disputed transmitted content; each reconstruction question has different evidence and authoritative records.", "why": "Join structured events by default, and capture exact content only for defined disputes under explicit scope, access, protection, and expiry controls.", "action": "Production acceptance requires a separate evidence-sufficiency gate and a minimization gate covering purpose, access, retention, exceptions, and tested disposition." }, "updatedAt": "2026-10-01", "url": "https://catalysphere.co.jp/en/articles/ai-agent-audit-log-minimization/", "machineReadableUrl": "https://catalysphere.co.jp/en/articles/ai-agent-audit-log-minimization/index.json" }, { "slug": "ai-agent-delegated-authorization", "title": "Reusing a Human Login Cannot Govern an Agent's Authority", "summary": "Instead of reusing a human session, separate the principal from the agent and evaluate a delegation that binds resource, action, audience and lifetime at execution time.", "quickAnswer": { "what": "A raw human session or indistinguishable broad bearer credential cannot support separate attribution, attenuation and agent-only revocation.[1][4]", "why": "Separate authorization from authentication by evaluating principal, workload, resource, action, audience, conditions and expiry at the execution boundary.[2][3][5]", "action": "Before production, test denial, expiry, changed requests, replay and revocation; narrow automation when the agent cannot be suspended independently." }, "updatedAt": "2026-10-01", "url": "https://catalysphere.co.jp/en/articles/ai-agent-delegated-authorization/", "machineReadableUrl": "https://catalysphere.co.jp/en/articles/ai-agent-delegated-authorization/index.json" }, { "slug": "ai-automation-capacity-conversion", "title": "Interviews Suggested About 90% Less Time. More Capacity?", "summary": "After AI-enabled RPA was introduced, interviews suggested monthly research time fell about 90%. Later, only 10–20% of estimated freed time was explained as new work or extra output. This note shows how to measure the time actually freed and where it went.", "quickAnswer": { "what": "In one anonymized workflow, interviews suggested about a 90% time reduction; these figures alone do not justify scaling or recognizing a financial benefit.", "why": "The percentages have different denominators; logs, case volume, quality, and review, correction, and maintenance time were not confirmed, so they do not establish profit or causality.", "action": "Approve only a bounded measurement phase: first check workflow time, quality, and hidden work, then trace where the net freed time went." }, "updatedAt": "2026-10-01", "url": "https://catalysphere.co.jp/en/articles/ai-automation-capacity-conversion/", "machineReadableUrl": "https://catalysphere.co.jp/en/articles/ai-automation-capacity-conversion/index.json" }, { "slug": "rpa-redesign-operational-handoff", "title": "Rethinking RPA: Execution Choices and Operational Handoffs in the AI Era", "summary": "After work has been redesigned, decide how remaining steps combine core systems, APIs, RPA, AI, and people—and what proves the operation can run without its builder.", "quickAnswer": { "what": "RPA is not an obsolete substitute for AI; it is one option for interface work with explicit rules and verifiable completion.", "why": "Core systems, APIs, RPA, AI, and people are not five exclusive choices; they combine different roles in record, connection, action, judgment, and approval.", "action": "Create an execution and handoff acceptance record for each step; test continuity without the builder, then return unmet work to joint operation, narrowed scope, or delayed handoff." }, "updatedAt": "2026-10-01", "url": "https://catalysphere.co.jp/en/articles/rpa-redesign-operational-handoff/", "machineReadableUrl": "https://catalysphere.co.jp/en/articles/rpa-redesign-operational-handoff/index.json" }, { "slug": "ai-workflow-redesign", "title": "How Far Should Workflows Be Redesigned After AI Adoption?", "summary": "Treat faster local tasks as only one signal; redesign bottlenecks, exceptions, expert handoffs, and the allocation of released capacity across the full workflow.", "quickAnswer": { "what": "After AI adoption, redesign roles by generation, evaluation, exception judgment, and approval rather than replacing an entire job.", "why": "A faster individual step does not increase system outcomes when downstream capacity, review load, demand, and exception handling remain unchanged.", "action": "Measure end-to-end lead time, confirmed quality, rework, exception age, and expert load, then repair the step where congestion moved." }, "updatedAt": "2026-10-01", "url": "https://catalysphere.co.jp/en/articles/ai-workflow-redesign/", "machineReadableUrl": "https://catalysphere.co.jp/en/articles/ai-workflow-redesign/index.json" }, { "slug": "rag-poisoning-defense", "title": "The knowledge base of RAG becomes a searchable attack surface.", "summary": "We mitigate the risk of malicious documents entering the top search results and guiding answers through write history, index monitoring, contradiction detection, post-generation verification, and isolation with re-indexing.", "quickAnswer": { "what": "RAG using external evidence creates an attack surface where a small number of malicious documents are prioritized in search for specific questions.", "why": "We control write permissions, history, search results, and answer verification separately, rather than relying solely on safety instructions to the model.", "action": "After discovering contamination, we do not merely delete the document but reconstruct the index, track affected answers, and issue corrections." }, "updatedAt": "2026-10-01", "url": "https://catalysphere.co.jp/en/articles/rag-poisoning-defense/", "machineReadableUrl": "https://catalysphere.co.jp/en/articles/rag-poisoning-defense/index.json" }, { "slug": "multi-agent-debate-reliability", "title": "Gathering any number of AIs does not result in independent wisdom.", "summary": "Do not trust multi-LLM discussions based on vote counts; instead, determine the added value by assessing the diversity of initial answers, conformity, problem drift, retention of correct answers, and the difference from simple voting.", "quickAnswer": { "what": "Because multiple answers derived from the same model may have correlated errors, the number of agents cannot be treated as the number of independent pieces of evidence.", "why": "Although discussions may improve on certain tasks, they can lose the initial correct answer due to conformity or problem drift.", "action": "Compare against single reasoning, self-consistency, and majority vote with the same budget, and adopt only under conditions where there is a clear added value." }, "updatedAt": "2026-10-01", "url": "https://catalysphere.co.jp/en/articles/multi-agent-debate-reliability/", "machineReadableUrl": "https://catalysphere.co.jp/en/articles/multi-agent-debate-reliability/index.json" }, { "slug": "ai-code-security-review", "title": "Code written by AI must be reviewed for speed and safety separately.", "summary": "Do not equate the productivity of generated code with safety; instead, treat vulnerabilities, user safety perception, static and dynamic testing, and independent review as gates for adopting high-impact code.", "quickAnswer": { "what": "Even if code generation is fast, it does not necessarily mean fewer vulnerabilities, and outcomes depend on the issues addressed and the selection of proposals.", "why": "The user's perception of safety and the actual reduction of vulnerabilities through testing are distinct metrics.", "action": "For authentication, cryptography, input processing, and similar areas, require records of the generation source, dedicated testing, and independent review as adoption conditions." }, "updatedAt": "2026-07-24", "url": "https://catalysphere.co.jp/en/articles/ai-code-security-review/", "machineReadableUrl": "https://catalysphere.co.jp/en/articles/ai-code-security-review/index.json" }, { "slug": "synthetic-training-data-model-collapse", "title": "Recursive training on synthetic data can lose distribution tails before the average collapses", "summary": "Instead of uniformly prohibiting the reuse of synthetic data, we design learning data operations that separate the replacement, accumulation, and selection of real data to preserve rare cases and diversity.", "quickAnswer": { "what": "Synthetic data does not automatically cause model collapse, but recursive learning that replaces real data can lose the tails of the distribution and diversity.", "why": "This is because degradation depends not on the material name, but on how the retention of real data, reuse across generations, sample selection, and errors from finite data overlap.", "action": "Preserve every sample’s provenance and generation, test rare groups, tails and diversity separately from average performance, and quarantine deteriorating sources or categories before restoring the verified baseline and retesting." }, "updatedAt": "2026-10-01", "url": "https://catalysphere.co.jp/en/articles/synthetic-training-data-model-collapse/", "machineReadableUrl": "https://catalysphere.co.jp/en/articles/synthetic-training-data-model-collapse/index.json" }, { "slug": "model-data-drift-monitoring", "title": "Production models age even when the code does not change", "summary": "Monitoring only changes in input distribution does not reveal actual harm; design multi-layered monitoring and response procedures that connect data, predictions, decisions, and delayed outcomes.", "quickAnswer": { "what": "Drift monitoring is a mechanism that tracks whether changes have harmed model performance and business judgments, rather than merely detecting distribution differences.", "why": "This is because harm-free input changes and degradation where relationships change while inputs remain the same cannot be distinguished by a single metric.", "action": "Stratified monitoring of pipelines, inputs, predictions, actions, and outcomes determines investigation, degradation, and retraining based on impact and reversibility." }, "updatedAt": "2026-10-01", "url": "https://catalysphere.co.jp/en/articles/model-data-drift-monitoring/", "machineReadableUrl": "https://catalysphere.co.jp/en/articles/model-data-drift-monitoring/index.json" }, { "slug": "prompt-injection-enterprise-controls", "title": "Prompt-injection defense starts with permission boundaries, not instruction detection", "summary": "Filtering suspicious text alone cannot protect AI that reads external data and executes tools; we organize multi-layered defense through attack evaluation, structured input, and official classification.", "quickAnswer": { "what": "Prompt injection cannot be resolved by textual warnings alone; it must be prevented at system boundaries encompassing permissions and data flows.", "why": "This is because models cannot fully separate trusted commands from commands within external data, and when tool permissions are present, erroneous following can lead to actual operations.", "action": "We layer separation of commands and data, least privilege, permitted operations, pre-execution confirmation, adaptive attack testing, and audit logs." }, "updatedAt": "2026-10-01", "url": "https://catalysphere.co.jp/en/articles/prompt-injection-enterprise-controls/", "machineReadableUrl": "https://catalysphere.co.jp/en/articles/prompt-injection-enterprise-controls/index.json" }, { "slug": "multimodal-document-accuracy", "title": "Answering questions about a PDF does not prove document understanding", "summary": "Being able to extract text via OCR is distinct from correctly linking layout, graphs, tables, and body text; we organize the unit of business verification based on research in document question answering.", "quickAnswer": { "what": "While multimodal AI assists with document search and drafting, any use case that determines values or evidence without verification requires individual, step-by-step validation.", "why": "This is because different capabilities are required beyond character recognition, including placement, legends, units, correspondence between tables and body text, arithmetic, and integration across multiple locations.", "action": "We separate evaluations by document type and question type, designing the system to verify the page, region, original value, and calculation process for each answer." }, "updatedAt": "2026-10-01", "url": "https://catalysphere.co.jp/en/articles/multimodal-document-accuracy/", "machineReadableUrl": "https://catalysphere.co.jp/en/articles/multimodal-document-accuracy/index.json" }, { "slug": "small-model-routing-economics", "title": "Using the strongest model for every request is not always the best investment", "summary": "Designing a system that sends all requests to a high-performance model, or one that fixes them to a low-cost model, is not necessarily optimal; drawing from cascade and router research, we consider routing under quality constraints.", "quickAnswer": { "what": "Rather than deciding on a single model size for the entire company, it may be more rational to choose a low-cost path or a high-performance path for each request.", "why": "This is because the difficulty of requests and the strengths and weaknesses of models are heterogeneous, leaving room to reduce expensive calls while maintaining quality.", "action": "We measure the internal input's quality floor, misrouting loss, retries, and latency, and verify rules that escalate to the safer side when confidence is low or the request is high-impact, out of scope, or unverifiable." }, "updatedAt": "2026-10-01", "url": "https://catalysphere.co.jp/en/articles/small-model-routing-economics/", "machineReadableUrl": "https://catalysphere.co.jp/en/articles/small-model-routing-economics/index.json" }, { "slug": "ai-agent-workflow-reliability", "title": "AI-agent failures persist as state across long workflows", "summary": "Even with high accuracy in a single response, errors accumulate in tasks involving continuous planning, search, tool execution, and state updates; we consider the boundaries of reliable automation based on real-world environment benchmarks.", "quickAnswer": { "what": "The reliability of an AI agent is determined not by the impression of the conversation, but by whether the final state is correct and reproducible for the same request.", "why": "This is because in long processes, small errors in planning, observation, tool selection, arguments, business rules, and state confirmation are carried over to the next operation.", "action": "Incorporate reversibility in sub-processes, state verification, permission limits, human approval, iterative testing, and immediate stop into the workflow." }, "updatedAt": "2026-10-01", "url": "https://catalysphere.co.jp/en/articles/ai-agent-workflow-reliability/", "machineReadableUrl": "https://catalysphere.co.jp/en/articles/ai-agent-workflow-reliability/index.json" }, { "slug": "rag-grounding-reliability", "title": "To what extent can RAG improve the factual accuracy of generative AI?", "summary": "Providing evidence via search does not necessarily eliminate errors; this article outlines methods for verifying RAG in business operations by distinguishing between search omissions, inconsistencies with evidence, and oversights in long contexts.", "quickAnswer": { "what": "While RAG can create pathways to external knowledge, it does not currently guarantee with sufficient certainty that answers remain faithful to the retrieved evidence.", "why": "This is because separate failures in search recall, document quality, evidence placement, and generation-time faithfulness are connected in series.", "action": "We separately score search and evidence utilization and final answers, and establish criteria for safely withholding questions that cannot be answered." }, "updatedAt": "2026-07-24", "url": "https://catalysphere.co.jp/en/articles/rag-grounding-reliability/", "machineReadableUrl": "https://catalysphere.co.jp/en/articles/rag-grounding-reliability/index.json" }, { "slug": "ai-coding-productivity-context", "title": "When does AI coding accelerate development, and when does it slow it down?", "summary": "Average effects of AI coding are not enough for investment decisions. Research from laboratories, enterprises, and mature open-source projects reveals the work, personnel, and measurement conditions that determine whether development speeds up.", "quickAnswer": { "what": "While AI coding may accelerate tasks with clear specifications that are easy to verify, it can slow down maintenance work requiring deep contextual understanding.", "why": "This is because the burden of explanation, waiting, verification, review, and correction varies by task, not just the generation time of the output.", "action": "We categorize work by verifiability and contextual depth, comparing the period from initiation to post-merge rework with and without AI assistance." }, "updatedAt": "2026-07-24", "url": "https://catalysphere.co.jp/en/articles/ai-coding-productivity-context/", "machineReadableUrl": "https://catalysphere.co.jp/en/articles/ai-coding-productivity-context/index.json" } ] }