{"rewrite":{"id":"r_ecf6fc5a9dbee9f9cd31eb52","clusterId":"c_8dcc165e301fdd5bf94b7719","slug":"fraim-runs-legal-knowledge-poc-with-money-forward","model":"deepseek-v4-1-flash","headline":"FRAIM Runs Legal Knowledge PoC with Money Forward","summary":"FRAIM ran a proof-of-concept with Money Forward between April and June 2026, turning internal legal consultation logs into reusable structured knowledge. Using correct-answer labels for roughly 18,000 consultation categories and several months of consultation threads, FRAIM built playbooks and citation knowledge graphs, then let category-specific AI agents draw on them. Staff approvals and corrections feed back into playbook updates and classification retraining.","whyItMatters":"The PoC tests whether legal advice buried in chat threads can be turned into a maintained asset rather than relying on the memory of whoever answers, with human sign-off kept as the final step.","webCardHtml":"\u003cp\u003eThe design leans on FRAIM AI Agent, the company\u0026#39;s document-work agent, with an eye toward deployment on the FRAIM Agentic Platform for organizations that work alongside AI. Past logs went through three steps: analysis that breaks consultations into intent, question and answer, points to confirm, and basis for judgment; a model that classifies consultations using category metadata as labels; and a citation knowledge graph linking consultations, answers, confirmation axes, FAQs, and evidence. Playbooks hold confirmation checklists, anticipated FAQs, and citations of the basis for judgment. Agents search similar consultations, present points to confirm, and surface evidence, with staff giving final confirmation.\u003c/p\u003e","blueskyPost":"FRAIM ran a legal knowledge PoC with Money Forward from April to June 2026, using about 18,000 consultation category labels and several months of threads to build playbooks and citation graphs that category-specific AI agents draw on, with staff confirming the final answer.","twitterPost":"FRAIM and Money Forward ran an April to June 2026 PoC turning legal consultation logs into playbooks and citation knowledge graphs, with roughly 18,000 category labels and human confirmation kept as the last step before an answer stands.","threadsPost":null,"newsletterBlurb":"FRAIM and Money Forward ran a proof-of-concept from April to June 2026 on reusing internal legal consultation knowledge. Correct-answer labels for roughly 18,000 consultation categories and several months of threads fed playbooks, citation knowledge graphs, and category-specific AI agents. Staff approvals and corrections feed back into the playbooks and classification model.","attributionJson":"[{\"source\":\"ASCII.jp\",\"url\":\"https://ascii.jp/elem/000/004/442/4442378/?rss\",\"title\":\"FRAIM Conducts Proof-of-Concept with Money Forward on AI for Utilizing Legal Consultation Knowledge\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":4552,"outputTokens":574,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1791951735,"createdAt":"2026-10-14T04:18:41.000Z","publishedAt":"2026-10-14T04:21:10.000Z","updatedAt":"2026-10-14T04:21:10.000Z"},"cluster":{"id":"c_8dcc165e301fdd5bf94b7719","canonicalTitle":"FRAIM、マネーフォワードと法務相談ナレッジ活用AIの実証実験を実施","representativeArticleId":"a_189636e94e32d396559d9158","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[],\"studios\":[],\"people\":[],\"type\":\"business\",\"domain\":\"industry\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-10-14T01:00:00.000Z","lastSeenAt":"2026-10-14T01:00:00.000Z","updatedAt":"2026-10-14T04:21:06.000Z"},"attribution":[{"source":"ASCII.jp","url":"https://ascii.jp/elem/000/004/442/4442378/?rss","title":"FRAIM、マネーフォワードと法務相談ナレッジ活用AIの実証実験を実施"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":[],"studios":[],"people":[],"type":"business","domain":"industry","is_roundup":false},"keyFacts":null}