[{"data":1,"prerenderedAt":119},["ShallowReactive",2],{"q_Hx9vvpvoNiUodL4TrPdL0x6WR_S9lm0CuLlsW5eF4":3,"_apollo:default":117,"_apollo:identified":118},{"seo":4,"posts":15},{"social":5,"openGraph":11,"__typename":14},{"twitter":6,"__typename":10},{"cardType":7,"username":8,"__typename":9},"summary_large_image","dassault3ds","SEOSocialTwitter","SEOSocial",{"defaultImage":12,"__typename":13},null,"SEOOpenGraph","SEOConfig",{"nodes":16,"__typename":116},[17],{"id":18,"slug":19,"title":20,"uri":21,"excerpt":22,"locale":23,"featuredImage":26,"tableOfContents":34,"content":40,"date":41,"translations":42,"author":43,"tags":56,"globalTags":78,"brands":90,"keywords":101,"seo":104,"__typename":115},"cG9zdDozMTc0OTM=","closed-is-not-learned-how-ai-helps-turn-quality-records-into-answers","Closed Is Not Learned: How AI Helps Turn Quality Records into Answers","\u002Fbrands\u002Fbiovia\u002Fclosed-is-not-learned-how-ai-helps-turn-quality-records-into-answers","\u003Cp>Earlier signals, better investigations.\u003C\u002Fp>\n",{"locale":24,"__typename":25},"en_US","Locale",{"node":27,"__typename":33},{"large":28,"__typename":29,"medium_large":28,"thumbnail":30,"srcSet":31,"sizes":32},"https:\u002F\u002Fblog-assets.3ds.com\u002Fuploads\u002F2026\u002F10\u002F102326-quality-and-marie-blog-banner.jpg","MediaItem","https:\u002F\u002Fblog-assets.3ds.com\u002Fuploads\u002F2026\u002F10\u002F102326-quality-and-marie-blog-banner-150x150.jpg","https:\u002F\u002Fblog-assets.3ds.com\u002Fuploads\u002F2026\u002F10\u002F102326-quality-and-marie-blog-banner-300x186.jpg 300w, https:\u002F\u002Fblog-assets.3ds.com\u002Fuploads\u002F2026\u002F10\u002F102326-quality-and-marie-blog-banner.jpg 587w","(max-width: 300px) 100vw, 300px","NodeWithFeaturedImageToMediaItemConnectionEdge",[35,36,37,38,39],"The gap between closure and learning|the-gap-between-closure-and-learning-0","A problem of scale, not diligence|a-problem-of-scale-not-diligence-1","What learning requires|what-learning-requires-2","Where MARIE fits|where-marie-fits-3","A better measure of a quality system|a-better-measure-of-a-quality-system-4","\n\u003Cp>Every quality system holds early signals of what could go wrong next. The problem is where those signals sit: buried in records already closed, filed and forgotten.\u003C\u002Fp>\n\n\n\n\u003Cp>A deviation is investigated, a CAPA is signed, and the record is closed. By every measure the quality system tracks, the job is done. Yet the organization has learned very little, because almost no one will read that record again.\u003C\u002Fp>\n\n\n\n\u003Ch2 class=\"wp-block-heading\" id=\"the-gap-between-closure-and-learning-0\">The gap between closure and learning\u003C\u002Fh2>\n\n\n\n\u003Cp>Quality systems are largely built to close records. Cycle time, overdue counts and closure rates prove that every event was handled. Handling an event and learning from it are different things, though. A closed record is a finished argument written for an auditor, and it is rarely retrievable in a way that helps the next investigator.\u003C\u002Fp>\n\n\n\n\u003Cp>The cost takes several forms. Investigators spend the first hours of a new deviation rebuilding context: has the equipment appeared before, did a similar complaint arrive from another site, did an earlier CAPA already address the issue? Categorization drifts, so the same event becomes a “process deviation,” an “equipment issue” or an “operator error” depending on the analyst, the site and the week. And systemic issues, built from events that each look minor, stay invisible until a regulatory finding, a delayed batch release or reputational damage to the brand makes the pattern impossible to miss.\u003C\u002Fp>\n\n\n\n\u003Cp>Life-sciences quality teams are data-rich and insight-poor. Each record is complete, but collectively the records are mute.\u003C\u002Fp>\n\n\n\n\u003Ch2 class=\"wp-block-heading\" id=\"a-problem-of-scale-not-diligence-1\">A problem of scale, not diligence\u003C\u002Fh2>\n\n\n\n\u003Cp>This is not a failure of diligence. A skilled investigator can weigh a single case with remarkable depth, but no one can hold hundreds of records, written by different authors across different sites and years, in mind at once and notice the faint resemblance among a few of them.\u003C\u002Fp>\n\n\n\n\u003Cp>Scale also exposes a structural blind spot: every record has an owner, but no pattern does. A deviation belongs to the investigator assigned to it. The resemblance between that deviation and two others, closed months apart at different sites, belongs to no one, because it lives between records and not inside any single one.\u003C\u002Fp>\n\n\n\n\u003Cp>More reviews, longer checklists and additional reporting do not solve a problem of this shape. They ask people to work harder at something human attention cannot do at volume. What it needs is support that reads across records, giving patterns a place to surface.\u003C\u002Fp>\n\n\n\n\u003Ch2 class=\"wp-block-heading\" id=\"what-learning-requires-2\">What learning requires\u003C\u002Fh2>\n\n\n\n\u003Cp>Moving from closing records to learning from them starts not with technology but with a change in what an organization expects of its own quality data. The change rests on a few shifts.\u003C\u002Fp>\n\n\n\n\u003Cp>\u003Cstrong>From archive to connection. \u003C\u002Fstrong>An archive stores each record on its own. Connected quality data relates every new record to the past ones, surfacing the most similar cases the instant a record is opened.\u003C\u002Fp>\n\n\n\n\u003Cp>\u003Cstrong>From search to question. \u003C\u002Fstrong>Keyword search assumes the investigator already knows the right terms, and categorization drift means they often don’t. Investigators think in questions: what is going on with this record, why does it look familiar, what is stopping it from moving? Quality data should meet investigators there, in plain language.\u003C\u002Fp>\n\n\n\n\u003Cp>\u003Cstrong>From reactive to proactive. \u003C\u002Fstrong>Most quality metrics describe the past, and most quality responses begin only after a problem has happened. Proactive quality management needs a comparative view, showing which events resemble each other across products, sites and time before the pattern earns a name in an inspection report.\u003C\u002Fp>\n\n\n\n\u003Cp>None of this removes the investigator. It moves the effort from assembling context to evaluating it, which is the part of the work that depends on experience.\u003C\u002Fp>\n\n\n\n\u003Ch2 class=\"wp-block-heading\" id=\"where-marie-fits-3\">Where MARIE fits\u003C\u002Fh2>\n\n\n\n\u003Cp>\u003Ca href=\"https:\u002F\u002Fwww.3ds.com\u002Fproducts\u002Fbiovia\u002Fmarie\" target=\"_blank\" rel=\"noreferrer noopener\">MARIE\u003C\u002Fa>, BIOVIA’s scientific Virtual Companion, was built for these shifts inside the Biopharma \u003Ca href=\"https:\u002F\u002Fwww.3ds.com\u002Fproducts\u002Fbiovia\u002Fquality-management\" target=\"_blank\" rel=\"noreferrer noopener\">Quality Management\u003C\u002Fa> solution. It is the conversational layer that lets users question everything the solution’s quality workflows produce.\u003C\u002Fp>\n\n\n\n\u003Cp>A question in plain language returns a structured answer: what is happening, why it matters and suggested next steps. In investigations, MARIE can:\u003C\u002Fp>\n\n\n\n\u003Cul class=\"wp-block-list\">\n\u003Cli>\u003Cstrong>Summarize a record\u003C\u002Fstrong> so its history is clear without reading every entry.\u003C\u002Fli>\n\n\n\n\u003Cli>\u003Cstrong>Identify blockers\u003C\u002Fstrong>, showing where a record has stalled and why.\u003C\u002Fli>\n\n\n\n\u003Cli>\u003Cstrong>Compare records\u003C\u002Fstrong>, finding the similarities and differences between a current event and past ones.\u003C\u002Fli>\n\u003C\u002Ful>\n\n\n\n\u003Cp>Behind this sits the solution’s AI-assisted analysis of historical quality data, which spots recurring deviations, identifies patterns, suggests likely root causes and flags emerging risks earlier. Each suggestion is a starting point for the investigator to evaluate, not a conclusion.\u003C\u002Fp>\n\n\n\n\u003Cp>Take a new deviation on a packaging line. Instead of a morning spent pulling old records, the investigator asks whether anything similar has happened before and receives the most similar past events, what they had in common and how those investigations concluded. The investigator still decides whether the resemblance is real, but starts there instead of at zero.\u003C\u002Fp>\n\n\n\n\u003Cp>In a regulated discipline, restraint matters as much as capability. MARIE never alters records, and human authors and reviewers stay in control of every decision. It works from a single source of truth with data integrity, digital continuity and automated audit trails built in, and every answer traces back to its source. All of this operates within the secure \u003Cstrong>3D\u003C\u002Fstrong>EXPERIENCE platform, where an organization’s data stays under its own control.\u003C\u002Fp>\n\n\n\n\u003Ch2 class=\"wp-block-heading\" id=\"a-better-measure-of-a-quality-system-4\">A better measure of a quality system\u003C\u002Fh2>\n\n\n\n\u003Cp>When quality metrics are next reviewed, the more useful question is not how quickly records closed but how much the organization learned from them. How long does it take an investigator to learn whether a deviation has happened before? How consistently would two analysts describe it? How early would a pattern across sites become visible?\u003C\u002Fp>\n\n\n\n\u003Cp>The organizations that answer these questions well will not be the ones with the most records. They will be the ones whose records work together, so that early warning signs surface well before an inspector, a recall or a headline forces the question.\u003C\u002Fp>\n\n\n\n\u003Cp>To see MARIE in action watch the \u003Ca href=\"https:\u002F\u002Fevents.3ds.com\u002Fmeet-marie-scientific-virtual-companion\" target=\"_blank\" rel=\"noreferrer noopener\">webinar\u003C\u002Fa> series.\u003C\u002Fp>\n\n\n\n\u003Chr class=\"wp-block-separator has-alpha-channel-opacity\" \u002F>\n","2026-10-27T11:00:00",[],{"node":44,"__typename":55},{"nicename":45,"description":46,"slug":47,"name":47,"firstName":48,"lastName":49,"avatar":50,"__typename":54},"jaya-tikhe","Jaya Tikhe has extensive experience in early drug discovery and project management. As a medicinal and organic chemist, she has a proven track record of delivering high-quality drug candidates across CNS, immunology, inflammation, and oncology. She currently leads product marketing for Dassault Systèmes’ Small Molecule Drug Discovery, Manufacturing, Quality, and Regulatory solutions.","Jaya Tikhe","Jaya","Tikhe",{"default":51,"url":52,"__typename":53},"mm","https:\u002F\u002Fblog-assets.3ds.com\u002Fuploads\u002F2025\u002F04\u002Fjaya-tikhe-96x96.jpg","Avatar","User","NodeWithAuthorToUserConnectionEdge",{"edges":57,"nodes":70,"__typename":77},[58,65],{"isPrimary":59,"node":60,"__typename":64},false,{"slug":61,"name":62,"__typename":63},"company-news","Company News","Taxonomy_topic","PostToTaxonomy_topicConnectionEdge",{"isPrimary":66,"node":67,"__typename":64},true,{"slug":68,"name":69,"__typename":63},"science","Science",[71,74],{"id":72,"name":62,"uri":73,"__typename":63},"dGVybTo4NTUz","\u002Ftopics\u002Fcompany-news\u002F",{"id":75,"name":69,"uri":76,"__typename":63},"dGVybTo4NjEx","\u002Ftopics\u002Fscience\u002F","PostToTaxonomy_topicConnection",{"nodes":79,"__typename":89},[80,85],{"id":81,"name":82,"uri":83,"__typename":84},"dGVybTo4ODM0","3DEXPERIENCE platform","\u002Ftags\u002F3dexperience-platform\u002F","Taxonomy_tag",{"id":86,"name":87,"uri":88,"__typename":84},"dGVybTo4ODMy","AI","\u002Ftags\u002Fai\u002F","PostToTaxonomy_tagConnection",{"edges":91,"nodes":98,"__typename":100},[92],{"isPrimary":66,"node":93,"__typename":97},{"slug":94,"name":95,"__typename":96},"biovia","BIOVIA","Taxonomy_brand","PostToTaxonomy_brandConnectionEdge",[99],{"name":95,"slug":94,"__typename":96},"PostToTaxonomy_brandConnection",{"nodes":102,"__typename":103},[],"PostToTaxonomy_keywordConnection",{"title":105,"metaDesc":106,"opengraphAuthor":107,"opengraphDescription":106,"opengraphTitle":20,"opengraphUrl":108,"opengraphSiteName":109,"opengraphPublishedTime":110,"opengraphModifiedTime":111,"twitterTitle":107,"twitterDescription":107,"readingTime":112,"metaRobotsNoindex":113,"__typename":114},"Closed Is Not Learned: How AI Helps Turn Quality Records into Answers - Dassault Systèmes blog","Discover how MARIE helps life sciences teams uncover patterns in deviations, improve CAPA investigations, and enable proactive quality management with AI.","","https:\u002F\u002Fblog.3ds.com\u002Fbrands\u002Fbiovia\u002Fclosed-is-not-learned-how-ai-helps-turn-quality-records-into-answers\u002F","Dassault Systèmes blog","2026-10-27T11:00:00+00:00","2026-10-27T11:00:39+00:00",5,"index","PostTypeSEO","Post","RootQueryToPostConnection",{},{},1793099542847]