[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-a-new-way-to-fake-enterprise-data-for-training-ai-agents":10,"sections":35},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":30,"feedback":34,"feedback_at":22,"cost_usd":34,"total_tokens":34},10832,"a-new-way-to-fake-enterprise-data-for-training-ai-agents","A New Way to Fake Enterprise Data for Training AI Agents","SAP researchers had an AI agent simulate enterprise software to generate training data, sidestepping the schemas and seed data real company systems require.","Researchers built an AI agent that generates fake enterprise data by actually using it, running real operations through simulated business software instead of synthesizing database rows from scratch.\n\nThe approach, called Synthesis Through Simulation, has an LLM agent call policy-enforcing APIs inside simulated enterprise environments rather than generating a table directly. Because every action has to pass through the same system that defines what counts as valid, the output is structurally correct by default. The team's domain-agnostic agent, the Generalist Populator, hit 0.88 average marginal fidelity and 100 percent constraint satisfaction across ten test environments, without ever seeing a database schema. Traditional statistical synthesizers couldn't even run on seven of those ten environments, because they need real seed data to get started. The framework, all ten environments, and the generated datasets are open-sourced on GitHub under SAP's account.\n\nTraining tool-calling agents for enterprise work is bottlenecked less by compute than by access: real company data and schemas are usually locked down for legal and business reasons. This method generates usable training data without needing that access at all. It also undercuts an assumption worth noting elsewhere: agents given schema access failed 82 percent of trajectories on the airline booking environment, tripped up by brittle task composition, while the schema-free agent did not share that weakness.\n\nStill, this is a self-reported benchmark on ten environments the researchers built themselves. Simulated enterprise chaos and the real thing are not always the same animal.","[\"synthetic-data\",\"ai-agents\",\"enterprise-ai\",\"open-source\"]","2026-10-09T04:00:00.000Z","2026-10-09T17:22:02.841Z","2026-10-09T17:22:07.095Z","published",null,[],"ai",[26,27,28,29],"synthetic-data","ai-agents","enterprise-ai","open-source",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.10549",0,{"sections":36},[37,40,44,49,54,59,63,68,73,78,83,88,93,98],{"name":38,"slug":24,"count":39,"latest_published_at":18},"AI",6608,{"name":41,"slug":42,"count":43,"latest_published_at":18},"Security","security",926,{"name":45,"slug":46,"count":47,"latest_published_at":48},"Policy","policy",486,"2026-10-08T22:40:11.000Z",{"name":50,"slug":51,"count":52,"latest_published_at":53},"Deals","deals",474,"2026-10-08T22:00:00.000Z",{"name":55,"slug":56,"count":57,"latest_published_at":58},"Hardware","hardware",229,"2026-10-08T20:47:10.000Z",{"name":60,"slug":61,"count":62,"latest_published_at":18},"Science","science",192,{"name":64,"slug":65,"count":66,"latest_published_at":67},"Consumer Tech","consumer-tech",181,"2026-10-08T23:26:35.000Z",{"name":69,"slug":70,"count":71,"latest_published_at":72},"Startups","startups",117,"2026-10-08T16:45:00.000Z",{"name":74,"slug":75,"count":76,"latest_published_at":77},"Software","software",114,"2026-10-08T17:57:01.000Z",{"name":79,"slug":80,"count":81,"latest_published_at":82},"Dev Tools","dev-tools",105,"2026-10-07T16:59:11.000Z",{"name":84,"slug":85,"count":86,"latest_published_at":87},"General","general",66,"2026-10-09T04:46:11.000Z",{"name":89,"slug":90,"count":91,"latest_published_at":92},"Gaming","gaming",58,"2026-10-08T20:08:45.000Z",{"name":94,"slug":95,"count":96,"latest_published_at":97},"Reviews","reviews",34,"2026-10-08T14:00:22.000Z",{"name":99,"slug":100,"count":101,"latest_published_at":102},"How-To","how-to",8,"2026-10-05T09:00:00.000Z"]