[{"data":1,"prerenderedAt":28},["ShallowReactive",2],{"nr-en-ki-modelle-schlagen-buchhalter-strukturierte-aufgaben":3},{"slug":4,"title":5,"dek":6,"date":7,"time":8,"publishedAt":9,"updated":10,"updatedAt":10,"dateFmt":11,"updatedFmt":10,"kind":12,"tier":13,"author":14,"authorName":15,"topics":16,"tracker":10,"trackerLabel":10,"headlineStat":22,"image":23,"ogImage":24,"imageAlt":5,"csv":10,"minutes":25,"words":26,"html":27},"ki-modelle-schlagen-buchhalter-strukturierte-aufgaben","AI Models Beat Licensed Accountants on Structured Tasks","A Mercor study shows current AI models are faster and more accurate than certified accountants on bookkeeping tasks – but they still can't close the books without human oversight.","2026-10-02","15:44","2026-10-02T15:44:00+02:00","","October 2, 2026","daten","standard","ideal-syka","Ideal Syka",[17,18,19,20,21],"AI models","Accounting","Benchmark","Automation","Auditing","AI models outperform accountants on structured tasks – but still can't independently close the books","\u002Fnewsroom\u002Fimg\u002Fki-modelle-schlagen-buchhalter-strukturierte-aufgaben.webp","\u002Fog-nr\u002Fki-modelle-schlagen-buchhalter-strukturierte-aufgaben.en.png",2,444,"\u003Cp>AI models have surpassed licensed accountants on structured bookkeeping tasks. A new study from Mercor pitted 12 licensed accountants with an average of five-and-a-half years of professional experience against leading AI systems. The result is striking: while the best models scored around 37 percent accuracy 18 months ago – matching the average accountant – they now solve the same simplified tasks nearly flawlessly. At the same time, AI systems cost a fraction of what accountants do.\u003C\u002Fp>\n\u003Ch2>Key Takeaways\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>\u003Cstrong>Claude Opus 5.5\u003C\u002Fstrong> leads the full APEX Accounting Benchmark with \u003Cstrong>61.8 percent\u003C\u002Fstrong> of evaluation criteria met, followed by other models such as \u003Cstrong>GPT-6 Astra\u003C\u002Fstrong> (57.9%)\u003C\u002Fli>\n\u003Cli>On simplified tasks, AI models outperform accountants in \u003Cstrong>speed and accuracy\u003C\u002Fstrong> – a leap forward in just 18 months\u003C\u002Fli>\n\u003Cli>On the complex APEX Benchmark with \u003Cstrong>160 tasks\u003C\u002Fstrong> across 10 simulated companies, \u003Cstrong>no model solves all tasks completely\u003C\u002Fstrong>\u003C\u002Fli>\n\u003Cli>Nearly \u003Cstrong>60 percent of tasks\u003C\u002Fstrong> are not fully solved by any model – AI still requires human oversight\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch2>The Benchmark: Simplified vs. Reality\u003C\u002Fh2>\n\u003Cp>Mercor tested two scenarios. On simplified APEX Accounting Benchmark tasks, AI's superiority is clear. The full benchmark is a different beast: 160 tasks spread across 10 simulated companies, developed by over 40 professionals averaging 11 years of experience. Here, the models hit a ceiling.\u003C\u002Fp>\n\u003Cp>Top performers Claude Opus 5.5 and GPT-6 Astra each score just above 60 percent of evaluation criteria. That sounds respectable – but it's incomplete. Nearly 60 percent of tasks aren't fully solved by any single model. Translation: without human oversight, AI can't independently close a fiscal year.\u003C\u002Fp>\n\u003Ch2>What AI Can Do – and Can't\u003C\u002Fh2>\n\u003Cp>Mercor acknowledges the tested tasks showcase what AI does best: \u003Cstrong>detailed information retrieval and precise instruction-following\u003C\u002Fstrong>. Real accountant work is broader. The study excluded:\u003C\u002Fp>\n\u003Cul>\n\u003Cli>Client communication\u003C\u002Fli>\n\u003Cli>Colleague consultations\u003C\u002Fli>\n\u003Cli>Domain-specific contextual knowledge\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>These aren't technicalities – they're what makes accountants irreplaceable. Yet the study also notes: \u003Cstrong>significant productivity gains across the industry are foreseeable\u003C\u002Fstrong>.\u003C\u002Fp>\n\u003Ch2>What This Means for You\u003C\u002Fh2>\n\u003Cp>For German mid-market companies and agencies, the signal is clear: AI won't replace accounting, but it will radically accelerate it. Routine tasks – capturing receipts, allocating accounts, consistency checks – will become faster and cheaper. That frees capacity for work requiring human judgment: strategic financial planning, anomaly detection, advisory services.\u003C\u002Fp>\n\u003Cp>Whoever invests in AI-powered accounting processes now gains competitive advantage. Those who wait risk competitors already capturing cost savings. The study proves it: the era of "AI isn't ready yet" is over – at least for structured financial workflows.\u003C\u002Fp>\n\u003Ch2>Sources\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fthe-decoder.de\u002Fki-modelle-uebertreffen-buchhalter-bei-strukturierten-aufgaben-sind-aber-noch-weit-von-eigenstaendigkeit-entfernt\u002F\">The Decoder (DE)\u003C\u002Fa>\u003C\u002Fli>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fthe-decoder.com\u002Fai-beats-licensed-accountants-on-speed-and-accuracy-but-still-cant-close-the-books-without-supervision\u002F\">The Decoder\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cem>Editorially owned by \u003Ca href=\"\u002Fen\u002Fautor\u002Fideal-syka\">Ideal Syka\u003C\u002Fa>. Sources and method: \u003Ca href=\"\u002Fen\u002Fredaktion\">Newsroom & method\u003C\u002Fa>. Tips and corrections: \u003Ca href=\"mailto:ai@i6eal.de\">ai@i6eal.de\u003C\u002Fa>.\u003C\u002Fem>\u003C\u002Fp>\n",1790962003643]