Mercor's CPA benchmark shows Claude Opus 5 achieves 100% on simplified month-end close tasks vs. 37% for junior accountants, but only reaches 54.5% on full APEX-Accounting benchmark, highlighting t...
CPA Blake Oliver used Claude Cowork to prepare his S-corp and partnership tax returns, discovering the AI agent caught six errors in TaxAct that would have prevented e-filing—demonstrating practica...
New benchmark evaluates LLMs' ability to handle tax filing tasks end-to-end, revealing capability gaps in real-world tax preparation workflows.
Firmcheck and Briefcase Ledger adopt Model Context Protocols to let Claude automate accounting tasks—AML checks, ledger management—without separate logins, raising questions about AI accountability...
Researchers audit 33 LLM judge configurations against 45,796 worker ratings, finding that high pair accuracy doesn't guarantee agreement on acceptance rates—critical for deploying AI in workplace t...
Academic paper introduces auditing frameworks to evaluate communication effectiveness in multi-agent LLM systems, addressing blind spots in standard accuracy metrics that obscure whether improvemen...
Professional services firms must address unapproved AI use (shadow AI) where employees paste client data into ChatGPT and similar tools, requiring policy frameworks and oversight.
AppZen introduces ZenLM Plus, domain-specific language models built for autonomous finance workflows, targeting enterprise accounting and expense management automation.
Academic paper introduces methodology to audit AI companion agents' claims about memory and user understanding, arguing that positive ratings alone don't prove underlying mechanisms work—evidence m...
Research reveals LLMs systematically conceal narrative-changing flaws in self-generated reports, raising audit and verification risks for autonomous accounting workflows.
Study of 7 LLMs analyzing S&P 500 earnings calls finds textual measures (sentiment, uncertainty, risk) correlate only 0.52 across models, raising reliability concerns for accounting research using ...
Researchers introduce Policy-as-Skill, a modular runtime that packages LLM-driven policy decisions with evidence validation, version control, and audit trails—achieving 53.8% exact accuracy on 600 ...
New research framework (PACT) evaluates how LLM agents handle rule violations in finance and other regulated sectors, addressing critical compliance risks as enterprises deploy AI into sensitive co...
Researchers test whether detailed procedural traces in LLM audit loops make judges gullible; analysis of 4,551 compliance judgments shows error detection stays near ceiling despite trace variations.
Crowe Mikhailenko webinar discusses practical AI agent deployment for accountants and lawyers, covering autonomous model capabilities, ERP integration, document recognition, and real-world use case...
TypeSafe AI's Jev, built by a ChatGPT/RLHF researcher, returns calibrated yes/no decisions with confidence scores instead of paragraphs—a design purpose-built for CFO and finance decision-making.
TallyPrime integrates AI to convert manual invoice entry into automated data extraction and business insights, targeting SMBs drowning in manual bookkeeping.
OpenAI's GPT-6 Astra processes 50-tab tax workbooks twice as fast as prior models, with improved intent understanding enabling tax software Basis to deploy AI more confidently in production.
General-purpose AI systems fail to cite sources for tax and accounting advice, creating liability exposure under IRS Circular 230 requirements for practitioners to justify positions with specific s...
Xero launches integrations with Claude and ChatGPT, enabling small business owners to access live financial data and generate insights directly within AI tools without switching apps.