AI真相幻觉：企业级AI错误与数据可信度研究（英）

ForewordArtificial intelligence now dominates every boardroom. Everyone’s racing to deploy it but almost no one is asking the only question that matters:Can you trust the data?In the enterprise, accuracy isn’t optional. It’s the difference between clarity and chaos. When AI guesses, companies get: bad insights, bad decisions, compliance exposure, and a dangerous illusion of confidence.With autonomous agents, the stakes spike even higher. One bad data point becomes ten. Ten become a hundred. Risk compounds. Operations break.Terzo was built for this moment.Our belief is simple: AI must be grounded in truth. Not hype. Truth.That’s why NirvanAI, Terzo’s latest product release, is different. It’s not a chatbot sitting on top of messy PDFs—it’s a full-stack enterprise AI platform powered by a 99%-accurate data foundation with humans in the loop.Clean contract, spend, and revenue data. Trusted answers. Real outcomes.If you want GenAI that actually moves the needle, you start with the data layer. Data-layer intelligence + autonomous agents is the new operating system for finance and Terzo is already building it.— Brandon Card CEO, Terzo1The Age of AI Illusion Under the Spotlight: Revealing the Biggest AI Risks Cracks in the Act: Hallucination Rates of AI ModelsBehind the Scenes: How Often AI is Actually CheckedThe Confidence Trick: 5 Examples of AI HallucinationThe Reveal: What Really Builds Trust in AIGet Real Financial Results with NirvanAI NirvanAI: Built on Real Data, Not Assumptions345678109Table of Contents2AI has cast a spell over the business world—capturing imagination, budgets, and headlines. But the results tell a different story. Nearly every company is experimenting with AI, yet fewer than half can show real financial return. The issue isn’t ambition—it’s misplaced trust. What appears intelligent often hides inaccuracy, bias, and incomplete data. Leaders now face a choice: stay dazzled by the illusion, or look behind the curtain and demand proof. Because with AI, confidence without accuracy is a risk in disguise.The Age of AI Illusion88%of companies use AI in at least 1 business function39%report a company-wide financial impact** Measured using EBIT (earnings before interest and taxes). Source: McKinsey, online survey of 1,933 participants conducted June 25 to July 29, 2025.3Even the best illusions have their glitches. Just over half of companies have already seen AI backfire, and the most common risk, by far, is inaccuracy.Leaders are taking this risk seriously. Today, 54% of organizations are now working to mitigate AI’s inaccuracies, recognizing that every wrong output erodes trust, wastes resources, and compounds strategic risk.Every model makes mistakes, but some create errors much more often than others. Revealing the Biggest AI RisksUnder the Spotlight:14%Explainability11%Personal / Individual Privacy10%Cybersecurity8%RegulatoryCompliance8%Intellectual Property Infringement7%Unauthorized or Unintended Action7%Equity and Fairness6%Workforce