Execs Want GenAI, But Aren't Ready
Enterprise executives are eager to implement generative AI, but a new analysis finds most companies lack the in-house expertise and unified data strategy to succeed. This gap often leads to stalled projects and “AI-washing,” as leaders struggle with the real-world governance and trust issues required to scale AI tools effectively.
While enterprise adoption of generative AI tools jumped from 11% in early 2023 to 65% by 2024, a significant portion of projects fail to move beyond the pilot phase. One study found 42% of companies abandon most of their AI initiatives before they reach production, a sharp increase from 17% the prior year. The primary obstacle remains a shortage of in-house talent, with 42% of organizations citing inadequate generative AI expertise as a major challenge. The demand for specialized roles like machine learning engineers, data scientists, and prompt engineers far outstrips supply, leading to inflated recruitment costs and project bottlenecks. Poor data quality is another critical failure point, with 58% of IT managers identifying it as the main obstacle to their AI projects. Many firms lack the unified data platforms and rigorous data hygiene practices necessary to train reliable models, leading to AI systems that produce biased or nonsensical results. One appliance manufacturer's AI service agent, for example, merged instructions from over 100 different manuals, resulting in a "complete mess" for customers. The costs associated with GenAI are also a significant hurdle, with 70% of executives reporting that it's driving up their IT budgets. A custom enterprise-grade AI solution can cost anywhere from $300,000 to over $1 million, with ongoing cloud computing expenses, data preparation, and integration adding tens of thousands more. This pressure to innovate, combined with a lack of readiness, has led to a rise in "AI-washing"—the practice of exaggerating AI capabilities to appear more advanced. In 2025, over 50,000 layoffs were publicly attributed to AI, yet research suggests less than 1% of total job losses could be directly traced to the technology, with many companies simply using AI as a convenient excuse for financially motivated cuts. Governance remains a persistent challenge as generative models produce open-ended content that is difficult to evaluate for accuracy and safety. This has led to high-profile failures, such as a Microsoft-powered chatbot for New York City that incorrectly advised business owners to break the law. The path to successful implementation often involves starting with smaller, well-defined use cases, like automating tedious reporting processes, rather than attempting a complete operational overhaul. Top use cases currently driving adoption include code generation for software development, customer support chatbots, and enterprise search functions. Ultimately, successful AI integration is treated as a continuous business capability, not just a technical project, requiring ongoing monitoring, clear ownership, and cross-functional accountability to manage financial, legal, and reputational risks.
