Opinion
AI transformation starts with building employee trust
Santhosh Keshavan, Chief Technology and Operations Officer at Voya Financial, believes that the success of AI transformation depends on whether employees are ready for and trust new technologies. Through approximately five hours of AI foundational training, company-wide participation in design, and internal innovation competitions, Voya achieved nearly 100% of employees obtaining literacy certification and 98% of employees using Copilot an average of 24 times per week within three months, successfully shifting employee mindsets from concern to curiosity.
4 leadership pain points that stall AI pilots — and how to fix them
AI spending is projected to grow 47% to $2.5 trillion by 2026, but most pilot projects fail to scale. A Zapier survey shows that only 13% of organizations have widely deployed AI projects. This article points out that leadership control issues are the core obstacle, covering four aspects: business problem ownership, technology fit and validation, full-cost evaluation, and operational model design. By clarifying business outcomes, adopting full-cost metrics, and establishing control layers and learning loops, enterprises can improve pilot success rates.
How CIOs can help build executive AI literacy
尽管企业设定了宏大的 AI 目标，高管 AI 素养仍严重不足。Gartner 数据显示，仅 21% 的 C 级高管认为同行“精通 AI”。本文基于 Gartner 总监分析师 Gladys Yeo 的观点，剖析高管 AI 素养的内涵，并提出 CIO 应从保护、优先、准备、定位四个维度入手，构建高管 AI 判断力，以支撑企业 AI 战略的落地与长期竞争优势。
Cost-per-token worked for AI’s first wave — but not the next
Chen Goldberg, Executive Vice President of Products and Engineering at CoreWeave, believes that token-based billing was effective in the early stages of AI, but when faced with diverse production-grade workloads, this metric cannot capture the true economics. He suggests that enterprises shift to evaluating GPU-hour utilization, fault detection speed, and elasticity costs, and make decisions through real workload testing.
How CIOs can evaluate enterprise software vendors
Most enterprise software evaluations begin with a feature checklist and ultimately yield a decision that appears confident but is inherently fragile. Sudhakar Shivaraju, Vice President of Global Real Estate Systems at JPMorgan Chase, draws on two formal selection experiences across industries and years to point out that the real risk lies not in choosing the wrong vendor, but in evaluating the wrong things, or mis-weighting the right things under real constraints. He advises CIOs to segment functional domains by actual usage scenarios (security and compliance, integration and API architecture, user experience and adoption resistance, vendor support and roadmap transparency, AI capabilities, total cost of ownership) and set weights based on their own constraints rather than generic templates. At the same time, vendor roadmaps should be treated as primary evidence rather than marketing material, distinguishing confirmed commitments from visionary statements; support responsiveness and roadmap transparency should be scored separately; AI capabilities should be examined as an independent domain, probing whether they are based on the enterprise's real data, how the privacy model works, and whether they are already in production; and switching costs such as data migration and user retraining should be honestly quantified.
The workforce advantage CIOs can’t ignore
The success of enterprise AI transformation depends not only on technology but also on the AI literacy of the workforce. A World Economic Forum report shows that by 2030, 39% of core skills will change, and nearly 60% of the global workforce will need retraining. However, current AI education models face issues such as a gap between theory and practice and a lack of personalization. This article proposes an immersive, hands-on learning framework, emphasizing changes in work methods as the measure of success, to help enterprises build an adaptive workforce.
Scaling AI calls for a revamped operating model
Early AI pilots are exciting, but transforming experiments into production environments that generate returns on investment is difficult for infrastructure and technology leaders. This article points out that enterprises need to focus on production readiness and truly scale AI applications by restructuring operational models, strengthening governance, and driving organizational change.
Why AI regulation is now an operating model
By 2026, AI regulation has moved from principle-based discussions to enforceable timelines, state-level laws, and contractual requirements. The EU AI Act is taking effect in phases, while US states and industry regulators are issuing specific rules. CIOs need to treat compliance as a design constraint and establish unified enterprise AI control systems to address lifecycle management challenges in a multi-regulatory environment.
Why CIOs must integrate governance into enterprise AI
Enterprise AI adoption is accelerating, and risks are rising accordingly. Gartner data shows that generative AI spending will increase by nearly 40% this year. Traditional governance models struggle to cope with non-deterministic AI architectures. CIOs need to adopt an architecture-first strategy, treating governance as a foundational technical requirement, and integrate six types of technical controls including guardrails, observability, traceability, centralized gateways, AI catalogs, and wrappers, while combining frameworks such as the NIST AI RMF, EU AI Act, and ISO/IEC 42001 to achieve responsible AI.
Why CIOs need to focus on AI guardrails to boost adoption
As enterprise agentic AI reaches a tipping point, more than half of technology leaders expect to remove humans from decision loops within a year. However, ungoverned AI can lead to issues such as hallucinations and inappropriate outputs. Anisha Vaswani, Chief Information and Customer Officer at Extreme Networks, writes that CIOs need to place AI guardrails on par with the technology itself, using strategies such as observability, security governance, and human-in-the-loop to ensure system safety while pursuing rapid ROI. The article emphasizes that cross-departmental collaboration and employee AI literacy are two pillars for the effective implementation of guardrails.