Turning AI ambition into sustainable value with trust

Trust is a sentiment earned through years of consistent action and performance, even in challenging circumstances. As global growth patterns evolve, trust has emerged as the key differentiator. This holds true across major technology waves, with artificial intelligence (AI) reaching a critical juncture in its development. Deloitte's recent initiatives with organizations in government and the private sector in Southeast Asia, with Singapore serving as the business and financial hub, indicate that scaling AI effectively, responsibly, and with clear business value is now the focus.

However, trust is not solely built by technology. It is established through expertise, context, and human judgment. While AI can generate outputs, identify patterns, and expedite decisions, people are ultimately responsible for understanding those outputs, employing professional judgment, and ensuring accountability for the resulting decisions.

In the past, trust in business organizations and government agencies relied on visible relationships, established reputations, clear contracts, and predictable policy frameworks. Yet, in an era where more decisions are influenced by data, models, platforms, agents, and automated workflows, these foundations alone are insufficient.

Organizations that struggle to scale AI often grapple with similar challenges: trust is lacking. To build trust, organizations must clearly delineate the boundaries between human and machine decision-making, provide transparency into how recommendations are generated, maintain robust audit trails, and ensure human oversight when risks escalate.

Current AI models are often too complex for people to comprehend every calculation behind the output. Instead, a systematic approach is needed to test, compare, and judge where AI can be relied upon, where its limitations lie, and where decisions must remain entirely human.

As AI becomes more integrated into business processes, organizations will be assessed not only on whether AI functions effectively but also on whether it operates consistently, cost-efficiently, equitably, and transparently when questioned. The effectiveness of AI systems hinges on the quality of the data they are built upon. Robust data governance and meticulous information management are thus vital strategic assets rather than merely technical considerations.

While organizations may resist waiting for flawless data, it is crucial not to overlook its limitations. Incomplete, outdated, or poorly managed data can result in unreliable outputs or AI-generated inaccuracies.

Ultimately, trust in AI stems not from flawless data or models but from individuals who recognize the limitations and employ judgment accordingly. Governance is often perceived as an obstacle to innovation, but the opposite is true. Across Southeast Asia, trust has long been a crucial catalyst for economic development. Although regional economies vary in their stages of development and regulatory maturity, those that cultivate confidence through robust institutions, clear rules, and consistent standards are often better positioned to attract investment, foster innovation, and maintain long-term growth.

Singapore's reputation as a trusted financial and business hub exemplifies this principle – it was not achieved despite stringent oversight but because of it.

As organizations deploy increasingly capable AI agents, governance frameworks must evolve beyond conventional technology controls. Boards and executive teams must have visibility into where AI is utilized, what decisions are being influenced, emerging risks, and accountability. Questions regarding data quality, the economics of tokens, model performance, cybersecurity resilience, third-party dependencies, and regulatory compliance cannot be addressed in isolation.

Trust in AI ultimately depends on leaders demonstrating that these risks are understood, managed, and subject to appropriate oversight. Governance must transcend the technology function, encompassing strategy, risk management, operations, legal, compliance, and workforce transformation. This is particularly crucial because agentic AI does not merely automate individual tasks; it increasingly operates within workflows that intersect multiple functions and systems, impacting approvals, customer interactions, operational decisions, and financial processes. As AI becomes embedded in how work is conducted, governance can no longer be an afterthought.

Historically, organizations focused on risks associated with technology failure. With AI, the focus must expand to include incorrect reasoning, bias, unintended actions, over-reliance on automated decisions, and a lack of transparency. Continuous monitoring, testing, and refinement are essential for reliable performance. Organizations must evaluate AI systems not only under ideal conditions but also in complex, real-world scenarios.

Human feedback loops and ongoing observation are critical for preserving trust over time. A key insight from practical implementations is that autonomy does not diminish the importance of human oversight.