Managing gen AI risks | Deloitte Insights

As generative AI continues to gain traction in various sectors, organizations find themselves navigating a rapidly evolving landscape. This comes with numerous risks that must be carefully managed to ensure sustainable growth and compliance. Recent insights from Deloitte shed light on various marketplace risks associated with generative AI, emphasizing the importance of strategic oversight and evolving infrastructures in this digital age.

### Regulatory Uncertainties

At the forefront of concerns for organizations is the issue of regulatory compliance. According to Deloitte’s recent report on the State of Generative AI in the Enterprise, over half of the organizations surveyed cited regulatory uncertainties as their biggest concern. Regulations vary widely across different regions, dictating aspects such as data usage, security, and privacy—crucial components when navigating generative AI risks.

For example, in the United States, Executive Order 14110 mandated federal agencies to enforce reporting requirements on organizations developing advanced AI models. However, regulatory dynamics can shift rapidly. Subsequent executive orders have reversed such mandatory reporting, which creates a more complex landscape for organizations to traverse. Companies must continuously adapt to these changing regulatory environments, ensuring that they remain in compliance while also fostering innovation.

### Computing Infrastructure Risks

The scalability of generative AI demands robust computing resources, which place significant strain on existing infrastructures. As data centers increasingly rely on electricity, aging public utility grids struggle to catch up with growing demand. According to a survey by Deloitte, various power and utility industry respondents identified infrastructure limitations as a primary hurdle in supplying reliable power to data centers.

The forecast for the upcoming years indicates that electricity consumption by data centers, which currently accounts for about 6% to 8% of annual electricity generation, could surge to between 11% and 15% by 2030. This escalation poses significant logistical challenges, particularly in areas where the supply of electricity has already hit critical levels.

### Opportunities Amid Risks

While risks abound, so do opportunities. Investment firms and tech providers are increasingly focused on expanding data centers, often facing challenges such as permit acquisition and limited electricity supply. For instance, in multiple areas of the United States, including Northern Virginia and Pittsburgh, limited electricity availability has delayed new projects.

Moreover, global supply chains also play a role in determining the operational resilience of data centers. Shortages in critical components—ranging from power transformers to servers—can exacerbate delays and escalate costs. Organizations must deftly navigate these uncertainties, balancing investment decisions while ensuring operational reliability.

### Vendor Lock-In and Application Flexibility

One notable risk in the generative AI landscape is vendor lock-in. As advancements in AI models outpace organizational adaptation, businesses may find themselves tied to specific suppliers, limiting their flexibility to modify or upgrade their technological landscape. This concern is compounded by the high demand for advanced hardware, which, as seen with NVIDIA’s GPUs, can lead to stock shortages and long lead times.

Organizations focusing on a singular vendor run the risk of missing pivotal technology upgrades, a substantial downside in an industry characterized by rapid innovation. Strategic decisions made today will echo through the future landscape—organizations are encouraged to diversify their hardware sourcing and explore compatibility across platforms.

### Value Realization Concerns

A crucial consideration for organizations investing in generative AI technology is the return on investment (ROI). Deloitte’s survey revealed that approximately one-third of respondents believe failing to realize anticipated value from their generative AI endeavors could slow overall marketplace adoption. Significant upfront expenditures related to model training and necessary computing infrastructure can deter organizations from fully implementing these technologies, leading to stagnation in progress.

### Emerging Solutions to Marketplace Risks

To mitigate these risks, organizations are exploring innovative strategies. One approach involves reducing computing demand by leveraging small language models (SLMs) for specific tasks, which can be both cost-effective and energy-efficient. Companies like Salesforce have set a precedent by utilizing smaller models to balance efficiency and performance.

### Strategic Infrastructure Decisions

The current landscape calls for strategic investments in on-premises data centers that can complement cloud solutions, offering organizations greater flexibility and control over workloads. Some enterprises are investing substantial capital, with forecasts predicting that the AI data center switching equipment market could reach $1 billion by 2027.

However, challenges remain, particularly regarding energy access. Many organizations find themselves over-evaluating the cost benefits of owning infrastructure without accounting for related power consumption and cooling requirements. A hybrid model often proves more effective, with organizations outsourcing intensive computing tasks to the cloud while retaining simpler processes on-premises.

### Innovative Energy Solutions

As traditional energy grids strain under increased loads, organizations are shifting toward renewable energy sources and microgrids to fortify their operations. Flexential’s 2024 AI Infrastructure report highlights a trend of companies leaning towards third-party colocation data centers for their AI needs, reflecting a significant shift away from on-premises reliance.

Moreover, companies are exploring alternative energy sources, such as advanced nuclear technology and micro-nuclear strategies, which are becoming increasingly viable for powering data centers. These initiatives promise a greener future while addressing the pressing power demands posed by burgeoning generative AI systems.

### Building Trust and Governance

Finally, operationalizing trust in AI utilization is critical. Deloitte’s Trustworthy AI framework emphasizes the importance of establishing robust governance structures that address ethical, legal, and operational dimensions of AI deployment. Companies are encouraged to take a cohesive approach, minimizing obsolescence risks while maximizing the beneficial impacts of emerging technologies.

### Conclusion

The landscape of generative AI is rife with risks and uncertainties, but with thoughtful strategies and innovative solutions, organizations are well-equipped to navigate these challenges. By staying abreast of regulatory changes, optimizing infrastructure investments, diversifying vendor reliance, and prioritizing energy sustainability, businesses can capitalize on the immense opportunities presented by generative AI while effectively managing the associated risks. As this sector evolves, a proactive approach will be paramount in ensuring continued innovation and growth.

Source link