Embracing AI in Southeast Asia: The Strategy for Avoiding Cost Overruns

news

Embracing AI in Southeast Asia: The Strategy for Avoiding Cost Overruns

Author: Yash Thakker, Associate Director – Cloud Consulting, APAC, Searce

As artificial intelligence (AI) continues to revolutionise industries worldwide, Southeast Asia (SEA) finds itself at a critical juncture. A recent study by Deloitte states that developing economies in the APAC region are actively implementing generative AI at a faster pace, with 30% higher share of gen AI users embracing AI with more enthusiasm compared to developed nations with more digitally native workers. The report cautions that businesses that fail to adopt AI will feel the impact; CEOs and senior leaders should not only focus on integrating generative AI to enhance efficiency but also reconsider their processes to adapt to the AI surge and avoid disruption.

Interestingly, despite the increased usage of generative AI by employees, businesses may not be maximising the full benefits of their investments in AI. In fact, only half of surveyed employees felt they were fully utilising the potential of generative AI. Out of Searce’s 200+ customers in South East Asia, only 5% of them have Gen AI use-cases deployed in production.

With companies jumping onto the AI bandwagon, it’s imperative that they are clear on their strategies for AI, maximising their investments to achieve business results and to generate revenue.

We’ve observed four categories of companies on the AI journey:

AI Explorers: companies exploring initial use cases with varying degrees of data readiness to harness the power of AI and machine learning (ML)
AI Augmented: companies seeking to drive operational efficiencies, using AI & ML to support their operations, with AI secondary to go-to-market (GTM) strategies
AI Powered: companies utilising AI for competitive advantage, using large language models (LLM) to power their primary GTM offerings with products/services seamlessly embedded with AI AI Disrupters – companies creating new markets, producing LLMs or core AI products for external parties to utilise We observe that many enterprises fall under the AI Explorer category while most of the investment is going to AI Disrupter organisations. This has created an immediate urgency for organisations to adapt to products/services built by AI Disrupters but the lack of clarity on impactful use-cases is missing.

Adoption Framework

To drive the strategic adoption of AI amongst our clients, we have been deploying the above framework to drive the strategic adoption of AI in businesses. The journey begins with the Discovery phase, where organisations define use cases, conduct design thinking workshops, and create innovation prototypes. This lays the groundwork for understanding how AI can address specific business needs. The second phase focuses on establishing a solid data foundation and building a compelling business case. This involves checking and upgrading the data infrastructure, calculating ROI, defining expected outcomes, and establishing AI foundations. The third phase is about execution, with MVP launches, A/B testing, and business case validation. Finally, the framework culminates in scaling and optimising AI solutions, building machine learning operations (MLOps) end-to-end, and scaling business cases. By following this structured approach, companies can mitigate risks, align AI initiatives with business objectives, and avoid costly missteps in their AI adoption journey.

Cost Levers during Adoption

The adoption of AI and machine learning technologies involves several key cost levers that organisations must consider for effective budgeting and implementation.

Technology Costs encompass the core infrastructure needed to run AI systems. This includes expenses for LLMs, CPUs, GPUs, and SaaS subscriptions.

Additionally, organisations need to factor in costs for API gateways, data acquisition, storage, and processing. The implementation of testing frameworks and MLOps pipelines also falls under this category. Security and Compliance form another crucial cost centre, covering data privacy measures, regulatory approvals, and potential issues such as legal pushback and litigation. This area also includes the development of ethical and explainable AI systems, which is becoming increasingly important. Lastly, organizational costs involve addressing the skills gap through training, managing change within the company, facilitating cultural shifts, and adapting business models and GTM strategies. By understanding and planning for these diverse cost levers, organisations can create a comprehensive framework for assessing and managing the financial implications of AI adoption, ensuring a more strategic and cost-effective implementation.

Measuring the ROI

The framework presented in this image outlines a systematic approach to calculating the ROI of AI adoption, which can be crucial for understanding and managing the costs associated with implementing AI solutions.

The journey begins with reimagining existing processes through an AI lens, identifying key impact KPIs and their current costs, and mapping out potential business cases for the project. This lays the foundation for a comprehensive understanding of where AI can add value. The identified KPIs and costs associated will anchor the overall ROI calculations.

The second step involves selecting appropriate technologies, and benchmarking KPIs with AI to create best and worst-case scenarios, which may involve proofs of concept, pilots, or high-impact, low-effort (HumanInTheLoop) initiatives. Partnering with the right set of consulting organizations and using the right technology vendor has a significant impact.

The third step focuses on mapping costs across all three cost levers, identifying both one-time and recurring expenses, and validating the business case by mapping these costs to an ROI model benchmarked against the identified KPI costs.

The final step is about execution with tight governance, continuous monitoring of benchmarked KPIs, and leveraging MLOps to improve metrics continually. This structured approach ensures that organisations not only understand the full spectrum of costs associated with AI adoption but also have a clear path to measuring and optimising their return on investment.

Overall, the AI adoption should ideally lead to lower costs of operations/increased revenue or an innovation that supports growth metrics for the organisation.

Managing your AI Deployments

AI adoption involves more than just initial implementation – ongoing management is crucial for maintaining performance and value. AI models, once trained, remain static, but they operate in a dynamic world where reality continuously diverges from the initial training data. This mismatch leads to a gradual decrease in model accuracy over time. To counter this, periodic retraining becomes necessary to maintain performance levels.

Effective AI management requires continuous monitoring and reinforcement. Organisations need to regularly assess model outputs, collect new data, and prepare it for model updates. This ongoing process is essential for keeping AI solutions relevant and accurate. These guidelines emphasise the importance of viewing AI adoption as a continuous journey rather than a one-time deployment. By recognising the need for persistent maintenance and adaptation, companies can better prepare for the long-term commitment required to maximise the benefits of their AI investments.

The Path Forward

For SEA businesses, the path forward involves embracing AI not just as a tool, but as a strategic asset. This requires a commitment to continuous learning and adaptation. Companies must invest in training their workforce, ensuring they have the skills and knowledge to harness AI’s full potential. This investment pays off in the form of increased efficiency, innovation, and competitive advantage.

By partnering with AI experts and leveraging platforms like Google Cloud, which is crucial, businesses can access cutting-edge technology and insights that will help them future-proof their businesses, enabling different functionalities to be worked into existing infrastructure as technology and business needs evolve. Working with the right partners can provide the technical support, workforce training and resources needed to build robust AI capabilities, ensuring that companies can adapt and thrive in an increasingly AI-driven world.