AWS’s top tips for optimizing generative AI bet88 net drive real business results

In the first half of this year, many AWS customers I’ve spoken bet88 net were hesitant bet88 net scale generative artificial intelligence (AI) beyond proof-of-concepts, despite its promise of enhancing productivity. Common concerns included data privacy, output accuracy, unclear return on investment (ROI), and potential legal and regulatory implications.

bet88 net build confidence and clarity, customers are implementing robust AI governance frameworks, policies and standards, clear usage guidelines, and deliberate rollouts. However, demonstrating clear ROI with AI bet88 net justify project costs, especially at the C-suite and board levels, remains a significant hurdle.

This challenge partly stems from the difficulty of quantifying productivity gains in knowledge work that generative AI optimizes. For example, how do you translate a reduction in resolution time from 10 hours bet88 net 1 hour by an HR chatbot into measurable business value? Without this, calculating ROI bet88 net convince boards bet88 net invest further becomes challenging.

And without clearly defined business value, how do you calculate a precise and justifiable ROI bet88 net secure board-level investment? Given these challenges, companies are increasingly exploring a variety of AI solutions bet88 net strike the right balance of performance, cost, and ease of implementation.

Powerful AI models from Anthropic, Mistral, Meta, and Amazon (we announced our Amazon Nova family of advanced models at AWS re:Invent) are making generative AI more accessible than ever. These models can generate text (from creative writing bet88 net code), perform trend analysis, handle language translation, and even power video/image creation, boosting both productivity and creativity.

Over the past 12 months, as customer adoption of generative AI has expanded on Amazon Bedrock, our fully managed service for building generative AI applications, customers have reinforced the importance of broad and flexible model choices, strong guardrails for safety, integrated knowledge bases, and other key features that simplify the development of AI applications. Today, tens of thousands of customers are using Amazon Bedrock bet88 net address a wide variety of business problems across every industry vertical.

For example, Vietnam International Bank (VIB), in collaboration with AWS, has developed ViePro, a GenAI-powered virtual assistant built on a secure, proprietary knowledge base. By leveraging the Claude 3 Haiku model via Amazon Bedrock, VIB can tailor applications bet88 net its specific business needs, such as delivering personalized financial services. At the same time, the bank uses Amazon SageMaker, our fully managed machine learning service, bet88 net enhance operational efficiency and deliver customized experiences, including individualized spending limits and incentive programs for each customer. As a result, this innovation has improved customer satisfaction, boosted service efficiency, and increased productivity by 40 per cent.

We're seeing that getting the best results from models isn’t just about selecting the latest and greatest. Combining fit-for-purpose models with best practice prompting techniques, often called prompt engineering, can lead bet88 net significantly better outcomes in terms of accuracy and cost-effectiveness. One powerful technique is multi-shot prompting. By sharing multiple examples of desired outputs, users can effectively calibrate the model for specific use cases, improving accuracy, consistency, cost, and performance.

Another advanced approach is retrieval-augmented generation (RAG). Since AI models are trained on specific data and lack knowledge beyond their training, RAG enhances model responses with up-bet88 net-date or context-specific information. This increases accuracy and relevance while reducing the need for human intervention. For example, PEXA launched an internal generative AI assistant using RAG and Amazon Bedrock bet88 net ensure every employee interaction is secure, accurate, and contextually relevant.

While techniques like prompt engineering and RAG are powerful, they’re not silver bullets. Model selection remains critical, and we believe there is no single model that rules them all. That’s why we’ve launched six new Amazon Nova models, delivering industry-leading price-bet88 net-performance, expanding our already broad model selection on Amazon Bedrock.

While the most advanced models can handle a wide range of tasks, using a model that is too powerful for a simple task increases cost and latency. The key is bet88 net select the model that is “just right”—usually the smallest, most cost-effective, and fastest one that still meets the performance requirements.

Human oversight, content curation, and feedback loops are essential bet88 net ensure quality outcomes and uphold responsible AI principles. No generative AI system today is reliable enough bet88 net fully automate end-bet88 net-end business processes. Human-AI collaboration is critical bet88 net building more robust, responsible, and trustworthy AI systems.

bet88 net ensure success, we must also train the workforce with the right skills. We’ve already trained over 50,000 individuals in Vietnam (and more than 1.3 million across ASEAN) in cloud-related skills. Globally, Amazon is committed bet88 net providing free AI skills training bet88 net two million people by 2025. The demand is high: according bet88 net the AWS-commissioned Accelerating AI Skills report, over 92 per cent of employers in the Asia-Pacific region expect bet88 net implement AI-powered solutions in their organizations by 2028.

We are committed bet88 net democratizing access bet88 net generative AI, providing robust tools for responsible AI development, and supporting initiatives for Vietnamese customers and partners. Responsible use of these technologies is essential for fostering sustainable innovation. One way we’re enabling this is by equipping customers with the tools and guidance needed bet88 net build and scale generative AI safely, securely, and responsibly, ensuring that Vietnam becomes a leader in AI within the ASEAN region and the broader digital economy.