According to AWS research, generative AI is unlike any previous technological change and is fundamentally reinventing the way businesses operate at breathtaking speed.
In Vietnam, for example, manufacturing output is increasing by nearly 10% annually, thanks to decades of digital transformation and industrial automation. AI has achieved similar changes in a matter of months. 61% of Vietnamese companies that have already implemented AI report an average revenue increase of 16%, and 58% expect cost savings of around 20%.
However, despite billions of dollars of investment, most organizations still struggle to move from pilot to production to deployment. In fact, Gartner research shows that in 2024, 60% of GenAI POCs will be abandoned upon completion.
The difference between an AI experiment and success is not choosing the right large-scale language model. It’s more than that.
“Through our work with partners and customers at various stages of their AI journey, we have observed consistent patterns that separate successful implementations from stalled implementations,” said Kirsten Gilbertson, APJ Head, SAP GTM and ASEAN Partner Organization Leader, AWS.
Therefore, Gilbertson noted that organizations that are successful in moving from pilot to production focus on four interrelated pillars and, importantly, recognize that technology is only one of them.
The first is to strategically build a data foundation. Just having data is not enough. How you organize, manage, and activate your data makes a big difference. Leading organizations have implemented three specific practices: It’s about combining all your data, labeling and organizing it so it’s easy to find, and putting controls in place to ensure only the right people (or agents) have access to sensitive data sets.
Highly regulated industries such as financial services and healthcare have the advantage of existing governance frameworks that can accelerate AI efforts. “However, if you are an organization starting from scratch, start by working backwards from your specific use case, rather than trying to integrate your entire data warehouse,” says Gilbertson.
For example, carriers may start by linking network performance data with customer service tickets and billing records for one purpose: predicting service degradation before customers experience problems. Once that use case provides value, you can decide which additional data connections are most important and expand from there.
The second is to build trust through security and verification. “In enterprise AI, trust is not just a nice-to-have; it is the foundation that determines whether an investment moves from pilot to production,” Gilbertson said. “Organizations face a dual challenge: They need AI systems that are secure enough to protect sensitive data, yet accurate enough to make critical decisions.”
For example, consider a healthcare provider with more than 700,000 members. Customers call us in their most vulnerable moments and need medical advice or insurance information. The opportunities that AI offers are huge, allowing us to support our customers faster, in any language, 24/7. But in this situation, a single illusion can cause real harm and destroy the trust that has taken years to build.
Leading organizations are moving beyond “trust and verify” to “verify, then trust.” They implement multiple layers of validation, including checking input for malicious content, validating output against known facts and policies, and continuously monitoring for drift or unexpected behavior. Emerging technologies such as automated reasoning, a mathematical approach used for decades in chip design and security verification, allow AI output to be checked against defined rules, reducing hallucinations by 99% in some cases. This validation-first approach accelerates innovation rather than slowing it down, allowing teams to experiment more boldly when they know guardrails will catch errors before they reach customers.
Third, we need to transform not just our technology, but our culture. The biggest impediment to AI adoption is not technology, but change management. Organizations are built around complex processes and have employees who manage those processes. It takes an intentional culture change to get individuals to step back and rethink how these processes can be automated end-to-end or handled by agents.
Success requires both top-down efforts and bottom-up implementation. Leaders must go beyond words and demonstrate visible commitment, while employees need space and support to rethink their workflows.
Vietnam Technology and Commercial Corporation Bank (Techcombank) is an example of this approach. Rather than simply deploying generative AI tools, the bank built a complete enablement strategy around them. We started with a pilot of 50 developers using Amazon Q Developer, achieved 80% team satisfaction, and then rapidly expanded to 600 IT developers with 100% active engagement. This expansion has increased the team’s production by 40% quarter over quarter.
Using this GenAI-powered assistant, the bank also accelerated the development of its flagship digital banking application, Techcombank Mobile, significantly reducing development time. Over 70% of developers report saving 5-10 hours each week. These efficiency gains will enable Techcombank to offer an enhanced digital banking experience to its 16.5 million customers across Vietnam.
“AI automates many tasks while also creating new opportunities and enhancing the human potential of others,” Gilbertson said. “The most successful organizations are transparent about this transformation and are investing in reskilling their workforce to succeed in an AI-enhanced environment.”
The final step is to work with the right experts. “While some organizations have the resources and expertise to build generative AI capabilities entirely in-house, most are finding that strategic partnerships can accelerate the transition from pilot to production,” Gilbertson said. “The question is not whether you can go it alone; the question is whether it is the fastest path to realizing value.”
The right partner brings three important benefits. The technical expertise to navigate the rapidly evolving AI landscape, the domain knowledge to apply AI to specific industry and regulatory environments, and the change management experience to drive large-scale adoption.
The data supports this. Organizations working with partners with deep AI expertise and proven customer success moved AI projects into production on average 25% faster than organizations working without a specialized partner. In a context where speed to value often determines competitive advantage, that acceleration can be decisive.
In Vietnam, AWS partners offer ready-made solutions to help accelerate GenAI adoption in areas such as Intelligent Operations Automation (AIOps), document processing and analytics, and AI-powered customer engagement.
“Successful organizations typically approach generative AI as a business transformation, not just a technology implementation,” Gilbertson asserted. “Organizations that thrive are not those with cutting-edge models, but those that recognize AI success require equal investment in technology, people, and processes.”
