Generative AI will remain a key tech trend this year, but the focus will shift to maximising ROI and driving tangible business outcomes rather than just experimentation.
“The majority of generative AI’s value and applications will continue to be realised in four key business functions: marketing and sales, product and service development, service operations and software engineering,” Sachin Chitturu, partner and Southeast Asia leader of QuantumBlack, AI by McKinsey, tells The Edge Singapore.
He continues: “Moreover, the organisations we recently surveyed say they are already pursuing new ventures enabled by generative AI across various use cases such as co-piloting, decision-making, hyper-personalisation and distribution.”
Chitturu says generative AI will increasingly be integrated with other forms of AI, such as applied AI, to drive superior outcomes. Applied AI leverages models to tackle classification, prediction and control challenges, automating, enhancing or expanding real-world business applications.
For instance, a digital marketing company might use generative AI to create content. However, the real value emerges when applied AI analyses user engagement insights and feeds them back into the generative AI model, enabling it to produce more engaging and effective content.
Advancements in generative AI will also accelerate the development of general-purpose and humanoid robots. “By incorporating AI-based large language models (LLMs) and large behaviour models, robot control systems can help the machines understand and respond to verbal input as well as mimic human movement. These characteristics make humanoid robots viable for tasks in a variety of sectors, including manufacturing, health and social care, retail, hospitality and customer service,” says Chitturu.
See also: Nvidia chips, Trump's tariffs and AI's future
Matthew Oostveen, vice president and chief technology officer for Asia Pacific and Japan at Pure Storage, also foresees generative AI being increasingly used in industrial settings or applications to improve efficiency. This calls for “new” approaches, such as large quantitative models (LQMs) based on hard scientific equations instead of web data.
“While LQMs use the same storage and GPUs as LLMs, they are trained on a different kind of data and demand deeper access to bespoke datasets. This will bring its own set of challenges, including governance — specifically around how to train the models with proprietary data that needs to be kept confidential even between departments,” says Oostveen.
Generative AI is also expected to become more autonomous or agentic. “Agentic AI can think for itself without requiring prompts from a user. While generative AI tools can reduce workloads, agentic AI offers far more to businesses in terms of efficiency and productivity,” says Nicholas Lee, chair of SGTech. For instance, AI agents can handle basic customer enquiries and the initial stages of customer engagement before handing off more complex cases or qualified leads to human workers.
See also: Foxconn’s mega-AI plant ready in a year despite Trump tariffs
Building blocks
With Gartner predicting that at least 30% of generative AI projects will be abandoned after proof of concept by the end of 2025, unlocking its potential demands careful planning.
“Organisations considering enterprise-wide generative AI adoption must first develop a clear AI strategy. They must identify business problems generative AI can solve and structure a solution that aligns with their broader company vision. [They also need to evaluate] if they have people in-house to make it work, with the necessary skills to implement and manage the generative AI solution,” says Chitturu.
Ensuring infrastructure readiness is another key imperative. Cisco’s AI Readiness Index shows that only 30% of organisations in Southeast Asia have the necessary GPUs to meet current and future AI demands, while just 39% possess the capabilities to safeguard data in AI models with end-to-end encryption, security audits, continuous monitoring and rapid threat response.
“We’re seeing companies [struggling to] tackle gaps in computing, data centre network performance and cybersecurity, amongst other areas. This is a concerning factor as companies anticipate significant increases in workloads. Companies must invest in scalable and adaptive infrastructure that can handle AI computational demands and embrace emerging technologies like cloud for efficiency and speed of AI deployments,” says Tay Bee Kheng, president of Cisco Asean.
According to Joseph Yang, general manager of high-performance computing and AI for APAC and India at HPE, an IT infrastructure optimised to support AI will also help mitigate unplanned costs caused by unexpected challenges related to operational complexities, security risks and resource inefficiency.
Organisations should, therefore,invest in robust, efficient hybrid cloud infrastructure designed for AI.
To stay ahead of the latest tech trends, click here for DigitalEdge Section
Yang says: “The world needs native AI systems that optimise everything across the AI lifecycle, regardless of whether the workload is on-premises, in a colocation facility, the public cloud, or at the edge. A flexible, efficient and secure hybrid cloud model will also enable organisations to have better data visibility, enhanced control and protection and streamlined data management across environments.”
Since the quality of a generative AI’s output depends on the data it is fed, having access to relevant and reliable data is vital. “Organisations will need to look at how they can integrate their unique, existing business data with generative AI, taking into account their business’s unique data and process context to make more informed decisions about their people and operations,” says Verena Siow, president and managing director for SAP Southeast Asia.
Responsible and secure AI
Despite its benefits, generative AI raises ethical and cybersecurity concerns. In response, several countries and regions introduced policies and frameworks last year to ensure its responsible use, including the EU AI Act and Singapore’s Model AI Governance Framework for generative AI.
“The adoption of generative AI depends on trust. Secure data handling procedures and open AI ethics frameworks should be given top priority by businesses in order to address privacy issues, foster public confidence and adhere to changing legal requirements,” says Amir Sohrabi, regional vice president for Asean-Korea and head of Digital Transformation for Emerging EMEA and Asia Pacific at SAS.
Agreeing, Siow advises organisations to partner with tech companies that “adhere to an ethical framework for the development, deployment, use and sale of AI systems and complement the rules and regulations of national and international governments.”
Generative AI could also introduce new cybersecurity vulnerabilities, including backdoors. Moreover, the same technology can be harnessed to craft more sophisticated and frequent attacks. A study revealed a 233% surge in deepfake-related tools on dark web forums between 1Q2023 and 1Q2024. These tools can generate hyper-realistic deepfakes, posing significant risks, particularly during election cycles globally.
Vinod Shankar, Accenture’s Security Lead for Southeast Asia, advises organisations to invest in secure AI solutions to strengthen resilience throughout the AI lifecycle. This involves implementing frameworks, policies and processes for secure AI practices, safeguarding access, data, models and infrastructure and building trust by designing and using red-teaming, adversarial simulations and security diagnostics to identify vulnerabilities and prevent unauthorised AI deployments, including LLMs.
He adds: “Robust response and recovery, especially for critical infrastructure like healthcare and telecommunications, requires minimising operational disruptions. Investing in cloud-based solutions can drastically reduce downtime after cyberattacks, restoring essential operations within hours instead of weeks.”
As generative AI’s energy demands rise, Yang emphasises the need to embed efficiency at every stage of its development and deployment.
He continues: “Organisations will need to scrutinise multiple dimensions of efficiency from data, software, equipment, energy and resource perspectives. By taking a systematic approach to unlocking efficiency gains across the entire AI lifecycle, organisations will be able to better reduce the environmental impact of their generative AI.”
