In 2025, Agentic AI is poised to transform the business landscape across Asia, driving unprecedented innovation and efficiency. As organisations increasingly adopt AI technologies that mimic human-like decision-making, the focus will shift from basic automation to establishing agentic capabilities.
Companies will harness AI to optimise operations, enhance customer experiences, and foster data-driven strategies. Emerging trends will include the integration of AI in traditional industries, a surge in data governance initiatives, and an emphasis on ethical AI practices.
This evolution will empower organisations to navigate complex challenges and seize new opportunities in an ever-evolving market.
Establishing Agentic capabilities
According to Don Ong, head of innovation at Advantest, establishing genuine agentic capability hinges on two critical elements: data and process. "For any company, it always starts with data. Having accurate and curated data is very important," he asserts.
The second crucial element is the process, precisely the sequence in which data is written into the database. Ong highlights the potential of robotic process automation (RPA) as the "hands and legs" of a robot, complemented by large language models (LLMs) as the "brain."
Actionable steps for CIOs and CTOs:
Data audit and cleansing: Conduct a thorough audit of existing data assets, identifying gaps and inaccuracies. Implement data cleansing processes to ensure data is accurate, consistent, and reliable.
Process mapping and optimisation: Map out key business processes, identifying automation and AI integration opportunities. Optimise these processes to ensure data is captured and written into databases consistently and accurately.
RPA implementation: Explore RPA's potential to automate repetitive tasks and streamline workflows. Integrate RPA with LLMs to create more sophisticated agentic capabilities.
Facilitating Information Workflow
Effective collaboration between human teams and AI agents is crucial for maximising the benefits of agentic AI. Ong stresses the importance of "clean, curated, vectorised data" as the foundation for successful AI implementation. He advises organisations to maintain data quality by ensuring processes accurately write data into the database.
Overcoming data silos
Many organisations, particularly in Asia, still grapple with analogue data capture methods. Ong suggests a two-pronged approach: digitisation and automation. Digitisation involves converting manual processes into digital formats, while automation streamlines these digital processes.
A report by McKinsey notes that "breaking down data silos and establishing a unified data architecture" is essential for successful AI adoption.
Actionable steps for CIOs and CTOs:
Data governance framework: Establish a robust framework defining data ownership, quality standards, and access controls.
Data integration strategy: Develop a plan for integrating data from disparate sources into a unified data warehouse or data lake.
Training and upskilling: Provide opportunities for employees to develop the skills needed to work effectively with AI agents.
Data cleaning and organisation best practices
Cleaning and organising data is a critical step in preparing for agentic AI. Ong recommends establishing relationships between data sources and relational databases, coupled with machine learning techniques such as unsupervised learning and clustering.
Leveraging machine learning
Unsupervised learning can help group data based on trends and traits, allowing humans to review and validate the groupings. However, Ong cautions that "cleaning the data and curating it is a very taxing, human habit kind of activity."
Actionable steps for CIOs and CTOs:
Data profiling: Use data profiling tools to analyse data quality and identify inconsistencies.
Data standardisation: Implement data standardisation processes to ensure data is consistent across different sources.
Human-in-the-Loop validation: Incorporate human review and validation into the data cleaning to ensure accuracy and relevance.
Measuring data quality
Measuring data quality is essential for ensuring the effectiveness of agentic AI. Ong suggests feeding data into a large or in-house language model to test its understanding and identify potential hallucinations.
Data as a textbook
He uses the medical textbook analogy, where information is presented clearly, concisely, and non-repetitively. "If the same data point appears in multiple places in your database, that's when the large language model gets confused," he explains.
Actionable steps for CIOs and CTOs:
Define data quality metrics: Define specific data quality metrics that align with the performance goals of agentic AI initiatives.
Implement data monitoring: Implement data monitoring tools to track data quality metrics over time.
Regular testing and validation: Regularly test and validate the performance of agentic AI models using different datasets to identify potential data quality issues.
Training AI for effective data utilisation
Ong recommends unsupervised learning as a popular method for training AI. This involves feeding data into a machine learning algorithm and allowing it to identify patterns and clusters without predefined outputs.
Iterative refinement
Once the data is clustered, humans can review and refine the results as needed. This iterative process ensures that the AI is trained on clean, accurate, and relevant data.
Actionable steps for CIOs and CTOs:
Experiment with different training methods: Experiment with other machine learning techniques to identify the most effective methods for training AI on specific datasets.
Incorporate human feedback: Incorporate human input into the training process to improve the accuracy and relevance of AI models.
Continuous monitoring and improvement: Continuously monitor the performance of AI models and refine the training process as needed.
A unified approach to addressing data silos
Addressing data silos requires a top-down approach, with CIOs and CDOs setting guidelines and establishing project teams to consolidate and curate data across the organisation.
No simple solution
Ong acknowledges that "there's no simple way around this. Much hard work needs to go in to make it work." Industry experts echo this sentiment, emphasising the importance of a comprehensive data strategy and a commitment to ongoing data governance.
Actionable steps for CIOs and CTOs:
Establish a data council: Establish a data council comprising representatives from different business units to oversee data governance and integration efforts.
Develop a data roadmap: Develop a data roadmap that outlines the steps required to consolidate and curate data across the organisation.
Invest in data integration tools and technologies to streamline the process of connecting and transforming data from disparate sources.
Overcoming implementation challenges
One of the biggest challenges in implementing agentic AI is redefining human roles and responsibilities.
Don Ong
"It's not easy trying to get humans or people to give up part of their job or part of their work and say this is going to be done by robots, by software." Don Ong
Redesigning job scopes
To address this challenge, organisations must redesign job scopes, retrain employees, and reallocate them to higher-value work. This requires a proactive approach to change management and a commitment to supporting employees through the transition.
Actionable steps for CIOs and CTOs:
Communicate the benefits of AI: Communicate the benefits of AI to employees, emphasising how it can enhance their productivity and create new opportunities.
Provide training and support: To help employees develop the skills needed to work effectively with AI agents.
Redesign job roles: Redesign job roles to focus on higher-value tasks that require human creativity, critical thinking, and emotional intelligence.
Preparing for Agentic AI: A three-step approach
Ong summarises the steps organisations need to take to prepare for agentic AI:
Data readiness: Getting data silos together and curated, relational, and vectorised so it's ready for GenAI.
Process optimisation: Getting processes correct and simplified to ensure data remains accurate, clean, curated, and vectorised.
AI awareness: Understanding the latest developments in AI and identifying the right "brain" to add to the "hands and legs."
Advice for non-IT professionals
For non-IT professionals, Ong advises embracing the inevitable adoption of AI in the workplace. "It's not a matter of if it will happen; it's a matter of when it's going to happen. So, my advice is if you can't beat it, join it," he says.
Upskilling and data literacy
Ong encourages non-IT professionals to upskill themselves and better understand data and processes, as these skills will be increasingly in demand.
Agentic AI presents a significant opportunity for Asian organisations to drive innovation, efficiency, and growth. By focusing on data readiness, process optimisation, and human-AI collaboration, CIOs and CTOs can navigate the challenges and unlock the full potential of this transformative technology.
Click on the PodChats player and listen to Ong describe how Agentic AI will change the game of business innovation in 2025.
What steps should organisations take to establish a true agentic capability that allows AI to learn and adapt from internal processes?
How can CIOs effectively facilitate and control the flow of information between human teams and AI agents to enhance collaboration?
What strategies can organisations employ to capture data from analogue processes, ensuring that Agentic AI has access to comprehensive information for learning?
What are the best practices for cleaning and organising data to ensure Agentic AI can deliver meaningful insights and value?
What metrics should organisations use to evaluate the quality of their data?
What methods can organisations use to train Agentic AI effectively using their data?
How can CIOs address data silos within the organisation to create a unified data infrastructure that supports Agentic AI?
Where do organisations fail in their understanding and implementation of AI?
Steps that an organisation needs to prepare for the Agentic AI.
What is your advice for the non-IT people as regards Agentic AI and its value to the employee, the business.
Allan is Group Editor-in-Chief for CXOCIETY writing for FutureIoT, FutureCIO and FutureCFO. He supports content marketing engagements for CXOCIETY clients, as well as moderates senior-level discussions and speaks at events.
Previous Roles
He served as Group Editor-in-Chief for Questex Asia concurrent to the Regional Content and Strategy Director role.
He was the Director of Technology Practice at Hill+Knowlton in Hong Kong and Director of Client Services at EBA Communications.
He also served as Marketing Director for Asia at Hitachi Data Systems and served as Country Sales Manager for HDS’ Philippines. Other sales roles include Encore Computer and First International Computer.
He was a Senior Industry Analyst at Dataquest (Gartner Group) covering IT Professional Services for Asia-Pacific.
He moved to Hong Kong as a Network Specialist and later MIS Manager at Imagineering/Tech Pacific.
He holds a Bachelor of Science in Electronics and Communications Engineering degree and is a certified PICK programmer.
