Making predictions for the coming year, research analyst Forrester says that the allure of quick wins and getting immediate ROI from AI has dominated a lot of corporate discussions the past months, but many are guilty of overlooking the need for a comprehensive, long-term business strategy and effective data management practices when they look at adopting AI.
What do businesses need to do to get their AI implementation done right? What are the blindspots in their AI strategies they need to address?
iTNews Asia speaks with Charlie Dai, VP, Principal Analyst at Forrester, to delve further into how companies should plan their AI strategies going into 2025.
iTNews Asia: While AI has made rapid strides this year, Forrester says there is no "easy button" for AI and advises enterprises to be prepared to put their noses to the grindstone when developing their AI strategy. What gaps have companies overlooked when implementing AI?
Dai: The key gaps can be seen in several areas:
1. Strategic alignment: Many companies fail to establish a clear, strategic vision for their AI initiatives. This lack of alignment can lead to disjointed efforts, wasted resources, and a lack of focus on achieving specific business goals.
2. Data quality and governance: Poor-quality data is a major obstacle to successful AI implementation. Companies often overlook the importance of data quality and governance, which can lead to inaccurate or biased results.
3. Talent and skills: The lack of skilled AI professionals is a significant challenge for many organisations. Companies often struggle to find individuals with the necessary technical expertise to develop and maintain their AI systems.
4. Governance and risk management: As AI systems become more complex and interconnected, the risks associated with their use also increase. Companies often overlook the need for robust governance and risk management frameworks that can help them identify, assess, and mitigate potential risks.
5. Experimentation and learning: Companies often overlook the importance of experimentation and learning in their AI initiatives, which can limit their ability to innovate and adapt to changing market conditions.
6. Integration with existing systems: Many companies struggle to integrate their new AI systems with their existing business processes and technologies. This lack of integration can lead to inefficiencies, delays, and other problems that can hinder the success of their AI initiatives.
7. Scalability: As companies scale up their AI initiatives, they often struggle to ensure that their systems can handle increased demand without compromising performance or reliability. This lack of scalability can limit the ability of companies to achieve their business goals and can lead to costly downtime or other problems that can harm their reputation or bottom line.
iTNews Asia: What can businesses do to address the gaps you described above?
Dai: They have to develop a comprehensive AI strategy that aligns with their overall business objectives and is supported by senior leadership. Often this includes
- Investing in data management tools and processes that ensure the quality, accuracy, and security of their data
- Investing in training and development programs that help their employees acquire the right skills
- Developing comprehensive risk management plans that include clear guidelines for managing potential risks associated with AI
- Creating a culture of experimentation that encourages employees to take risks and learn from their mistakes
- Investing in integration tools and processes that help them integrate their new AI systems with their existing technologies
- Investing in scalable infrastructure solutions that can handle increased demand without compromising performance or reliability
iTNews Asia: Have businesses been too impatient in wanting to see quick results from their AI investments? How much time should we give when trialling, developing and implementing AI before seeing a positive business return?
Dai: Yes – in 2024, many organisations expected immediate ROI on their AI projects, and this had led to a premature scaling back of investments. This impatience can be detrimental to the long-term success of AI initiatives, as it may result in underinvestment in necessary infrastructure and resources, as well as a lack of focus on building a solid strategy aligned with business goals.
In 2025, it's crucial for enterprises to be deliberate and pragmatic in their approach to AI. This means focusing on building a strong foundation for AI, investing in the right infrastructure and resources, and prioritising projects that will deliver tangible business value over time.
By taking a more measured and strategic approach to AI, enterprises can ensure that they are setting themselves up for long-term success in this rapidly evolving field.
The time it takes for an enterprise to see a ROI from AI implementation can vary significantly depending on various business, technology, and organisational factors, such as the purpose of AI implementation, maturity of AI technology, quality of data, level of integration between the AI system and other systems, and data and AI literacy based on the organisational culture.
- Charlie Dai, VP, Principal Analyst, Forrester
iTNews Asia: If businesses take a measured and pragmatic approach, what will be key to their organisational success? How can they make sure AI can keep pace with innovation?
Dai: To keep pace with innovation, a business-aligned data and AI strategy must incorporate inputs like guiding principles, an operating model, an approach to talent, and frameworks for governance and technology to set priorities and manage stakeholder expectations.
Their AI platforms must support operational, transactional, and analytical systems for the smooth deployment of next-generation applications, and their data must have appropriate controls for access, privacy, security, and regulatory compliance.
They must also have a mechanism to prioritise, evaluate, assess, and adapt to the pace of innovation in the AI techniques and other emerging technologies. There is also a need to build an adaptive data organisation and culture. The organisation must foster an environment of curiosity and creative thinking while developing a culture of trust that supports AI adoption.
iTNews Asia: In the US market, you have cited that hallucinations, finding quality training and challenges around governance and data protection as their largest concerns when adopting AI. Do you feel the same issues prevail for enterprises in APAC? Why or why not?
Dai: Hallucination refers to the phenomenon where a foundation model generates incorrect, non-sensical, not real, or fabricated information lacking provenance in response to a query or input.
APAC companies have similar concerns on model hallucination, quality training, and challenges around governance and data protection. AI’s readiness in data management is a major gap. The architecture nature of foundation models, which are pre-trained with publicly available internet and synthetic data, often reflects biases, outdated and misinformation common on the internet. This creates hallucination by default.
Together with the lack of high-quality data and the gaps in modern data management, this will lead to inaccurate results and negatively impact the decision-making process for businesses. In addition, as AI adoption increases, so does the need for robust governance frameworks to ensure that AI systems are used responsibly and ethically.
Companies must also address data protection issues, such as ensuring that customer data is protected and used in accordance with applicable laws and regulations.
On the other hand, APAC firms are also facing unique challenges that are not necessarily present in the US market. For instance, linguistic complexity is a key barrier for regional adoption in APAC, APAC businesses are also concerned about data localisation requirements and the potential impact of new regulations on their AI initiatives.
The lack of skilled talent in AI and data science in many APAC countries can hinder their ability to adopt and implement AI technologies effectively. Unlike the US, we also need to understand there are cultural differences and varying levels of technological maturity across different countries in APAC. These can influence the types of concerns that enterprises have when adopting AI.
iTNews Asia: You have observed that many enterprises to date have hit transient roadblocks in applying GenAI in ways that meet expectations, and predicts they will double down on predictive AI applications in 2025. What is driving this swing back towards predictive AI and why?
Dai: Firstly, Generative AI (Gen AI) has faced challenges such as AI readiness of data, complexity of RAG, and the gap in ROI expectation. These challenges have led many enterprises to be more deliberate in genAI adoption.
Secondly, leading firms in major industries have been working on predictive (discriminative) AI for over a decade using traditional architecture and small models, which has shown more reliable and consistent results compared to genAI.
Predictive use cases like predictive maintenance, customer personalisation, supply chain optimisation, and demand forecasting have proven their business value. As a result, enterprises are doubling down on foundation models to modernise predictive AI applications by progressively replacing and/or supplementing traditional ones, driving more business outcome with a unified approach across hybrid architecture.
Lastly, the integration of predictive AI with genAI can enhance the overall performance of AI systems. By leveraging both predictive and generative AI technologies, such as the adoption of both foundation models and non-foundation models, as well as the introduction of AI agents powered by genAI to automate predictive business cases, enterprises can create more accurate and efficient solutions that address complex business problems, driving the growth of predictive AI applications in 2025 and beyond.
