The landscape for AI adoption continues to evolve. Two years after GenAI was introduced and the excitement it created, the focus is now moving toward a serious emphasis on execution and results.
As AI adoption continues to increase, businesses are also beginning to see interesting nuances and patterns emerge, with varying maturity levels across countries and regions.
Describing AI’s evolution and speaking exclusively to iTNews Asia in the first of a two-part interview, Romain de Laubier, Managing Director and Senior Partner, Asia-Pacific Chair, BCG X, says, “AI been around for nearly 10 years and gone through several cycles. Before OpenAI and ChatGPT, we were starting to see a bit of an AI fatigue – companies tried, but the value was not clear, and for many, it was a hard transformation.”
“GenAI has given the technology a second wind and reinvented AI. It has made AI accessible and easy to understand. People understood intuitively what AI can do.”
de Laubier says the Asia-Pacific region today holds the lead in AI adoption globally, particularly in predictive or traditional AI. However, the region lags behind in GenAI adoption, with many companies preferring to wait to be able to better discern the value of GenAI and its use cases before deciding to go ahead.
While AI adoption is primarily led by the private sector, in the public sector Japan and Singapore stand out globally as among two of the top three governments in the world leading AI spending. Singapore has always been driven to stay ahead in the game as a digital nation, de Laubier explains, while Japan is now playing catch up after being late in digitisation previously.
In the Asia Pacific, 29 percent of companies have trained over a quarter of their workforce on AI and GenAI, which leads the world in workforce training and upskilling. Singapore and Japan, in particular, stand tall in AI upskilling and is outpacing the other countries, with 44 percent of companies in Singapore and 38 percent in Japan having trained their workforce on AI and GenAI tools.
According to de Laubier, globally and in APAC, 2025 is a pivotal year as we are now witnessing AI production moving at scale beyond proof of concepts (POCs).
“Companies are now seeing the value of AI; they’ve done the small experiments and now want to scale to the full organisation. They are taking a deep look into all their business processes and seeing where they can embed AI everywhere. That is where the future is – how you take a function and reinvent it using AI and GenAI.”
The next few years are going to be about proving the value exists
A recent AI Radar global study from BCG found that while three-quarters of CEOs and C-suite executives believe there is value from AI and GenAI, there remains a gap between ambition and reality as only a quarter are seeing any meaningful value from their existing AI initiatives.
The task ahead for CEOs and C-suite executives is going to be about proving AI’s value and overcoming the difficulties of organisational change, including the resistance from employees on their jobs, says de Laubier.
“The next one to three years are going to be about proving that the value actually exists, and (business leaders) need to go through the hard reps of changing their organisations and ways of working. It’s not going to be a walk in the park, it’s still going to be tough,”
“People are also worried about the impact of GenAI and what it's going to mean for their jobs. There's going to be resistance and there's still a lack of understanding, from a technical standpoint, on what it takes to actually scale,” adds de Laubier.
de Leubier also notes widespread adoption will require effort in changing processes, skills and ways of working.
He foresees that the increasing use of AI in the workforce will have profound implications on every business function. For example, the profile of users will change, new KPIs will be used to measure performance, skill sets will change and companies will have to evaluate how many people they need.
While many companies will be focused on execution and results in 2025, and leading companies are already scaling AI cases, bringing GenAI to full implementation across organisations remains challenging.
Despite these concerns, de Leubier says GenAI has continued to mature, and the limitations that GenAI had at the beginning – around hallucinations, cybersecurity, and data privacy – have already been fixed by the providers and vendors.
What can we learn from companies that are seeing success?
Looking back at the AI adoption journey of organisations, de Laubier says BCG has found many companies aiming too low and diluting their efforts when they put too many bets in different functions. He said future-ready companies have shown they have been successful when they focus on a targeted set of AI initiatives in core functions.
“When adopting AI, the philosophy of ‘let's try many things, and see if it sticks’ has proven to be a failed mode. When you focus on very small things, you don't move the needle.”
Citing the BCG AI Radar report, de Laubier says today, 35 percent of companies are still focusing on AI experimentation. About 49 percent of business leaders currently perceive AI as a value source, while 16 percent say AI has provided business value.
There is a need for companies to move beyond small-scale experiments to achieve significant and tangible results. “Companies that do well often focus on the big things. They don’t have 100 POCs, instead, they focus on AI in just a few areas, and they go for functional transformation. They take one or two functions (or business processes) and transform them by leveraging AI and GenAI,” he adds.
These transformed core functions are then rapidly scaled; companies are able to upskill their teams and then able to measure their operational and financial returns.
“There is enough evidence to go beyond the small approach, and the time to play around has passed. Companies need to choose their investments strategically and look at where they can get a competitive advantage,” says de Laubier
“What we've seen is that when companies have invested ahead of time, they have generated results and have ample proof points that if you do it the right way, you can deliver, and they're going to have more money to reinvest in their transformation. (By doing so) they have continued to race ahead.”
Do we need to fix our data to get AI to work?
A fallacy in AI adoption is the belief that you need to get your data right at the beginning, says de Laubier.
“You can imagine the type of information you need, but in reality, that is close to impossible. What matters is the business objective, as things will change.”
We're not in a traditional IT world anymore where you have the business writing down the requirements, IT translating them and coming back with a system in three years. When you do that, the system that finally comes out will never fit because the business requirements will have evolved multiple times.
- Romain de Laubier, Managing Director and Senior Partner; Asia-Pacific Chair, BCG X
“You can, bit by bit, transform your core IT system to make sure that it's modernised, and allow more use cases and access to data as time goes by,” he adds.
The value of a 10-20-70 principle
BCG’s AI Radar report surmised that for companies winning with AI, it is often a sociological challenge as much as a technological one. The soft stuff – reimagining workflows, upskilling talent, and driving organisational change – is turning out to be the hardest to do.
BCG’s findings show that successful companies allocate more than 80 percent of their AI investments to reshape core functions and invent new offerings, while the rest of companies put 56 percent of their AI investments on smaller-scale, productivity-focused initiatives. Leading companies can generate 2.1 times greater ROI on their AI initiatives than their peers.
“Top performing organisations adopt a 10-20-70 framework to unlock AI’s business potential, allocating 10 percent to algorithms, 20 percent to improve data and technology, and 70 percent to transforming people, processes, and culture and organisational change,” says de Laubier.
“While tech is difficult, it's measurable and visible. That makes it the easiest to grasp. You can see the improvement in your algorithm’s performance, and you can see how your tech is performing.”
What’s the most difficult part of AI transformation?
de Laubier says the ‘soft part’ is often the most difficult aspect of the transformation (organisation, process and people etc) as that is where most companies fail because it requires difficult changes at senior levels and the resilience to continue despite setbacks.
While the transition is never easy, de Laubier says successful transformation companies often show they have a clear strategic vision, strong top-down support, and are prioritising the right investments.
“It's very difficult to focus the organisation if you don’t have a clear business objective. If you’re doing it only as an enabler, it’s not going to get the right level of attention.”
“You must have the right tech stack underlining the initiative. You also need to do both your AI transformation and your tech transformation at the same time,” de Laubier adds.
On the talent gap, de Laubier says upskilling existing employees is becoming a necessity, especially for smaller organisations that cannot afford to hire many AI experts. These upskilling programs require long-term commitment, typically taking 18-36 months to yield significant results.
“You need to invest in the right set of talents and adjust your compensation strategies as trained employees become more valuable in the job market, which may be a difficult for some business leaders.”
