�۶���Ƶ

Imagine a day when your digital assistant not only schedules meetings but also proactively resolves scheduling conflicts, crafts nuanced responses to critical emails and orchestrates global research collaborations—all with minimal human input. In today’s corporate landscape, agentic AI systems are already transforming supply chain logistics by dynamically rerouting shipments in response to global disruptions, streamlining compliance in real-time financial transactions, and delivering hyper-personalized customer experiences at scale. These advanced agents are no longer just tools; they are autonomous collaborators, redefining productivity, creativity, and strategic decision-making for knowledge workers and enterprises alike.

Nicolas Gaudois, Head of APAC Tech Research, �۶���Ƶ, moderated a panel on AI at �۶���Ƶ Asian Investment Conference which covered the latest developments in field over the past year:

Agentic AI has the potential to shake up everything from automating customer service to how we get from A to B in autonomous vehicles. Agents also promise to bring about a new chapter in enterprise adoption of AI models.

There is clearly a high level of interest within the corporate sector to deploy AI. A �۶���Ƶ survey of 125 IT executives and data engineers revealed that over 60% have implemented some level of AI deployment, though only 14% of large enterprises have achieved production at scale¹ .

There are signs however, that we could be about to see greater large-scale penetration amongst corporations. Raj Sundaresan, CEO, Altimetrik, an AI solutions company shared at AIC:

Over the past year, there has been an incredible evolution among our customers, as we see the usage of Gen AI transitioning from experimental pilots and boardroom demonstrations to secure and scalable applications, with particular interest in copilots and custom agents.

Sundaresan gave a case study of a life sciences company applying predictive AI to help manage its supply chain. Employees can interact with the AI so that the necessary raw materials are always available, thereby preventing manufacturing disruptions. This resulted in a 30% increase in efficiency.

One way to conceptualize the adoption of AI agents is as a shift from “copilot” solutions guided by human inputs to “autopilots” that operate autonomously. As such, these agents will become key components of automated workflows, heralding a new wave of problem-solving capabilities that promise the autonomous completion of tasks.

��

Looking ahead, people are already asking big questions about the impact that AI agents will have on the world of business, and society at large.

One question is whether the prevalence of agents will lead to a concentration of economic power among a small group of companies, or a dispersion of power to more businesses.

The answer depends on how agents are deployed, said John Gibson, Chief Commercial Officer, Faculty AI.

One scenario envisions everyone having their own agent that is hyper-personalized to their individual needs and can independently perform a wide range of tasks – such as booking a restaurant, writing to the local government, etc. The other possibility sees fewer agents primarily operated by the suppliers of goods and services. In this scenario, an individual would interact with the agent of its restaurant or its local government.

Gibson suggests that in the latter scenario where fewer agents are distributed among a range of vendors, power will be dispersed among all these organisations, potentially having a democratizing effect. However, if individualized agents are available to every individual, these agents will likely be provided by a very small number of companies – thus concentrating power and increasing the risk of monopolies.

As the world sees the rapid evolution and uptake in AI, can supply keep up with demand? There is a vast infrastructure underpinning AI computing, as large numbers of data centers are being built across the world. The capex requirements are correspondingly large, with �۶���Ƶ forecasting more than USD400 billion in 2025 alone².

Over the last couple of years, the key concern was whether there are enough chips to equip the data centres and sufficient energy available to power them. But the easing of supply constraints will likely swing the debate among investors towards the relatively under-researched demand side of the equation: will the investment in AI infrastructure generate a satisfactory return on capital?

Gaudois said:

AI innovation is not slowing. If anything, it is accelerating, and it supports tremendous demand for semiconductors to build out the infrastructure. The key question going forward is whether there is enough demand to justify the enormous outlay.

�۶���Ƶ identified three pillars of demand for AI infrastructure and assessed the prospects for each one. The first pillar is large-language model training by major developers, which continues to absorb significant computing power, with little evidence of slowing down. The second pillar, representing consumer AI products, is also not showing any signs of being under pressure, as tech giants are successfully monetizing consumer engagement.

The third and final source of demand is the laggard, enterprise usage of AI, as many of the largest companies are still at the proof-of-concept stage of application. But when it gains momentum, it could become the strongest source of demand of them all.

Taken together, the three pillars suggest that demand will keep pace with supply for the foreseeable future. “The only risk is if demand from model training and consumer products cools down before enterprise use gains momentum,” said Gaudois. “But since we are already seeing many innovative applications of AI by major corporations, there is reason to believe that supply and demand dynamics will remain balanced over the coming years.

¹�۶���Ƶ, Q-Series, Will AI Demand Be Sufficient to Justify Investment?

²ibid