As artificial intelligence embeds itself deeper into investment research workflows, senior investment professionals and technology leaders face increasingly consequential decisions about where and how to deploy it. AlphaSense’s Owen Atkins, Head of Financial Services APAC, recently moderated a panel discussion to explore the decision process from the executive point of view.
The expert panel included:
- Ivan Lew, Executive Chairman at Shaw Investment APAC Pte Ltd
- Nirman Chhabra, Chief Investment Officer at Abiman Investment Singapore
- Chao Jen Chen, Head of AI, Managing Director at Fullerton Fund Management
Below, we recap the key takeaways from that discussion, exploring the practical realities of AI adoption for investment research, the importance of human judgment, and what the move toward agentic AI workflows means for investment teams.
AI Advances From Summaries to “Super Outliers”
The expert panel agreed that asset management has moved well past using AI to produce research digests. In fact, firms are now deploying specialized agents to scan hundreds of sell-side reports simultaneously. Rather than simply capturing the market consensus, these reports help portfolio managers uncover and exploit unique positions.
Chao Jen Chen explained that the real institutional value lies in isolating these opportunities, what he called “super outliers”: deeply contrarian viewpoints that leading portfolio managers use to challenge their internal house views.
AI is also transforming investment operations across the industry. Chen also shared that AI agents can automatically check investment eligibility of new credit issuance by comparing their term sheets against dozens of portfolio mandates, compressing work that once took hours into minutes. Ivan Lew described how his firm uses AI-controlled autonomous vehicle fleets to make port operations safer, freeing up considerable manpower. At the same time, Lew cautioned that “AI is as smart as the person that put the information in it.”
Where AI Alone Still Falls Short
Lew was not alone: All three panelists cautioned against blind AI implementation. Nirman Chhabra said it’s important to keep in mind that “AI is trained on the past and investment is about the future.” Chen illustrated his caution about AI’s output quality and consistency in an example: When a credit analyst on his investment team tried using a large-language model to place a regression line on a relative value scatter plot, the LLM placed the line entirely outside the data cluster, which is obviously wrong. Chen also noted that a LLM could produce wildly different results on the same dataset from one day to the next.
At the core of the issue is that generic AI tools lack the context to perform high-value-added tasks. The result is an output that may appear sound on the surface but often has severe structural issues that junior analysts could easily miss, Chen said. Lew pointed out a similar problem in physical diagnostics: AI can read and report power plant operational conditions in real time but is unable to identify the actual root cause of a fault. In root cause analysis, AI cannot replace the knowledge of an engineer with decades of hands-on experience.
AI can analyze data at scale, but the most valuable datasets are often proprietary and cannot be openly shared with competitors or the public. More importantly, AI cannot be held accountable. When investors, regulators, or boards ask for exact numbers and explanations, it is still the responsibility of senior leaders to provide the answers and own the outcomes.
Governance Risks Are Mounting
When foundation models became easy enough for individual analysts to build on, firms quickly discovered they had an inventory problem. Chen said that his investment team has accumulated over 100 AI agents, built by the analysts themselves. Left unmanaged, those tools represented a serious potential risk: data leakage, security vulnerabilities, and inconsistent outputs.
Chen shared that the solution to the agent explosion was not to ban these tools, but to formalize the path to production. The team is building a curated agent library that mandates multiple layers of review for each AI agent spanning IT security audits, model risk reviews for quantitative soundness, and dataset standardization.
AI as Human Amplifier, Not Replacement
All three panelists agreed that AI’s best role is that of an assistant, with a human retaining ultimate decision-making and legal responsibility. Chhabra said that portfolio managers should treat AI as a junior analyst: A PM would never publish a junior analyst’s write-up without reviewing it, nor should they trust an AI-generated write-up blindly.
The corollary is that the value of senior judgment is rising, not falling. Chen reinforced the point by mentioning that senior-level analysts consistently extract richer outputs from AI than their more junior peers. AI is amplifying their existing advantage, not creating one out of thin air.
I think senior leader insight is [still] very important and I don’t think AI can replace senior leaders anytime soon.
AI and Alpha Generation: Differentiation Still Key
What does the proliferation of AI across asset management mean for alpha generation? Chen thinks widespread AI adoption raises the industry's operational floor: If every firm has access to the same models, the baseline quality of research improves across the board.
But AI does not erase manager differentiation, he argued — if anything, it increases it. The pace of AI adoption still differs vastly across the industry: While leading asset managers have already built mature, curated AI agent libraries, others are still debating which platform to use. As a result, Chen said that “the [alpha] ceiling for every company is still different [and] will be determined by how well you adopt AI.”
Looking Ahead: The Human-in-the-Loop Approach
Ultimately, AI is seen as expanding human capability rather than replacing it. The panelists dispelled the notion of an AI-driven job apocalypse, agreeing that a human-in-the-loop approach appears most likely.
Chhabra pointed out a virtuous circle by which advancing technological capabilities are driving down costs, further encouraging usage and adoption to everyone’s advantage:
As the cost of each technology falls, the usage of that technology also increases exponentially. And therefore as the cost of compute is falling, people are using more and more compute, as is evidenced by the huge token usage, which is growing exponentially.
I have no doubt that there will be the same technological adoption curve [for AI]. And therefore I don't think it's going to be a case of replacing humans. It will help humans [as have] all the game-changing technologies of the last 150 years since the Industrial Revolution.
Maximize Your Advantage With AlphaSense
AI is redefining how asset managers deploy capital. AlphaSense puts you in the driver’s seat: We equip management teams and analysts with the content and AI tools they need to better identify, evaluate, and act on opportunities.
Discover how AlphaSense can transform your firm’s research workflows. Start your free trial today.
