This article first appeared in The Edge Malaysia Weekly on July 6, 2026 - July 12, 2026

ARTIFICIAL intelligence (AI) has become banking’s latest competitive frontier, with institutions racing to deploy chatbots, coding assistants and productivity tools to help employees draft emails, summarise documents and automate routine work. Yet most deployments remain incremental, focused on improving existing processes, as banks cautiously navigate governance, cybersecurity and regulatory concerns.

Hong Leong Investment Bank Bhd (HLIB) is pursuing a more ambitious path. Rather than treating AI as another workplace tool, the investment bank is redesigning parts of its operating model around what it calls an “institutional brain” — a network of specialised AI agents that preserve institutional knowledge, automate highly technical work and support decision-making across the organisation.

If successful, the initiative could offer a glimpse into how investment banks may eventually operate, with human professionals working alongside digital colleagues that retain institutional memory, collaborate across functions and continuously improve through experience.

The initiative, known internally as Project Alpha 2.0, did not begin in investment banking or sales. Instead, HLIB chose to tackle one of banking’s most persistent structural challenges: risk management.

“We didn’t start using AI because it was fashionable. We started because we had a real business problem,” says group managing director and CEO Lee Jim Leng.

Unlike many corporate functions, expertise in areas such as risk management, regulatory compliance and corporate finance is built over years of experience. Recruiting specialists is difficult, while much of their knowledge exists only in individual employees rather than documented processes.

“Risk people are very technical. They’re hard to recruit and even harder to replace,” says Jim Leng.

“When staff leave, they take that knowledge away. It then takes us a long time to train somebody else to reach the same level.”

Instead of viewing AI primarily as a cost-cutting tool, HLIB saw an opportunity to solve a deeper organisational problem — preserving decades of accumulated expertise, even as employees retire or move on. The bank’s answer was to codify that expertise into AI agents capable of learning internal processes, retaining institutional knowledge and making it continuously available to future employees.

Although the initiative began in HLIB’s risk management division, the platform has been extended to its sister company, Hong Leong Asset Management Bhd (HLAM), which comes under Hong Leong Capital Bhd (KL:HLCAP).

The financial group plans to expand its use of specialised AI agents into other revenue-generating businesses, creating a shared institutional knowledge platform across the organisation.

At the heart of the strategy is a simple idea: institutional knowledge should belong to the organisation, not individual employees. Rather than relying on the staff to remember past transactions, regulatory interpretations, client engagements or internal methodologies, HLIB wants that collective experience to remain with the bank.

“When I retire, I don’t want all my knowledge to disappear. We want to build institutional knowledge that stays with the bank,” says Jim Leng.

Chief risk officer Lee Wai Sing says the platform is designed to ensure continuity instead of replacing employees. “If someone leaves halfway through a transaction, normally the next person has to spend considerable time understanding what has happened,” he points out.

“With the AI agent, you simply ask it for the history of the engagement, what discussions took place and what still needs to be done. That institutional knowledge stays with the bank,” he elaborates on the philosophy that distinguishes HLIB’s platform from consumer-facing generative AI applications.

Instead of relying mainly on publicly available information, the system operates within the bank’s secure environment, drawing on proprietary policies, workflows, historical transactions and operational knowledge accumulated over years. The result is less like interacting with a chatbot and more like working with another colleague, says Jim Leng.

“I can ask Mala a question at 2am — I can’t ask my staff at those hours,” she says, referring to the bank’s AI orchestrator. “But Mala never tells me she’s tired.”

At HLIB’s risk management division, every employee is paired with a specialised AI counterpart. The digital workforce includes credit analysts, sustainability analysts and data analysts, each responsible for a specific function alongside their human colleagues.

“Think of it like a normal organisation chart,” says Wai Sing. “I have my human team and I have a digital team. Every human works together with an AI agent and all the agents report to one lead orchestrator [Mala].”

The workflow resembles supervising a team of junior analysts. Suppose management requires a credit assessment on a listed company. Rather than assigning multiple analysts to gather financial statements, monitor market developments and assess sustainability risks, Jim Leng simply instructs Mala.

The orchestrator automatically distributes the work among specialised AI agents. One retrieves financial statements and market data through authorised sources. Another evaluates historical performance and prepares the credit assessment. If environmental exposure is relevant, a sustainability specialist assesses carbon emissions, transition risks and other climate-related metrics.

The findings are then consolidated into a single report before being returned to the orchestrator, which emails the completed analysis to management — typically within minutes instead of days.

The same approach has transformed routine risk monitoring.

Before trading begins each morning, Mala automatically instructs a digital analyst to retrieve positions from the bank’s back-office systems, calculate market and liquidity risks, and prepare management reports. “It is already waiting for me in my inbox when I arrive,” says Jim Leng.

According to them, HLIB’s AI agents collaborate not only with people, but also with each other. If one specialist lacks information, it automatically requests assistance from another AI agent. When data is only available from a human employee, it sends an email requesting the information, waits for the response and resumes the workflow before completing the task.

“It behaves very much like another colleague,” says Wai Sing. “The AI agents have memory. They know each employee’s role and even understand how different people prefer to work. The more they interact with you, the better they understand you.”

The memory extends beyond individual interactions. Rather than searching the internet, the AI primarily draws on HLIB’s internal knowledge base, including policies, standard operating procedures, historical meeting materials, analytical frameworks and proprietary risk models.

When external information is required, it is obtained through authorised application programming interfaces (APIs) and reputable news sources, with every output fully traceable.

“Every number must have a source. If the agent gives me an answer, I must be able to trace where that information came from,” says Wai Sing.

The platform’s capabilities become most apparent during periods of market volatility.

Traditionally, conducting a bankwide stress test requires multiple teams working simultaneously. Market risk specialists analyse trading exposures, treasury evaluates liquidity, credit teams assess counterparty risk, while sustainability specialists examine climate and transition risks. Consolidating those assessments into a single report can take days, if not weeks.

HLIB’s AI platform now performs much of that coordination automatically.

Instead of assigning work to different departments, Wai Sing instructs Mala to initiate a stress test. The orchestrator immediately deploys a swarm of specialised AI agents. One evaluates market risk using updated macroeconomic assumptions. Another analyses liquidity, while others assess supply chain disruptions, counterparty exposures and climate transition risks. Their findings are consolidated into a single report before being presented to management.

“If there is another war tomorrow, I can immediately ask the agents to run another stress test. What previously required an entire village can now be completed overnight,” says Wai Sing.

The example illustrates how HLIB has moved beyond using AI to automate individual tasks. Instead, it has built a coordinated network of specialist agents capable of functioning much like an experienced risk management team, allowing management to model multiple market scenarios almost on demand.

One unexpected benefit of HLIB’s AI deployment has been its ability to challenge long-standing assumptions. During one implementation, an AI agent assigned to business continuity management refused to complete its task.

The reason turned out to be surprisingly simple. It had identified an error in an Excel spreadsheet that employees had been using for years.

“The formula had been wrong all along,” Wai Sing recalls. “Humans simply continued using the same template because it had always been done that way. But the AI started from scratch, checked every formula and told us it couldn’t proceed until we fixed the error.”

For management, the episode underscored another advantage of AI. Rather than inheriting institutional habits, the technology approached every assignment with a clean slate, scrutinising processes that employees had long accepted without question. “They look at things from a completely fresh perspective,” says Jim Leng.

The incident reinforced the bank’s view that AI’s value extends beyond automating repetitive work. Equally important is its ability to identify inconsistencies, question established practices and detect risks that may escape human attention.

The productivity gains have been substantial. According to Wai Sing, the deployment across HLIB and Hong Leong Asset Management has already saved more than 10,000 man-hours, enabling the group to absorb natural staff attrition without replacing every departing employee.

More importantly, the combination of human professionals and AI agents has dramatically increased the team’s capacity. “We can perform work equivalent to a risk management team of about 200 people. Tasks that previously took one week can now be completed in about five minutes,” he says.

The additional capacity has allowed the bank to undertake work that would previously have been constrained by a lack of manpower, while enabling experienced professionals to devote more time to judgement, governance and strategic decision-making instead of routine analytical work.

Despite those gains, management is keen to dispel the notion that AI is being deployed to reduce headcount.

“Our philosophy has never been about replacing people. It is about making our people more productive and allowing them to focus on higher-value work,” says Jim Leng.

Rather than displacing professionals, the technology is intended to augment their capabilities by taking over labour-intensive processes while leaving critical judgement and client engagement firmly in human hands.

That distinction remains central to HLIB’s AI strategy. While AI agents can retrieve information, perform analysis and generate recommendations, every output must still be reviewed and approved by a human before any action is taken.

“Decision-making always rests with humans. The agents are there to assist and augment us, not replace us,” says Wai Sing.

The bank believes this human-in-the-loop approach is particularly important in financial services, where accountability, regulatory compliance and fiduciary responsibilities cannot be delegated to algorithms.

Before rolling out its first AI agents, Hong Leong Capital established an internal governance framework covering model validation, cybersecurity, oversight and responsible AI use. The framework was benchmarked against international standards while ensuring alignment with regulatory expectations.

The group has also established an internal AI Centre of Excellence, which oversees AI governance, maintains an inventory of AI agents, validates new models and helps retrain employees for emerging AI-related roles.

The governance structure reflects management’s belief that AI deployment is as much an organisational transformation as it is a technology initiative.

“We actually give our agents staff IDs. The only things they don’t receive are salaries and bonuses,” Wai Sing laughs.

Whether HLIB’s experiment results in the institution retaining Jim Leng’s knowledge and abilities long after she retires, it is clear that the bank is no longer treating AI as another software tool but a permanent digital member of the organisation that retains institutional knowledge, works around the clock and becomes more capable with every interaction.

Rather than eliminating jobs, AI adoption merely reshapes the nature of work — creating new responsibilities in areas like AI governance, model validation and risk oversight — even as it automates existing ones, HLIB’s management stresses. Still, the fact that fewer human beings are required to complete tasks points to an industrywide shake-up in the making. Watch this space.

Save by subscribing to us for your print and/or digital copy.

P/S: The Edge is also available on Apple's App Store and Android's Google Play.