SINGAPORE – Quicker loan approvals and shorter customer onboarding time may become the norm as financial services firms worldwide and in Singapore turn towards artificial intelligence (AI) agents to handle more intelligent tasks.
Customer service work, such as answering questions and handling claims, as well as loan approvals, are some areas where AI agents are being tested in the finance industry, experts told The Straits Times.
While the technology is still evolving quickly, early applications are already showing potential in helping reduce routine tasks so bank employees can focus on higher-value work, they added.
Unlike traditional generative AI tools that need constant prompting, agentic AI or AI agents can make decisions, carry out complex tasks and manage workflows largely on their own.
Traditional AI, for example, might explain when you may qualify for a loan. Agentic AI, on the other hand, can now evaluate the customer, decide if the person qualifies and approve the loan within hours instead of days.
This could be significant given that manual data entry and paper-heavy processes tend to slow down loan approvals.
But banks can use AI agents to crunch the documents and do the initial risk analysis on a customer before handing the file to a human employee, said Mr Deb Deep Sengupta, area vice-president for South Asia at UiPath, a global software company that develops AI and agentic automation software.
For example, US-based Lake Michigan Credit Union used AI agents to handle data collection and file exceptions, reducing loan cycle times by 10 days, he noted. File exceptions refer to when a loan application or the file has missing, incorrect, or outdated information that prevents it from meeting standard approval guidelines.
Dr Paul Beaumont, partner and data scientist at McKinsey & Company’s AI arm QuantumBlack, said that another area is intelligent credit underwriting for mortgages, auto loans and small business loans.
In this instance, AI agents can automatically aggregate and analyse applicant data from various sources.
Dr Beaumont cited Germany’s Deutsche Bank as an example of using agentic AI to achieve faster loan approvals while enhancing risk assessments by incorporating alternative data sources.
Salesforce ASEAN’s vice-president and chief technology officer for solutions Gavin Barfield said that loan discovery, as part of the loan process, can be automated with AI agents while human loan officers focus on advising borrowers, building trusted relationships and finalising loan applications.
Loan discovery in financial services refers to the process of identifying, evaluating and applying for loan products tailored to a customer’s specific financial situation, typically through AI-powered apps or online platforms.
Customer service is another area where AI agents can make an impact.
Insurance companies, for instance, have deployed agentic AI for customer interactions, accelerating claims processing, said Amazon Web Services Singapore managing director Priscilla Chong.
She cited bolttech, a Singapore-based insurtech company, which uses agentic AI to power an advanced speech-to-speech chatbot that handles customer policy questions, processes routine claims and responds to inquiries with near-instant response times.
Insurer Singlife has also teamed up with Salesforce in October to launch an AI agent to boost customer service efficiency by providing faster and more accurate responses to queries.
This involves tapping Salesforce’s Agentforce platform to take in information from Singlife’s product manuals, training guides and other materials.
Typically, customer service executives would have to manually search through these materials to find relevant information before responding to customers’ queries.
Singlife is looking to expand the use of agentic AI to its financial adviser representatives, the firm said.
Another example is the Bank of Singapore, which launched an agentic AI tool in October to generate “source-of-wealth” reports, which detail a person’s or entity’s total assets and their origins, and gives clarity on the legitimacy of the customers’ assets.
The tool shortens the time it takes to generate such reports from the usual timeline of 10 days to as little as an hour.
As a result, the bank’s relationship managers can now spend more time engaging clients to better understand their financial needs and review their portfolio.
On the security front, AI agents allow for enhanced fraud detection and response.
“It can monitor transaction streams in real-time, identify anomalous patterns and instantly freeze compromised accounts, significantly reducing financial losses and protecting customers,” said Dr Beaumont.
One of the biggest impacts of AI agents in this area is their ability to clear hundreds of thousands of alerts in seconds, a task that would take a human analyst 30 to 90 minutes per alert, he noted.
AI agents are also proving useful in automating know-your-customer (KYC) processes and augmenting anti-money laundering processes, said Dr Beaumont.
Mr Sengupta noted that AI agents can handle the heavy lifting of client due diligence by automating identity verification, matching entity data, and collecting required documentation.
Future uses for agentic AI in finance include autonomous market analysis and trading with minimal human intervention as well as role-specific agents that can act as assistants to relationship managers and bank analysts, said experts.
“We are seeing banks develop entirely new products that don’t yet exist in the market,” Dr Beaumont added.
Even as the uses for AI agents grow, the human touch cannot be undermined. Mr Sengupta pointed out that human judgment remains critical for final decision-making.
“In practice, financial services institutions follow a model where the AI executes the groundwork, a human validates the findings, and the AI then completes the workflow,” he said.
Building rapport with customers also remains fundamentally human work, especially in areas like wealth management and financial advisory.
Ms Chong said: “Client relationships are built on trust, empathy, and deep understanding of individual circumstances – qualities that AI cannot replicate.”
Strategic decision-making under uncertainty also requires humans in the loop.
Complex, high-stakes decisions will still rest with humans, who can apply nuanced judgment and ethical considerations, even as AI provides data-driven recommendations, said Dr Beaumont.
