Jakub Jirsk - Fotolia
Singlife taps Salesforce Agentforce to drive agentic AI strategy
The Singapore insurer has begun rolling out Salesforce Agentforce to assist customer service teams, with plans to extend the use of the technology to financial advisers and direct customer interactions
Singapore-based insurer Singlife has started deploying Salesforce Agentforce to assist customer service teams as part of a broader move to harness the capabilities of agentic artificial intelligence (AI) across its business.
Through the platform, the company will use AI agents to parse complex insurance product documentation and serve accurate answers to human agents in real-time. While Singlife has previously experimented with generative AI for code generation, this marks the company’s first major foray into AI agents.
Romil Sharma, group head of technology and operations at Singlife, told Computer Weekly that the decision to go with Salesforce was driven by the scalability and ease of deployment offered by a software-as-a-service (SaaS) model, compared to building custom AI agents with platform-as-a-service (PaaS) offerings from hyperscalers.
Other factors that sealed the deal with Salesforce included resiliency and cost, according to Sharma. Following a selection process in August 2025 and contractual work a month later, the system is now being used by close to 30 customer service agents. By January 2026, the firm intends to expand the deployment to 100 agents.
Singlife’s Agentforce implementation relies on Salesforce’s Data 360 data platform, formerly known as the Salesforce Data Cloud. The company has ingested some 150 documents, such as product manuals, frequently asked questions (FAQs), and training guides, into the platform.
When a customer service agent launches a query, an AI agent retrieves relevant information to provide an answer with citations, alleviating the need for the human agent to sift through different versions of product documents and other information that have been modified over time.
Sharma noted that the goal of the agentic AI initiative is to improve customer service operations rather than reduce headcount. “Our game is not to cut down the people,” he said, adding that the technology allows human agents to handle higher volumes of queries as the business grows.
Governance and the learning curve
Singlife has implemented a strict governance framework, involving its risk team to evaluate agentic AI outputs. Sharma was candid about the challenges of maintaining accuracy during the rollout, targeting an 80% accuracy rate as a benchmark for success.
“AI is always learning,” Sharma said, adding that the implementation team noticed a fluctuation in accuracy as the user base expanded to a bigger pool of customer service agents with a greater variety of prompts and questions.
“When I did the initial verification, the accuracy was 80%. When we went with five human agents, the accuracy went down to about 60% to 70%,” he said. However, through context handling and human-in-the-loop feedback mechanisms – such as thumbs-up and thumbs-down ratings – the AI agents quickly learnt and returned to higher accuracy levels.
While more organisations are dipping their toes into agentic AI, most have yet to define a formal lifecycle management framework for AI agents – mirroring the way organisations manage the lifecycle of human employees and software applications.
Singlife is no exception. When asked how the insurer manages the full lifecycle of an AI agent, from onboarding and continuous training to decommissioning, Sharma admitted that this remains a work in progress.
“We’re not there yet, but that’s something we’re discussing already,” Sharma said. “Salesforce is also putting a governance framework for us, and in that, they’ve also recommended some lifecycle management practices and other things.”
For now, the company is limiting the use of AI agents to less risky questions, with plans to allow for more complex interactions as confidence in the system grows. It also intends to deploy AI agents for its network of financial advisers who require access to timely and accurate product information so they can address customer needs with clarity and confidence, Sharma said.
At some point, Singlife will also look at allowing customers to interact directly with AI agents via its website, reducing the need for them to contact customer service teams. However, Sharma stressed that this would only occur after the technology has been thoroughly proven with internal staff.
Beyond Salesforce, Singlife maintains a multi-cloud strategy, with Oracle, Microsoft Azure, and Amazon Web Services among its list of suppliers. The company has also been testing Amazon Bedrock and IBM watsonx for other agentic AI use cases, including underwriting and code generation.
With the growing use of AI agents, which is expected to drive more organisations to recalibrate their workforces, Sharma said Singlife is looking to reskill its people and get them “acquainted with all these changes so that they can remain relevant in the market.”
Read more about AI in ASEAN
- Malaysia’s Ryt Bank is using its own LLM and agentic AI framework to allow customers to perform banking transactions in natural language, replacing traditional menus and buttons.
- By unifying data on the Databricks platform, Vietnam’s Techcombank has built AI capabilities to deliver hyper-personalised offers to 15 million customers and expand its footprint beyond its traditional affluent base.
- Researchers at the National University of Singapore have created a wearable device that combines a camera with conversational AI powered by Meta’s Llama models to give sight to the visually impaired.
- Indonesia’s GoTo has migrated half its infrastructure to Alibaba Cloud, paving the way for AI initiatives to solve real-world business problems and support local languages.
