Generative AI (GenAI) could boost the global economy by up to US$4.4 trillion ($5.78 trillion) annually by 2040, according to McKinsey. The gains will stem from increased operational efficiency through task automation and augmented intelligence, as artificial intelligence (AI) provides insights to inform human decision-making.
Recognising the technology’s value, Manulife recently launched GenAI initiatives in Singapore to transform its business processes and elevate customer service.
The first is a Sales Agent Enablement tool that combines traditional AI, automation and GenAI to generate engagement ideas based on the customer’s last engagements in the past 12 months, life stage and needs, and recent news. Manulife agents can have more purposeful, needs-based conversations to deliver personalised customer experiences even as their client base increases. The tool was launched to a pilot group in May and extended to more than 2,000 agents in Singapore in July. It will be rolled out to Japan and other global markets soon.
Secondly, Manulife’s new Underwriting Assistant leverages GenAI to automate the analysis and summarisation of documents. “Underwriting is a core part of the insurance business, where underwriters decide how much coverage we can give clients based on their health and some of their behaviour. The underwriter’s core job is interpreting medical information and deciding coverage. But in reality, they spend a lot of time consolidating doctor’s reports and writing notes before coming to a decision,” says Mark Czajkowski, Manulife’s chief analytics officer for Asia and chief marketing officer for Singapore. As such, the Underwriting Assistant is designed to improve accuracy in decision-making while freeing up underwriters to focus on more complex cases, which ultimately help accelerate processing times for Manulife’s customers.
Czajkowski adds that GenAI will empower the insurer’s contact centre operators in Singapore to deliver more efficient and accurate client interactions by the end of this year. It will automate call summaries, perform daily trend analysis, and manage complex contract lookups within seconds, enabling operators to provide faster, more precise client responses.
Scaling up
See also: Navigating the workforce shift for a GenAI-powered future
Capturing GenAI’s full value requires infusing the technology throughout the entire organisation. However, moving GenAI prototypes to full-scale production is no easy feat.
“Insurance companies might struggle with limited access to sufficient high-quality data (wherein they find it difficult to obtain diverse datasets representing various risk scenarios, inconsistent or biased historical data) when looking at data sources to train their GenAI models,” explains Deepraj Emmanuel, director, head of strategic engagements for Asean at Kyndryl, an IT infrastructure services provider.
He adds: “Insurers also face reliability and accuracy risks with the AI model producing unrealistic or implausible results. This is a particular concern for insurance companies as AI-generated content used in the risk assessment or claims processing department, for example, requires highly accurate and dependable outputs. [Additionally, they may find it challenging to] integrate GenAI to legacy insurance systems and scale across multiple geographies and disparate business units, particularly given the plethora of applications and platforms used across various use cases within just one company.”
See also: Updated version of Singapore's unified QR code, SGQR+, launched by NETS
To address those challenges, Manulife has set up a GenAI programme. “The programme looks at updating our AI principles, which includes updating our governance policies and processes to enable more users to use GenAI and modernising our IT infrastructure to enable data integration and support GenAI’s nature of real-time data analysis,” shares Czajkowski.
Another important factor is to ensure GenAI’s impact on the business.
We’ve approached this from a global perspective. Identifying common use cases or opportunities important to multiple segments and markets across Manulife’s business makes it easier to scale GenAI. We don’t have to rebuild from scratch every time. So, we focus on building the core concept or use case once and replicating it quickly in other locations globally. This has also helped us optimise the cost [of scaling up GenAI initiatives].
Mark Czajkowski, chief analytics officer for Asia and chief marketing officer for Singapore, Manulife
Agreeing with him, Kyndryl’s Emmanuel adds that a robust AI strategy will help leaders identify business objectives for utilising generative AI across the organisation and build a roadmap for AI implementation. “With clear metrics to measure the impact and ROI from these initiatives, leaders can move quicker from ideation to prototyping and implementation.”
Ensuring AI’s accuracy and reliability
AI hallucinations (wherein AI models produce incorrect or misleading results) and model degradation (where models become less accurate over time) are among the top concerns surrounding GenAI.
To prevent such instances, Manulife has established responsible AI principles. “[Those principles] cover fairness, accountability and transparency and are cascaded across the entire organisation. We also have a very strong technical framework to help us ensure that the innovations we create (by ourselves or with vendor solutions) are of the required quality to manage the risk. Furthermore, we monitor the performance of our AI and machine learning models and have a model retraining strategy and timelines,” states Czajkowski.
Kyndryl’s Emmanuel advises insurers to ensure the accuracy and reliability of their GenAI models by:
- Maintaining high-quality, diverse training data to mitigate risks of model collapse and hallucinations. This includes regularly refreshing datasets with verified information and strictly filtering out AI-generated content from training data, thereby representing real-world scenarios on which the AI model is trained.
- Taking a hybrid approach and not relying exclusively on GenAI. Instead, a rule-based system for robust critical decision-making combined with human oversight and approval for sensitive and high-stakes processes is the best approach.
- Prioritising the development and implementation of explainable AI techniques and creating clear audit trails for AI-generated outputs is vital for insurers. This process provides transparent explanations of AI processes used by internal stakeholders, customers, and regulators.
- Establishing strong governance frameworks with clear guidelines for ethical AI use. As part of this governance framework, companies need to ensure compliance as AI regulations evolve.
- Participating and collaborating in industry/vertical-focused research and development on AI reliability for insurance, with a focus on creating best practices to advance AI safety techniques.
Emmanuel also emphasises the need to use AI and data responsibly. “At the cornerstone is the need to have stringent measures around data protection and security for sensitive customer data. Deploying advanced technologies like homomorphic encryption can assist in securing data analysis, while technology teams should regularly assess and update their cybersecurity posture. A strong data governance and quality framework will optimise for security, privacy, and quality of the data sets utilised; it should also provide clear guidance on which data is collected and what the retention policies are.”
Looking ahead
GenAI is set to transform all aspects of the insurance industry.
We’re already seeing several opportunities for their impact on insurers, and as the models become more sophisticated, so will the solutions the industry can offer to its customers. [Beyond technology, insurers should also] ensure they have AI talent and the right enablement for existing staff on AI technologies and use cases. This will help create an AI-centric growth culture among employees and empower them to lean on the technology for support.
Deepraj Emmanuel, director, head of strategic engagements for Asean, Kyndryl
Recognising that, Manulife offers user training to help its employees become more proficient in AI and be ready for future work requirements. There is also leadership training to help leaders understand their role in identifying AI and deploying use cases, learn responsible AI principles, reimagine future roles and more.
“This year is Manulife’s foundational year in proving that GenAI use cases can come to life and be ‘productionised’. What’s next for us is looking at ways to rapidly scale those use cases across all our markets, triaging the many GenAI prototype ideas coming from different parts of our business, and continuing to equip our workforce at all levels with skills for future roles,” says Czajkowski.
