Automatic for the people
AI-adopting insurers have been warned to ensure customer outcomes remain front of mind
By Claire Heaney
A blueprint for the insurance industry to harness generative artificial intelligence calls for companies to be transparent with consumers about the technology’s use.
While the tech could improve affordability, efficiency and equity in insurance, the AI for Better Insurance: Enhancing Customer Outcomes and Industry Challenges report warns there are risks if a consumer-centric rollout is not adopted.
It points to growing scepticism about the general insurance industry, younger generations questioning the need for and cost of cover, consumer demands for seamless, tailored and technology-driven solutions, and wariness about AI.
CSIRO, the national science agency, collaborated with the Insurance Council of Australia on the 40-page report, identifying seven key challenges and opportunities for the industry.
AI technologies including IBM’s hurricane forecasting tool, and smoke detection and bushfire response systems have already shown the potential benefits.
The report comes as the industry grapples with rising operational costs, geopolitical upheaval, cybersecurity risks, weather events linked to climate change, urban expansion into high-risk areas, construction and labour price rises, and cost-of-living pressures.
This has led to increases in premiums, claims and concerns over the insurability of at-risk properties.
Premium rises have forced people to reconsider or drop coverage, to underinsure or move away from automatic renewals.
The report warns that some automated decision-making systems suffer from a lack of transparency and flexibility.
It points to the robodebt scandal, where lack of oversight, transparency and adaptability led to major failures.
Tens of thousands of welfare recipients were affected by the federal government’s automated scheme, which incorrectly demanded they repay debts they did not owe, due to a faulty algorithm.
The report calls for strengthened governance to ensure responsible AI adoption, fostering collaboration and information-sharing across the industry and research organisations, greater resilience, and a strategic and proactive approach to using AI.
It urges the industry to build AI literacy skills to ensure it has a future-ready workforce of leaders and staff using the tool effectively.
About 46,000 people are directly employed in the industry, with thousands more engaged in related services such as brokerage, legal and technology development, the report notes.
The industry should clearly communicate with consumers to ensure it is a trusted partner, being transparent on its use of AI through regular audits and reporting to foster ethical and effective operations.
It also needs to tell consumers about AI’s potential to improve affordability, efficiency and equity, the report says.
Finally, the industry should embrace innovation to tailor insurance products to address emerging challenges such as climate disasters.
Actions could include automation, real-time decision-making, dynamic pricing and predictive analytics to improve operations and the consumer experience.
According to CSIRO, AI may be able to predict risks and help foster a more resilient insurance system, reduce human error and identify suspicious claims patterns.
In the report’s foreword, ICA chief executive Andrew Hall says AI is one of the most powerful enablers of digital transformation, but it presents both tremendous opportunity and a need for careful, responsible stewardship.
This will ensure any benefits are delivered fairly, safely and transparently.
“AI is already being applied in several areas across the industry and while there are limits to its suitability and use case, we are seeing early use cases in underwriting, pricing, claims triage, fraud detection and customer service automation,” Mr Hall says.
“These applications are helping to improve customer experience and outcomes, efficiency and accuracy, reduce manual workloads, and deliver faster service to customers.”
He says the general insurance industry, like others, is exploring the potential of AI to improve services and streamline operations, but there are risks around data privacy, unintended bias in automated decision-making and limited transparency.
“The way we adopt these technologies matters. We must take a consumer-centric, values-led approach to the development and deployment of AI, ensuring that the benefits of innovation are delivered fairly, safely and transparently.”
Mr Hall says the industry must work closely with government and regulators to create guardrails that promote safe and responsible adoption.
CSIRO, in consultation with the ICA AI Working Group, has identified five priority AI use cases for the Australian general insurance industry.
Automated claims processing and triage
The report says AI systems could provide deeper analysis of claims trends and hasten claims processing. Improved liability verification could lead to faster processing of positive customer cases, give greater accuracy in repair allocation and match customers to the right repairers quickly.
It has the potential to free up claims handlers’ time, so they can focus on higher-priority, more complex work and support customers.
But there are cybersecurity concerns around personal data and the potential for external data leakage.
“The cost associated with development and implementation of AI systems, along with cybersecurity challenges, could disproportionately impact smaller insurers, limiting their ability to adopt AI technologies,” the report says.
It recommends broader governance requirements, including disclosure of AI usage and regular reviews of AI model risk management.
Fraud detection and prevention
AI may pick up fraudulent behaviour more effectively. This can help lower operational costs, which may lead to cheaper premiums and resources being allocated to genuine claims.
The report says a technical challenge arises from the potential limitations of historical training data.
“Specifically, AI systems trained on historical fraud patterns may be unable to detect new AI-enabled fraudulent activities.
Additionally, insurers may not have sufficient volumes of training data, which may create blind spots in fraud detection capabilities.
“As fraud detection becomes more automated, there is a risk of losing valuable human knowledge and pattern recognition of fraud previously gained and embedded in well-understood manual processes.”
The report says industry collaboration to share data on fraudulent claims may enhance scam detection effectiveness.
Enhanced underwriting and risk assessment
Tailored pricing and improved risk awareness, operational gains and professional development are cited as opportunities.
But the report warns of bias and a risk of uninsurable areas, and loss of human expertise in underwriting decisions. Regular auditing of AI accuracy and fairness may help, but there are reputational and operational risks of mistakes.
Natural disaster impact prediction and response
There is potential for improved assessment and warnings, and faster servicing when dealing with disasters.
Customers may benefit from reduced gaps, quicker claims and immediate payments.
Enhanced loss prevention and early warnings are an advantage, but risk-based pricing may negatively impact insurance affordability for some customers.
Insurer may lose intellectual property if it is shared, resulting in hesitance among large players that have invested heavily in their own data collection systems.
Operational control and compliance
Centralised compliance checks and scalable population analysis, and consistent decisions are seen as positives.
But AI may weaken monitoring and regulatory reporting, and there is the risk of a lack of human judgment and shortage of subjectivity or nuance.
