On 13 April we had our Santam sponsored webinar that was presented by Munich Re to our members. Thabo Twala, IIG Vice-President welcomed all in attendance to a topic on Insurance Backed Performance Guarantees in an AI environment. Our two speakers, Jascha Prosiegel and Johannes Benedikter from the Munich Re “New Tech and AI” Underwriting department, presented on Munich Re’s appetite in taking risks in the area of new technologies and data-driven business models.
Jascha Prosiegel is the Senior Underwriter & Actuary who has M.Sc. Mathematics from the University of Mainz and is a member of the German Actuarial Association. He joined Munich Re in 2014 in different innovation and technology roles, from renewables over space to AI.
Johannes Benedikter is an Underwriter with a M.Sc. Business in Technology, from TU Munich and joined Munich Re in 2016 within Munich Re’s rotational Trainee program Explore. He has been working in different underwriting roles for innovative risk solutions since 2018.
The topic of insurance-backed performance guarantees is an interesting one as it focuses on the guarantees for the use of AI processes in businesses. In essence, insurance for protecting firms against liability on the failings of their AI is an innovative concept. As we move towards an era of more and more AI into application processes, the higher the exposure to risk is.
AI offers us the ability to reduce costs, improve quality and increase profits in our processes and therefore effectively delegates the responsibility to an algorithm. There is a risk of, whether the model will perform as promised, whether the results will be consistent and accurate over time even when things are changing or what happens when the service is not delivered. aiSure was put together to address these uncertainties by having this promise guaranteed with having formulated an insurance solution.
The insurance is structured in a way that based on your Service Level Agreements that you have with your client for performance of certain functions, based on the extent of the promise in percentage terms, a certain shortfall payout per case will be paid out. The bigger the buffer for the performance shortfall the lower the premium, while the bigger the payout per shortfall, the higher the premium.
The basic consideration that is taken into account when broadly structuring the insurance model is
Moral hazard problem
- Actions taken
- Observable quality
Adverse selection problem
- Information asymmetry / disproportionateness
- Buying motivation of the insurance
Alignment of Interest
- Benefits in claims situation
- Structure of steering
There is a difference between a Performance Guarantee versus a Warranty:
Performance Guarantee is where contracts meant to ensure one party’s ability to complete a predefined task within a predefined period according to a standard. In case of nonfulfillment, the guarantee has the right to make a claim for monetary compensation.
Warranty is where there is a legal obligation providing the warrantee with security that the purchased product performs as stated and all codes and regulations were followed during manufacturing. The end consumer has the right to claim even if the original warrantor went bankrupt.
Interestingly a paradox was created during a survey taken where 78% of people believed that they expect AI to disrupt their industry in the next 10 years but 88% of them do not have confidence in AI-based decisions and outputs. This creates uncertainty of AI buyers as the perceived risks could make the AI business case unviable for decision makers.
A form of AI that uses machine learning from data to detect patterns and learn to improve itself in experience without being explicitly being programmed. There are cases where if the data provided to the model is not correctly labeled will lead to AI failure. See below for an example where in the first row, the last cup is approved while a piece of handle is missing.
The effect of this may lead to a cup being approved without a handle as shown below resulting in an AI failure.
Not all AI use cases will fit for insurance backed by performance guarantee as there might be external factors that will affect the AI prediction model, these include where there is a component of manual review during the process or where credit or market risks may trigger a shock event causing the AI prediction model to fail.
Munich Re has a number of companies that it has since provided aiSure solutions that uses the insurance model for either New Tech (AI applications, internet of things applications, equipment as a service and liquidated damages) or Green Tech (solar power transformation, wind energy, energy storage & e-mobility and circular economy). A company called Fraugster (https://www.fraugster.com/) was highlighted that uses an AI engine that provides protection to companies that you never need to worry about fraud again. The guarantee is that if fraud occurs on your account, you will be given protection by a full chargeback / refund to the merchant that uses this platform to validate customers’ identities and assess fraud risk throughout the customer journey.
Another use case is that of a company called Ascent (https://www.ascentregtech.com/) where it provides protection to its customers against fines and regulatory risk. The use of AI as a tool to limit the probabilities of risk against fines and regulatory risk uses machine learning to automate regulatory knowledge and filters out the regulatory white noise by getting you to the right obligations that matter to the business.
Further companies highlighted during the webinar was Deep Instinct (https://deepinstinct.com) and ClearMetal (https://beta.clearmetal.com)
The process of Munich Re analysing an AI system and then insuring it for performance, the user can place more trust in the AI and scale it across the business with higher confidence. The AI user gets a monetary indemnification that mitigates a potential system underperformance and lets you focus on your business of selling your products or services.
Jascha and Johannes concluded the session and thanked the IIG for the opportunity.
Our MC Thabo thanked the speakers for their time and their insights and knowledge sharing. He then managed a short Q&A session. He closed the session by thanking Santam for their sponsorship of this insight session.
This article was written by Suren Kasil
This IIG Insight session was sponsored by:
