Artificial intelligence presents the insurance industry with a challenge unlike most technological developments before it. It is not merely a new tool that insurers may adopt to improve their operations. It is also a technology capable of changing the frequency, severity, observability and correlation of the risks they insure.
The same development may therefore create opportunity within an insurance business while changing the nature of its liabilities. Artificial intelligence may improve risk analysis, underwriting, claims management and operational efficiency. Yet its growing deployment may also introduce new exposures, create dependencies shared by apparently unrelated insureds and alter the assumptions upon which established insurance products and portfolios have been constructed.
For insurers, these developments raise a fundamental question: how should an industry built around understanding uncertainty respond when the technology itself is changing the nature of risk?
This is the question addressed by SINGULARITY®, an artificial intelligence consultancy framework provided by X from GENERAL INTELLIGENCE PLC to insurance providers. It examines the strategic, technical, commercial, actuarial and institutional implications of increasingly capable artificial intelligence, both as a capability within the insurance enterprise and as a source of emerging insurable risk.
The framework begins with a principle of commercial discipline: artificial intelligence is a means, not an end. The task is not to follow technological fashion or to assume that every new exposure can be accommodated within an existing insurance category. It is to determine where artificial intelligence can create durable value, where it changes the risks being assumed and where the evidence supports a responsible insurance proposition.
Two transformations, one insurance problem
The first transformation concerns the insurer itself. Artificial intelligence may extend analytical capacity, help identify patterns in large bodies of information, support underwriting and claims decisions, improve the use of specialist knowledge and change the economics of insurance operations.
These opportunities are significant, but they are not automatic. A technically capable system may fail to deliver commercial value. Its outputs may be unreliable under particular conditions, its implementation may introduce new dependencies, or its complexity may outweigh its contribution. The relevant question is whether the capability improves the quality or economics of the insurance business in a way that justifies its cost, risks and consequences.
The second transformation concerns the risk environment. As artificial intelligence becomes embedded in products, services, infrastructure and organisational decisions, it may create exposures that differ materially from those associated with conventional technologies.
An AI system may produce an error that affects a single user, or it may contribute to losses across numerous organisations relying upon a common model, software component, infrastructure provider or data source. A system may perform adequately in isolation while creating a significant accumulation exposure when deployed widely. Failures, vulnerabilities or unexpected behaviour may propagate through interconnected economic activities.
These possibilities do not establish that every AI-related risk will be systemic, nor that every emerging exposure will prove insurable. They do, however, make it important to examine not only the individual insured risk but also the dependencies connecting it to the wider technological and economic environment.
SINGULARITY® brings these two questions together: how artificial intelligence can contribute to the insurer’s business and how the insurer should understand the risks created or transformed by the technology.
The limits of yesterday’s assumptions
Insurance depends upon the disciplined classification, pricing, selection and transfer of risk. Its economic foundations require an insurer to form a sufficiently informed view of the exposure being accepted, the potential distribution of losses, the terms under which liability attaches and the capital required to support the undertaking.
Emerging artificial intelligence risks may challenge these tasks in several ways. The technology may develop faster than historical claims experience can accumulate. The circumstances in which losses arise may change as systems become more capable or more widely deployed. The evidence required to establish causation may be difficult to obtain. The same incident may affect multiple policyholders, while contractual boundaries between different forms of liability remain contested or uncertain.
Historical data remains important, but it may not adequately represent a risk whose underlying technology, adoption or operating environment is changing rapidly. In such circumstances, insurers may need to combine available loss evidence with technical research, scenario analysis, engineering knowledge, operational evidence and explicit consideration of uncertainty.
The challenge is not to abandon established actuarial and underwriting disciplines. It is to determine where those disciplines require additional evidence, different assumptions or new methods of assessment.
SINGULARITY® provides a structured basis for examining these questions before an insurer commits capital, develops a product or accepts an exposure that it cannot yet adequately understand.
From technological possibility to insurable risk
The fact that a risk exists does not mean it can be insured responsibly. A commercial insurance proposition requires more than a plausible description of a possible loss. The exposure must be defined sufficiently well to support decisions about coverage, pricing, limits, deductibles, exclusions, aggregation, claims handling, reinsurance and capital.
For AI-related risks, this may require insurers to investigate how a system is developed and deployed, what functions it performs, how it is supervised, which external services it depends upon and what happens when its performance deteriorates or its behaviour differs from expectation.
The analysis may also need to distinguish between different sources of loss: an individual system failure, a security incident, an erroneous output, an operational interruption, a failure of human oversight or a common disruption affecting numerous users of the same underlying technology. These scenarios may have different causes, consequences, evidential requirements and implications for coverage.
The difficulty is compounded when the relevant technology changes after the policy has been written. An insured organisation may alter its systems, introduce new models, change suppliers or expand its reliance upon automated processes. Such developments may materially affect the exposure originally assessed.
SINGULARITY® considers whether the available evidence supports a sufficiently clear and commercially sound insurance proposition. Where the exposure remains inadequately understood, the appropriate conclusion may be to investigate further, narrow the proposed coverage, impose relevant conditions, defer the decision or decline the risk.
An emerging risk does not become insurable simply because a market exists for transferring it.
The hidden danger of common dependencies
One of the most consequential questions for insurers is whether apparently separate risks are genuinely independent.
An insurer may hold policies covering many different businesses, sectors and geographical areas. On the surface, the portfolio may appear diversified. Yet if those insureds depend upon the same AI model, cloud infrastructure, software component, data provider or technical service, a common failure may affect several of them at once.
The resulting accumulation may not be visible when each insured is assessed individually. A system that appears reliable in one application may create a different risk when used at scale, particularly where multiple organisations depend upon the same underlying capability.
This is a familiar insurance problem in a new technological context: the relationship between individual exposures and the aggregate portfolio. But the speed of adoption, opacity of technical dependencies and limited historical evidence associated with emerging AI systems may make the task more difficult.
SINGULARITY® examines how insurers might identify such relationships, assess plausible loss scenarios and consider their implications for underwriting, portfolio construction, reinsurance and capital. This may involve looking beyond the immediate policyholder to the technical and operational dependencies on which the insured organisation relies.
The objective is not to assume that common technology inevitably produces correlated losses. It is to establish whether material dependencies exist, how they might behave under adverse conditions and whether the insurer’s exposure remains within its risk appetite.
Underwriting intelligence in a changing technological environment
The use of artificial intelligence in underwriting presents its own questions. Greater analytical capacity may help insurers interpret submissions, identify relevant characteristics, examine inconsistencies and investigate exposures. Yet the quality of the resulting decision depends upon the reliability of the information, the appropriateness of the analysis and the judgement applied to its outputs.
An insurer must understand not only what a system concludes, but also the conditions under which the conclusion is reliable, the limitations of the analysis and the consequences of error. Where artificial intelligence contributes to a consequential decision, validation, monitoring and clear allocation of responsibility become essential.
The framework considers how AI capabilities may be assessed in relation to underwriting, pricing, claims, fraud, portfolio management, customer service, research and operations. It also examines whether a proposed investment genuinely improves the insurance business or merely adds technological complexity.
The appropriate relationship between people and machines will vary by task. The objective is neither to preserve human involvement for its own sake nor to automate for its own sake. It is to establish where each can make the more appropriate contribution, with professional accountability retained where consequential decisions require it.
The economics of an AI insurance proposition
A new category of technological risk may attract commercial interest before the evidence required to understand it has matured. Equally, an insurer may identify a genuine opportunity to develop a valuable product, improve its service or strengthen its competitive position.
In either case, commercial enthusiasm must be tested against the underlying economics of the proposition.
For an emerging insurance product, relevant questions include the likely sources of loss, the adequacy of available evidence, the scope and clarity of coverage, the potential for accumulation, the availability and cost of reinsurance, the capital required and the insurer’s ability to manage claims. The assessment should also consider whether the risk can be monitored as the technology and its deployment evolve.
For an internal AI investment, the questions are different but related: whether the system performs reliably, whether the expected benefit is credible, whether the cost and dependencies are acceptable and whether the organisation can govern the capability effectively over time.
SINGULARITY® applies the same underlying discipline to both. A technically impressive proposition is not necessarily a commercially valuable one. The relevant test is whether the proposed capability or transfer of risk makes a worthwhile and enduring contribution to the insurance enterprise.
Research, uncertainty and independent challenge
The development of artificial intelligence cannot be understood solely by examining the products currently available. Advances in computational architectures, reasoning systems, intelligent agents, autonomy, evaluation and safety may change the capabilities available to organisations and the nature of their dependencies.
SINGULARITY® is informed by research undertaken by X from GENERAL INTELLIGENCE PLC into artificial intelligence from scientific, technical, organisational, economic and institutional perspectives. This research is considered alongside recognised standards, regulation, actuarial and insurance practice, operational evidence and other relevant sources.
The objective is not to predict the future with certainty. It is to distinguish demonstrated capability from aspiration, material progress from transient enthusiasm and technological novelty from strategic significance.
Where evidence is incomplete or competing interpretations exist, uncertainty must be identified rather than concealed. This is especially important in insurance, where confidence unsupported by evidence can lead to inadequate pricing, poorly defined coverage or an accumulation of exposures whose relationships have not been properly understood.
Independent challenge is therefore central to the framework. It examines whether the question has been defined correctly, whether the assumptions are supported by evidence, whether credible alternatives have been considered and whether the organisation will retain meaningful control over the technology or exposure.
The appropriate conclusion may be to proceed, to proceed differently, to investigate further, to defer or not to proceed.
A discipline for the age of increasingly capable AI
The word singularity is often used to describe a hypothetical point at which technological change becomes exceptionally rapid or difficult to anticipate. SINGULARITY® does not depend upon the assumption that such a point will occur at a particular time, or that any single forecast of the future is correct.
Its concern is practical: as artificial intelligence becomes more capable and more deeply embedded within economic activity, insurers need a disciplined means of understanding what those developments mean for risk, value and institutional responsibility.
The framework is founded upon five principles: long-term owner value, contribution, judgement, simplicity and ordinary decency. Together, they establish a basis for assessing technological opportunity without losing sight of commercial purpose, professional accountability and the obligations of an insurance institution.
For insurers, the strategic challenge is twofold. They must understand how artificial intelligence may improve their own capabilities and they must assess how the technology may alter the exposures they accept, the portfolios they construct and the capital they deploy.
The objective is not simply greater automation, greater computational power or greater insurance capacity. It is a more informed understanding of what should be done, what should be insured and what should remain beyond the boundaries of a responsible risk transfer.
Artificial intelligence is the means. SINGULARITY® is the capability. Insurance judgement is the discipline. The responsible understanding and transfer of risk is the objective.
FURTHER INFORMATION:
1. About SINGULARITY®
SINGULARITY® is an artificial intelligence consultancy framework provided by X from GENERAL INTELLIGENCE PLC to insurance providers. It addresses the strategic, technical, commercial, actuarial and institutional questions arising as artificial intelligence becomes more capable and more deeply embedded within economic activity.
2. About GENERAL INTELLIGENCE PLC
GENERAL INTELLIGENCE PLC is a Scottish company established in 1896. X is its trading name and registered trade mark. The company undertakes artificial intelligence research and consultancy concerning the strategic, technical and institutional dimensions of intelligent systems.
3. The central proposition
SINGULARITY® considers artificial intelligence in two related ways: as a capability that may create value within an insurance business and as a technology that may change the nature of the risks insurers underwrite. The framework examines both the opportunities and the exposures, without assuming that every new capability is valuable or every emerging risk is insurable.
4. The meaning of “singularity”
The term singularity is commonly associated with hypothetical scenarios involving exceptionally rapid or difficult-to-anticipate technological change. The consultancy framework does not require a particular forecast or date. It examines the practical implications of increasingly capable artificial intelligence for insurance, risk, capital and institutional responsibility.
5. Artificial intelligence as an insured risk
Artificial intelligence may introduce or alter exposures relating to erroneous outputs, operational interruption, security, liability, dependence upon third-party systems and the consequences of increasingly automated decisions. The nature and insurability of each exposure depend upon the circumstances, the evidence available, the contractual terms and the insurer’s ability to assess and manage the risk.
6. The two-sided insurance challenge
Insurers must consider how artificial intelligence may improve underwriting, claims, analysis and operations, while also examining how widespread deployment of the technology may change the frequency, severity, observability or correlation of losses. These questions are connected but require distinct forms of analysis.
7. Insurability
The existence of an exposure does not establish that it can be insured responsibly. Insurability requires a sufficiently informed understanding of the risk to support appropriate decisions about coverage, pricing, limits, deductibles, exclusions, accumulation, claims, reinsurance and capital.
8. Emerging risk and limited historical evidence
New or rapidly changing technologies may not have a sufficiently mature history of losses to support conventional analysis on its own. Insurers may need to consider technical research, operational evidence, plausible loss scenarios, expert analysis and explicit uncertainty alongside historical data. Such methods support assessment but do not eliminate uncertainty.
9. Common technological dependencies
Apparently separate insureds may rely upon common models, software components, infrastructure providers, data sources or other technical systems. A shared dependency may create the potential for correlated losses or accumulation across a portfolio. The framework examines whether these dependencies are material and how they should influence underwriting and risk management.
10. Portfolio accumulation and correlation
Portfolio risk cannot always be understood by assessing policies independently. Insurers must consider how exposures may interact under common events or shared dependencies and whether the resulting aggregate risk remains consistent with their risk appetite, reinsurance arrangements and capital resources.
11. Underwriting and pricing
Artificial intelligence may assist underwriting and pricing by helping analyse information, identify patterns and investigate exposures. The reliability of these contributions depends upon data quality, appropriate validation, the relevance of the analysis and professional interpretation. Technological capability alone does not establish that an underwriting decision is sound.
12. Claims and causation
AI-related losses may involve complex interactions between technology, human decisions, external services and operating environments. Claims assessment may require consideration of the circumstances of the event, the relevant evidence, contractual boundaries and the role of different parties. The framework does not assume that causation will always be easy to establish.
13. Coverage and contractual boundaries
Emerging exposures may require careful consideration of policy wording, definitions, exclusions, limits, deductibles and the relationship between different forms of cover. The objective is to understand what is being transferred, under what circumstances liability attaches and whether the contractual arrangement reflects the risk intended to be accepted.
14. Reinsurance and capital
AI-related exposures may have implications beyond individual policies, particularly where losses are correlated or accumulate across portfolios. The framework considers the relationship between underwriting decisions, reinsurance, capital requirements and the insurer’s broader risk appetite.
15. Artificial intelligence within the insurer
Potential applications include underwriting, pricing, claims, fraud, portfolio management, customer service, research and operations. Each should be assessed according to its intended purpose, reliability, commercial contribution, implementation requirements and governance needs.
16. Technology evaluation
Assessment of a proposed platform, model or application may include analytical performance, accuracy, reliability, robustness, resilience, security, privacy, explainability, information quality, data provenance, integration, intellectual property, supplier dependency, regulation, implementation and total cost of ownership.
17. Human judgement and accountability
The use of increasingly capable systems does not transfer responsibility for underwriting, claims decisions, portfolio exposures or the operation of an insurance business from the institution to the technology. Where AI contributes to consequential decisions, the basis for that contribution should be sufficiently understood to permit appropriate professional scrutiny.
18. Governance and institutional control
Governance may include decision rights, human oversight, testing, validation, monitoring, documentation, information controls, incident management and criteria for reviewing, modifying or discontinuing an initiative. Arrangements should be proportionate to the purpose of the system and the consequences of its failure.
19. Research and uncertainty
SINGULARITY® is informed by research into artificial intelligence from scientific, technical, organisational, economic and institutional perspectives. It considers computational capability, architectures, reasoning systems, intelligent agents, evaluation, safety, governance and regulation. Where evidence is incomplete or competing interpretations exist, uncertainty is identified rather than concealed.
20. Independent challenge
The framework examines whether the underlying question has been properly defined, whether expected benefits and underwriting assumptions are supported by evidence, whether credible alternatives have been considered, whether common dependencies and accumulation have been understood and whether the organisation will retain meaningful control over the resulting capability or exposure.
21. Commercial value
Potential value may arise through improved risk selection, stronger underwriting, greater analytical capacity, more effective claims management, better use of specialist knowledge, operational efficiency, improved client service, greater resilience or new insurance products. Benefits must be considered alongside costs, risks, dependencies and the possibility that technological complexity will exceed the value created.
22. The five principles
SINGULARITY® is founded upon five principles: long-term owner value, contribution, judgement, simplicity and ordinary decency. These principles connect commercial purpose with professional responsibility and provide a basis for assessing technological investments and emerging insurance propositions.
23. Proportionate consultancy
The scale of consultancy should reflect the significance and complexity of the question. SINGULARITY® may be applied through a focused strategic assignment, independent technology assessment, underwriting or insurability review, investment evaluation, AI risk assessment, portfolio or accumulation review, governance assessment or continuing strategic advice. The preferred scope is the smallest engagement capable of answering the question properly.
24. What the framework does not claim
SINGULARITY® does not assume that artificial intelligence will develop according to a particular timetable, that every emerging AI exposure will be insurable, that all shared technological dependencies will produce correlated losses or that AI can eliminate uncertainty. Its purpose is to assess opportunities and risks using evidence, professional judgement and proportionate scrutiny.
25. The SINGULARITY® test
The framework asks: What creates long-term value? What genuinely contributes to it? What risks are being created, transferred or changed? What requires professional judgement? Are apparently separate risks genuinely independent? Could common technological dependencies create material accumulation? Can an emerging exposure be sufficiently understood to support responsible insurance? Is artificial intelligence the appropriate means? Can the objective be achieved more simply? What are the enduring costs, dependencies and risks? Can the proposed course be properly governed and controlled?