Generative AI and the transformation of insurance knowledge work
AI won’t replace people, but people who use AI will.
Time will tell if this often-repeated phrase comes to pass, but it speaks to a larger truth: the nature of technology is disruption. Technology changes the work we do, how we perform it, and the lifestyle we enjoy as a result. Historically, waves of technological disruption (agricultural, industrial, digital) have favored and enabled intellectual and creative pursuits over physical labor.
In each wave, workers have had to adapt both their skills and mindsets to remain competitive in the marketplace. While some may fail to make the switch required of the times in which they live, the vast majority adapt. Through these prior waves of disruption, technology has played an assistive role to humans, with humans seeking to maximize the use of technology for current tasks, thereby freeing up time for higher-value work.
Role reversal: The relationship between humans and technology is changing
The emergence of large language models and generative AI as a business tool is changing this human/technology dynamic in two important ways:
- The capacity of generative AI solutions to synthesize and creatively present vast amounts of information impinges on the last bastion of human value-added knowledge and creative work. There will soon be fewer domains of higher-valued work to which a displaced human workforce can migrate.
- The ever-expansive and curious nature of generative AI solutions requires boundaries unlike those of prior technologies such as ensuring their ethical training, construction, and use. This requires a role reversal between humans and technology in which the human now plays an assistive role to the technology, and in which the human value add comes not through the maximization of the technology, but through its productive constraint.
In these ways, generative AI is not fundamentally a technology disruptor. Rather, it is a technology-enabled business disruptor aimed squarely at an enterprise’s business domain roles, and those which sit at the intersection of business and technology. Roles which can be described as performing knowledge work.
What is insurance knowledge work?
In the insurance industry, we do not sell a tangible product, but rather a promise of protection. Our work to build, price, and execute on the promises we make is knowledge-based in its nature. First described by Peter Drucker, so-called ‘knowledge workers’ are numerous within modern insurance organizations. They may have the title of analyst, consultant, architect, strategist, researcher, or a similar name in an organization, but their work is universally characterized by the following abilities:
- Gathering and synthesizing information into meaningful insights.
- Developing deep subject matter expertise.
- Problem-solving and generating novel solutions.
- Communicating effectively with diverse audiences.
This creative element of knowledge work – the ability to provide not just information but meaning, and the ability to generate novel solutions – has been, up until this moment in time, the sole domain of human actors. It is the primary means through which we augment and add value to information technology solutions that assist us.
Generative AI fundamentally changes this relationship between humans and technology. The ability of large language models and generative AI to become not just providers of information, but of context, meaning, and novel solutions at scale pierces the veil between humans and our technology by enabling the latter to engage in creative pursuits.
In so doing, generative AI is already changing insurance knowledge work as we now know it. Just as with prior business transformations, it will both vastly improve productivity and require workers to adapt. However, the speed at which this transformation will likely happen is much greater than prior waves.
The speed of knowledge work transformation
Recent research from Deloitte reveals that 79% of business and technology leaders expect generative AI to drive substantial transformation within their organization and industry over the next three years, and Gartner forecasts by 2026 over 100 million humans will engage robocolleagues to contribute to their work. Mere months after OpenAI’s public unveiling of ChatGPT, we observe large language models and generative AI solutions already finding their way into enterprise tools for use cases such as content generation, knowledge management, and software development. The adoption of these features could soon usher in corresponding displacement of current knowledge workforce skills and change the nature of work – especially in information-centric industries like insurance.
So what should current insurance knowledge workers do to contemporize skills and remain relevant in light of these shifting dynamics? How do we prepare to step into the value-added roles of the future?
The use of generative AI requires skills that have become less common in the workplace in recent years.
The role reversal between humans and technology in which humans play an assistive role and add value through productive constraint will require us to rebuild some basic competence in two important areas:
- Conversational intelligence: The give and take of a conversation where each party is providing data, context, and emotional cues more closely resembles the interactions large language models are optimized for. Unlike search engine algorithms, generative AI solutions (specifically chatbots) are extremely adept at altering what information is provided to a user and in what form. This requires the user interaction with technology to be more iterative, conversational, collaborative, and even playful than what we’ve become accustomed to, in order to get the most useful data and presentation back from our inquiries.
- Critical thinking: The highly variable nature of what information is provided to a user, and in what form, requires us to approach generative AI solutions with a healthy skepticism. Generative AI solutions are context engines that alter their responses to suit what they believe the user is asking for, and are known to suffer from bias, seasonal laziness, and hallucination; generating inaccurate information as if it were true. Users of generative AI must be mindful of these limitations and not mistake the perceived authoritativeness of AI outputs as absolute truth.
Do these things now
As with all innovation, knowledge workers who see change coming and adopt a growth mindset will be the ones most likely to thrive in a future embedded with generative AI solutions. Insurance knowledge workers should consider doing these four things to broaden understanding and competence with these game-changing tools.
- Learn: One need not be an expert in AI development to realize its benefits, but developing a very basic understanding of how large language models and generative AI solutions are built and trained is a good first step. Any number of resources are available from free YouTube videos to low-cost learning paths offered by online course providers, with the volume and quality of instruction growing rapidly.
- Play: There is no substitute for using generative AI solutions to understand their behavior and how they respond to inquiries with both information and context. Free versions of chatbots such as OpenAI’s ChatGPT and Google’s Gemini are great tools to start building conversational skills. Ask the tool to respond to a question or summarize something you’re interested in and observe how it responds to follow-up requests for more detail or for an answer to be provided in a specific format.
- Use: Whatever your day-to-day work may be, consider how use of an AI solution could augment your productivity. Is there research you’re engaged in that could be accelerated through generative AI? Business requirements or rules it could help author? A proposal or presentation it could check for accuracy and originality? By applying your own expertise and a critical eye to the AI solution’s contribution, you will begin to learn where its greatest value may lie for your current role. As you do, it is important to be mindful of any AI use policies your enterprise has in place, and ensure that your obligations to protecting customer data and trade secret information are never compromised.
- Share: Most enterprises now understand the potential of generative AI, and many have taken steps to either build proprietary small language models or pursue targeted proofs of concept (PoC). Going forward, moving from PoC to large-scale implementation will require a better understanding of specific return on investment for various use cases. By sharing what you learn through experimentation, you can establish yourself as a resource and contributor to the firm’s knowledge base. Choosing to engage, learn, and share are hallmarks of a growth mindset, and may likely be the behaviors that differentiate those who adapt and thrive in the future of insurance knowledge work.
Chad Sands MBA, CPCU, CLU, FLMI is an IT leader and architect at State Farm Insurance. The opinions expressed are the author’s own.