◯ Menu
To further AI understanding and adoption in the insurance industry, Lazarus is producing a series of articles titled Artificial Intelligence - Insights for Insurance, or “AI-II.” Several of these Insights will focus on prompting as prompting is essential to implementing enterprise-scale AI solutions. In this Insight, we will discuss some fundamental concepts while later articles will dive much deeper into prompting. The viewpoints expressed in this Insight come from a combination of Lazarus’ expertise in large language models (LLMs) and our hands-on experience.
Prompt Engineering is the practice of crafting instructions, commands, and questions to effectively utilize an LLM. Successful prompting is both art and science. Engineering connotes that properly engaging with an LLM needs to be a planned activity; more akin to a formal presentation than an ad-hoc conversation. In early 2023, there were junctures when it felt like there was a weekly announcement of an ever-larger LLM, with fervor implying bigger is always better. As 2023 went on, there was an increased realization that for many business needs larger LLMs, or even finetuned LLMs, were not a strict requirement. In many cases, better prompting brought better results from existing LLMs. LLMs are powerful tools, but that power cannot be fully realized without investing sufficient time and effort into learning how to interface with them, and prompting is a key part of this. We could compare it to hiring a brilliant engineer who does not speak your language. It is obvious the engineer is very intelligent, but the communication barrier stops you from accessing their skills.
The solution to this problem is not hiring an even more brilliant engineer. Instead, you would work to break down the communication barrier. Using this analogy, the solution is often more effective prompting, rather than pursuing ever-larger models.
Insurance companies need to develop prompt engineering expertise in 2024. Of course, there are other pressing issues for insurers–inflation, interest rate uncertainty, aging workforces, and casualty losses to name just a few–but if companies delay developing prompting skills they will find themselves at a disadvantage.
Leading technology companies–including Lazarus–have already invested in prompt engineering. However, prompting cannot be handled solely by technology partners. To fully utilize the power of LLMs, insurers must apply their specific industry expertise in prompt development. Some insurance companies have already begun investing in prompting skills.
As LLMs solve more business problems, proper engagement with LLMs will be critical for all knowledge workers. Examining the ubiquity of Google and spreadsheets in the modern corporation is instructive. No one asks “Who at this corporation is in charge of Googling?” or states “Let me speak to the VP of Spreadsheets.”
Just as with Google or spreadsheets, ancillary behaviors will need to be developed. For example, users know that Google search results will start with sponsored content and from organizations that have invested in SEO and users adjust to this reality.
As LLMs solve more business problems, proper engagement with LLMs will be critical for all knowledge workers.
With spreadsheets, any competent user will ask about provenance before using someone else’s spreadsheet (Who developed this? When was it last updated? What has it been used for?).
This same level of pragmatism will have to be applied to prompt engineering. In order to have this level of pragmatism, general prompting will have to be understood by all knowledge workers.
This is not meant to imply that unique Prompt Engineer roles completely go away. Like many skill sets, there will be evolution and likely a very rapid evolution for Prompt Engineers. While general prompting will need to become ubiquitous across the corporation, Prompt Engineers will evolve to take on the most complex prompting challenges where results could have an enterprise impact.
For example, prompt engineers will be needed for industrial-scale prompts where a prompt needs to be incorporated into the workflow for, say, 100 people relying upon the output for mission-critical work on a daily basis. The company needs to be very confident the prompt will scale as needed and will require a prompt engineer to sign off for this level of enterprise prompting.
Accordingly, one critical future role for dedicated prompt engineers will be oversight and control of enterprise-scale prompts. Heavily regulated industries, like Insurance, will evolve to a model where enterprise-grade prompts, as well as prompts generated by LLMs, get special governance. In these governed scenarios, the prompt engineer will verify that the prompts are still valid, that they are detailed enough, and that new situations haven’t arisen that cause the prompt to be outdated or return unexpected results. The prompt engineer becomes responsible for sign-off or remediation. The prompt engineer evolves as opposed to disappearing.
No one asks “Who at this corporation is in charge of Googling?”
Effective prompting is a vital component of implementing LLM solutions in the insurance industry. To keep up with the current state of AI technology, insurers should look to develop their prompt engineering capabilities in 2024.
Knowledge workers across all domains will need to learn prompting skills to effectively use LLM-based tools (general prompting). Dedicated prompt engineering professionals will not disappear, rather their responsibilities will shift towards large-scale and specialized prompting tasks (enterprise prompting).
Lazarus AI develops enterprise-grade foundation models for the insurance industry and beyond. Lazarus AI’s advanced APIs enable organizations to eliminate their processing bottlenecks and provide rapid time to value.
Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur.
Block quote
Ordered list
Unordered list
Bold text
Emphasis
Superscript
Subscript