AI in the Canadian Group Benefits Industry
with Jason Spector of Lucite
Part 2: Demystifying the LLM: or exploring the strengths and differences between the major models and why it matters to advisors
The promise of AI is enormous, but it can also be overwhelming. New tools seem to launch every week, each claiming to transform how we work. For advisors, the potential is real: automating tedious workflows, reducing administrative burden, and freeing up time to focus on what matters most – serving clients.
But not all AI is created equal. Some tools overpromise and underdeliver, while others introduce risks, whether through poor accuracy, lack of oversight, or unclear data practices.
For benefits advisors, the question is no longer whether to use AI, but how to adopt it responsibly. Advisors handle highly sensitive personal data, operate in a complex regulatory environment, and face high costs of failure if errors slip through the cracks. The right path forward lies not in generic AI tools applied without context, but in domain-specific solutions designed with industry safeguards and clear parameters.
In this three-part series, we’re unpacking the key considerations and exploring how Canadian group benefits advisors can incorporate AI into their practice, thoughtfully and effectively:
- AI Slop: and why you need to be on the lookout for it
- Demystifying the LLM: and the strengths and differences between the major models
- Beyond the Buzz: How Lucite Helps Advisors Win Back Their Time
2. Demystifying the LLM: or exploring the strengths and differences between the major models and why it matters to advisors
In the first AI blog in this series, we explored the rise of “AI Slop” and “Workslop” and the wide gulf between mediocre and high-quality outputs. In this issue, we’ll go deeper: what is a Large Language Model (LLM) and how can advisors get the best out of today’s AI tools?
What is a Large Language Model (LLM)?
You’ve probably heard of ChatGPT or Claude, but it is worth revisiting how these tools work in practice. All leading LLMs including GPT (OpenAI), Claude (Anthropic), Gemini (Google), and Llama (Meta) share the same foundational process. They are trained on massive text datasets to predict the next word in a sequence. Through repeated exposure and feedback, they learn to generate impressively humanlike responses. At a basic level, LLMs are like parrots. They learn to say the right thing in the right context, but they don’t “understand” what they’re saying in the way a person would. They excel at pattern recognition, not comprehension.
Which Model Is the Best?
There is no one-size-fits-all answer today as to which is the leading LLM. The best model depends entirely on the job to be done. Some are stronger at extracting structured data, others at interpreting ambiguous language, and some excel at writing and research.
Recent industry reporting and research has begun to narrow in on this topic. A 2025 survey published on arXiv, an open access platform where researchers from top universities and AI labs share early findings, reviewed dozens of studies on multi-model strategies and concluded that no single model performs best across all use cases. Another study presented at the IJCAI 2025 conference, on Dynamic and Adaptive Feature Generation with LLM, showed that dynamically selecting the right model within a task, rather than just at the start, can significantly improve both quality and efficiency. We expect the consensus around this topic to continue to build.
In practice, the best results come from combining multiple models, each contributing its strengths. But managing that manually, testing outputs across different chatbots or tools, is time consuming and inconsistent. That is one of the key reasons behind the rise in purpose-built AI platforms. Best-in-class AI software is assigning the right model to the right task and cross-referencing results in the background.
How Lucite Helps
Lucite is Canada’s foremost AI platform built for group benefits advisors to assist with streamlining plan design, renewals, quoting, and claims experience reporting. The team at Lucite has spent thousands of hours testing, benchmarking, and fine-tuning prompts specifically for the work advisors do every day, from analyzing booklets to formatting quotes to generating client-ready deliverables that are customized to each advisor. Lucite’s technical team of self-professed AI nerds eagerly review the latest model releases as they come out, so advisors don’t have to.
An increasing number of advisors partner with Lucite to automate their most tedious and mission critical workflows with the push of a button. Behind the scenes, Lucite orchestrates the most appropriate model for each job, all while ensuring client data remains secure. This gives advisors the speed, accuracy, and clarity of cutting-edge AI without the noise, risk, or trial and error of managing the tech themselves.
