AI is great, but what can I do with it?

AI has several powerful use-cases. But most recommendations for how to use it are quite over-complicated. I am going to give you 3 ways you can benefit from new AI technology as an actuary. I promise to keep it simple. Instead of teaching you how to build the Death Star with AI, I am going to suggest 3 services that have been helpful for me. Likely, this will all be outdated in a few months. Such is life.

Quick disclaimers:

• I am not paid to promote any of the products or services in this article. I am merely a satisfied user.
• Make sure you abide by your company's AI-related policies. If you acquire and/or use an AI product or service for your work, go through legitimate channels.

Okay, let's get into it.

1. Use Claude

Everyone has heard of ChatGPT. It's popular. But it has a cousin who is way better at programming: Anthropic's Claude.

Claude is widely considered the best large language model (LLM) family for programming, with Sonnet 3.7 being the leading model at the time of this writing. If you are writing small scripts to automate stuff or trying to figure out an Excel formula, using Claude as your go-to will help you crack the problem faster and more effectively.

Claude has a $20/month pro tier with access to Sonnet 3.7 as well as a very impressive Extended thinking mode that can program complicated mathematical algorithms.

A few ways Claude has been helpful for me:

1. Writing python scripts to automate tasks quickly,
2. Gaining an understanding of existing code,
3. Gaining an understanding of insurance products,
4. Figuring out how to hand wash my car. (Okay, this one is not actuarial.)

I haven't found Claude (or any LLM) helpful for understanding the actuarial requirements of regulations such as U.S. GAAP or STAT. Those you still have to learn the old-fashioned way for now.

2. Use Cursor

This is mainly for programmers.

Cursor is built on top of VS Code. It allows you to leverage AI models directly in your code editor. This has the powerful benefit of allowing you to easily inject multiple files into the context of a request. So if I need to perform a refactor spanning multiple files, I can use Cursor to make the request to any LLM (I always use the latest Claude model), then review all of the changes before I accept.

Cursor is widely acclaimed in the AI community as the top solution for an AI code editor. There is another application called Windsurf which is similar, but which I have not tried.

I would highly recommend Cursor over its bigger-name alternative: Github Copilot.

3. Use Open Router to try different models

Instead of paying $20/month to one model provider such as OpenAI or Anthropic, you can simply buy credits in Open Router and try any model you want. This allows you to easily switch between different models. And it is ultimately far more cost effective on average than a subscription to a model provider.

If you bought $20 of credits on Open Router, it might take you several months to a year to exhaust them, even with regular use.

Now this next part is really really only for programmers. Open Router also offers an API so that you can easily switch between models programmatically. This is helpful if you want to try and/or use different LLM's in your application.

Next Steps

That's it. I promised to keep it simple. Hopefully it was.
You can watch a YouTube demo of how Xval uses AI within our application to make life easier for our users.

You can try Xval for yourself at demo.xval.io. If you want to reach out to learn more, you can find me on Linkedin or email me a timothy dot quast @xval.io.