Anthropic has a new product specifically for financial analysis:

I haven't tried it yet, but apparently the Norwegian Sovereign Wealth fund is using it (see here)

I like it both ways...

I tend to think that there are broadly 2 ways different ways to use AI for research and analysis.

Open - where a Deep Research or similar model will trawl the internet to conduct research and analysis like an analyst (or 10) would to respond to your prompt. This is generally what people mean when they say LLM I believe and is the default you would get when using ChatGPT, or Google's Gemini, Perplexity, etc.

Closed - where you load the information you want like Annual reports, books, documents, investment checklists, websites, etc and ask it to interrogate ONLY THOSE SOURCES for whatever you want to know. Eg - Changes over 5 Annual reports in the Rem report, KAM, related party transactions, strategy, etc, etc.

Most providers have both options, the standard entry level is usually Open, then Closed are typically within that like CustomGPT, Perplexity's Spaces, Gemini Gems, and Google's NotebookLM (standalone).

So what to use?

Best to use more than one model if you can.

I use Google as my primary AI for a few reasons.

For the standard US$20 per month you get both Gemini (Open) and NotebookLM (Closed).

Gemini's Pro 2.5 is one of the top models according to benchmarks and I am used to it now and like it a lot.

NotebookLM has a lot of capacity to upload PDFs, URLs, etc, etc and a lot of ways to output - audio summaries / podcasts, Summaries, mind maps, etc. I like this even more.

One generic approach I use when looking at a new business is to load a standardised prompt into both Gemini and Perplexity Pro for a particular company. Then generate PDF outputs from the responses and load theses from each into NotebookLM and generate an audio summary.

That way in the space of about an hour (1/2 hour to run + 1/2 hour to listen) you get a podcast style run-down of the most important results from your prompt to steer your next steps in a more informed direction.

If you like the pod you can download it, ask notebook to generate a transcript, etc to archive for reference.

You can also customise the pod with more detailed prompts, make it longer, or shorter and even interrupt with your own questions during (interactive mode). Amazing!

Health warning

There is emerging and growing evidence that AI makes you stupid (paraphrasing).

It seems outsourcing your thinking to AI diminishes your ability to do it for yourself. Makes sense to me.

So it probably depends on how you use it?

A bit like your phone or social media, AI is designed to make you dependent on it, so you need to guard against letting it or it will reduce your ability to DYOR just like social media on your phone has shortened your attention span...

The story in today’s Fin Review “Will equity analysts be replaced by AI? That’s a billion-dollar question” opens with “Equity analysts are on the precipice of being hugely disrupted by artificial intelligence.”

https://www.afr.com/technology/will-equity-analysts-be-replaced-by-ai-that-s-a-billion-dollar-question-20250716-p5mfaf

Reminds me of the saying AI probably won’t take your job but someone who knows how to use AI might.

Gary Mishuris recently published an interesting article showing his AI use cases as a Portfolio Manager / Analyst - How AI Is Enhancing My Investment Process— and How It Can Help You

https://behavioralvalueinvestor.substack.com/p/how-ai-is-enhancing-my-investment

A more detailed presentation of Gary's approach is here - https://drive.google.com/file/d/1i-22YGdka0FATSFE_Svk9io3NaRG22dd/view

Of course any AI use cases should be aligned to your individual strategy but you could do a lot worst than starting here.

One strategy that should work for all in equity analysis and beyond is start by trying to get AI to help you with real world problems and tasks, be prepared for it to surprise you and your curiosity will eventually turn you into someone who knows how to use AI.

If you’re still not sure which AI to use for your purpose… ask AI. They (almost?) all have free versions to try.

Start with that real world problem and you’re on your way.

Fin-R1: A specialised LLM for Financial Reasoning and Decision Making

For those who can't make it through the day without using LLMs extensively or those who aspire to get there, here's something to keep on your radar:

Fin-R1 is a targeted artificial intelligence model that’s been specially designed to help solve financial problems. Think of it as a very smart computer program that can understand complicated financial language and data. It’s been trained using a two-step process: first, by learning from thousands of examples of financial reasoning, and then by fine-tuning its responses using techniques that reward it for giving clear, step-by-step answers. Even though it has a relatively small size compared to other models, it manages to perform really well in financial tasks.

What makes Fin-R1 especially interesting is its potential to transform financial decisions. Imagine a tool that can not only crunch numbers but also explain its thought process in an easy-to-understand way. In time, something like this might be a more effective productivity multiplier than a general LLM. Its ability to handle complex scenarios while remaining efficient could mean more accurate and trustworthy support for businesses and consumers alike, paving the way for smarter, more informed financial strategies.

Here's the paper: https://arxiv.org/html/2503.16252v1#S1

Also separately: I've recently completed the Azure AI Foundations certification which has some nice intro stuff. It's a good place to start if you're new. Note that it's very Microsoft / Copilot centric but thematically it's similar to other Large Language models. The Generative AI module that's part of that certification is here: https://learn.microsoft.com/en-us/training/paths/introduction-generative-ai/. (Just the Fundamentals course component)