Just can’t properly express my thanks for the AI forum. Well done to all contributants.

Einstein once said “if I had 60 minutes to solve a problem, I’d spend 55 minutes thinking of the right question to ask, for it will lead to the right solution” (something like that, at any rate)

It seems to me, the AI skill we need to master is the art of asking the right question.

as a suggestion, could all Straw people make their suggestion of solid, clarifying questions in designated ‘topic areas’ - over time this library would become worth the Strawman sub in itself. To kick it off, a skeleton of subject might need to be developed. My suggestion would be to use VALUE QUALITY MOMENTUM as the trunk and from there the branches for say quality might be (a) management (b) moat (Product differentials) etc.

i hope this suggestion gains traction, but hey, if it’s no good, go for the three pointer in the digital bin.

Could this be the best ad ever? for AI anyway, well done Google. loved it

https://www.youtube.com/watch?v=-7e6g11BJc0

Part 5 – Recap (and finale)

I built up the previous posts on this topic over time as I went, the below might be a more logical way to step through it…

This will be the starting point for today’s AI Meeting

Why use AI for Equity Research and Analysis?

I think AI is a huge deal and it will become table stakes for equity analysts – if it’s not already.

There’s a big Informational Edge from using the Internet to do Equity Research – all your competitors are using it. So if you don’t you’re at a disadvantage.

There’s a big Analytical Edge from using the Microsoft XL or similar to do Equity Analysis – all your competitors are using it. So if you don’t you’re at a disadvantage.

There’s a big Informational AND Analytical Edge from using AI to do Equity Research and Analysis – all your competitors are using it. So if you don’t you’re at a disadvantage.

Who’s driving?

For analysts that use the internet, XL and AI, it’s like they’re driving a car (using the latest available technology).

Not using the internet or XL is like walking – you’ll get overtaken every time.

Using the internet and XL is like getting the bus, there’s a lot of people on there with you and you’ll get there in the end but if you look out the window, cars are overtaking you.

Using AI as well puts you in the drivers seat.

What to use it for?

There’s endless AI use cases, most of which are outside investing.

I wanted to narrow the scope for such a huge subject to make it approachable and applicable.

So I looked at Generative AI for Qualitative Equity Analysis, because that’s what I’m most interested in and it’s the biggest use case for me today.

I didn’t consider any use cases outside of this but found plenty in my search.

Getting better at AI for Equity Research will get you better at using AI for life.

Why now?

There’s a proliferation of tools out there now and more coming but no burning need for most people to use them.

I don’t want to be like some elderly relatives who had no burning need to use smart phones or the internet when they gained traction but they now struggle to access the services they need because the service providers are trying to funnel all users through their cost effective app.

In time AI Agents will likely be the new apps, so now’s the time to start learning to drive.

Where to begin?

In a nutshell, 1) just get started then 2) keep going.

Try free versions and see where you end up.

There’s a tonne of tools so I started with the biggest, best known among them – ChatGPT, Claude, and Perplexity.

AI currently has a different business model to other tech like Google, Facebook, Twitter, etc who all monetise their use through advertising.

AI has a freemium model where a lite version is usually available for free. Then monthly subscription costs increase with usage – step changes in quality and quantity.

Which tool should I use?

This is still an evolving space but most are US$20 (A$30+) per month for the entry level premium versions.

Which one(s) you’ll want to use will depend on a few things – some are better at different things than others, there’s user preference and what other tech you’re using today.

If you use a lot of Microsoft like me, ChatGPT might be best, if you’re more into Google sheets, android, etc, Google’s AI tools should integrate better.

Claude seems the most ‘ethical’ depending on your beliefs. Perplexity ties all responses back to source for verification if you’re looking to rely on results directly from the AI.

How should I ask it?

Prompting (asking) is a critical skill to develop because this is how you interact with the AI model.

Prompt Engineering (how to structure your question) will play a big part in how useful the eye dropper of information you get back is, from the ocean that’s available.

Trial and Error can work here too but there’s a lot of info on Prompt Engineering available for free online, or you can even ask AI to help you write a prompt to get whatever information you’re looking for.

What should I ask it?

I would start with what’s most important to your investing strategy (for me, management, moats, competition, strategy, etc are a big focus).

I would then move onto the things you don’t like doing, are not good at or would ask an analyst that was working for you (for me that’s industry and competitive analysis).

The more you follow up with questions and commands on the responses you get, the more it becomes like a conversation with a busy analyst that might not know a lot about investing but does have lightning-fast access to a world of information.

What next

The ultimate aim is to embed AI as a tool in your investing process.

That way it gets used every time, even if for different things.

Then you get a lot of reps to build your fluency, so AI becomes a sharp tool in your kit and maybe even a trusted part of your ‘team’.

Q&A

There’s a lot I haven’t covered here, so keen to tap the Strawman brains trust for more use cases, questions, concerns, tips and tricks.

Part 4 – What else?

So much more to all of this… but a few key things to consider next.

It’s a fast moving feast… ChatGPT just released “Deep Research” last week. This has a huge amount of potential for Equity Research Analysts. Its this sort of thing that makes me think Gen AI will be table stakes for Equity Analysts in the near future if not already - https://openai.com/index/introducing-deep-research/

You can’t get it on the free version but ChatGPT Plus users will get it 'soon' for~A$30 per month (Pro users have it now @ ~A$300 per month).

Paywalled AI for Equity Research

So far we’ve only talked about the LLM’s that trawl the internet to answer your prompts. There are a lot of Gen AI tools behind paywalls and I expect we’ll see a lot more of these. I’m using the beta version of AI in Alpha-Sense (https://research.alpha-sense.com/dashboard) which is my primary research subscription tool but likely too expensive for most @ ~$20k p.a. (I have an unusual deal as I’ve been with them since way before they were Alpha-Sense).

I find their Gen AI to be super useful, works the same as ‘public’ AIs with prompts, etc but just on their paywalled data + public data, it’s all referenced back to source, etc.

Reason I mention this is that a lot of other providers like tikr, Stockopedia, Simply Wall St, etc, etc will no doubt be developing their own AI to upsell you. So you might want to see who’s up to where and pricing for these. There’s a few use cases on YT where people are trying to sell you what they have built (mostly AI screeners) which I’d be wary of, but still worth a look to see what’s available and how they approach it.

More thoughts on Prompting

I like @mikebrisy‘s metaphor of using AI as an analyst(s). Each AI Tool has differences in ‘personality’ that you can get to know over time.

Thinking like this, it may be best to use AI for the things you like doing least / to plug your weaknesses, or at least to shortcut / guide more dense qualitative analysis like Industry analysis, competition analysis, etc.

Follow ups / refinement is key. Just like the best interviews are conversations, so too with AI, make it a conversation, that’s what it's built for.

Coaching Mindset

If you treat each different AI model like an analyst you are managing and coaching, rather than one that you are using like a freelancer, that’s probably a better mindset.

That way, you're getting better at understanding the AI’s strengths and weaknesses and guiding them (within each query) as well as improving your own understanding about what they're best at and can be trusted with to get the best out of them.

Then you really are building a ‘team’ of trusted analysts (AI agents may have the potential to become trusted advisers – TBC).

Oher AI Models, experiments and customisations

Its even possible to build or use an existing Custom GPT based on ChatGPT or other models.

Here’s an example of one that uses 37 white papers of Michael Mauboussin – called Mauboussin GPT - https://chatgpt.com/g/g-678299b953b08191bc51bb112c6c6c1c-mauboussin-gpt. Here’s the Mauboussin GPT explainer - https://x.com/PVofFCF/status/1878139785383932050

Google playing catch up

I’ve not come across as much coverage of Googles AI tools and they had a famous fail when they launched their response to Chat GPT a couple of years ago, but looks like they are chasing hard.

NotebookLM by Google pitches itself as “Your Personalized AI Research Assistant” - https://notebooklm.google/. This is a closed environment where you can upload Annual Reports / Presentations, etc so you are only using that information and the models reasoning ability when you ask it questions. It can then do a range of things including “Generate Audio Overviews and listen on-the-go”.

Also Google Gemini Advanced (https://gemini.google/advanced) per post from @Dangles- https://strawman.com/forums/topic/10580#post-32410 who explains it there well.

After you exhaust your free account allotment, both of these Google AI’s are US$20 for paid versions after a free 1 month trial.

Next Level

I’ve been trying to stick to beginner / entry level concepts in these posts because that where I’m up to and suspect most Straw people might be too.

However, I feel like @Strawman gave us a great summary of the intermediate level when he said… https://strawman.com/forums/topic/10580#post-32656

It would be good to use this as the basis for a deeper dive when we zoom around the virtual table tomorrow.

Hope to see a lot of you there then!