First of many major AI inflection points?
With claudes excel plug in and agentic ai tools (and cowork) it seems to me we are at the first major inflection point for AI.
Its very possible it seems that way to me because I have only been passively following AI developments in the industry, since the last time I checked in (~6 months ago) AI has progressed so quickly, I can only imagine what tools will be dropping in the next 6 months. With all the resources flowing to AI just curious as to what other people are seeing at their firms in terms of adoption of these tools.
My firm is smaller more family office like so we're not at the forefront of adopting a lot of these tools just yet but its only a matter of time before we integrate.
Someone makes this post every 3 months, and here we are, ~4 years in, and AI is essentially still a toy with no financially viable use cases.
AI will be a marginal help in some edge cases. It isn't revolutionizing anything, any time soon.
So no one has adopted tools at your firm I take it? How would you describe your firm in terms of size?
Curious on 1) when your last exposure to AI tools was and what they were and 2) What do you see as its "marginal use" and "edge cases"
25+ on the development side.
I got pitched on some AI product at the beginning of March, so 2-3 weeks ago.
I don't see any real use for AI right now. You mentioned that AI has been helpful in underwriting in Excel. Well, that's a marginal utility in an edge use case. I assume you still vet the output of what Claude Excel gives you. So really you're simply saving a few minutes on one tiny, relatively meaningless part of a real estate transaction.
Hardly "revolutionary."
d
My current framework is that AI is best used in tasks where mistakes are low/risk and or immediately self-evident.
For example, if someone doing marketing used AI to scrape the internet and build an email list, it's fine if the AI is 95% accurate. 950/1,000 emails go out successfully, the other 50/1,000 bounce back as undeliverable (self-evidently wrong) and don't do any harm (no risk).
Where AI is dangerous is when it is put in position to "hallucinate" a few key aspects of a multi-step process. All the major models seem to have an issue with BS'ing and confidently giving wrong answers instead of indicating what they don't know. And in a spooky way, since their use of language is strong, you just assume they wouldn't make sub-human IQ level logic errors. But they will sometimes. They'll do 80-90% of steps in a project at "superhuman megamind" level and then sneak in a few logic errors or just straight up incorrect and/or ficitonal statements or numbers that no human would ever present to you. And they'll do it with all the confidence in the world, right alongside the outputs that are superhuman. So if those errors are not immediately self-evident, they will slip through.
This means that for tasks which don't make all errors self-evident, you must recheck every single step the AI takes and verify all its sources and assumptions and content of original sources. And that obviously destroys its efficiency and the utility of using it for those tasks. But it's either that or you don't check it and get embarrassed, which many people have. Some lawyers are using AI brilliantly to drive efficiencies, whereas others are being sanctioned for citing hallucinated AI "cases" that literally do not exist lol.
It's obviously already incredibly helpful to certain kinds of tasks. But for others, it can be a real minefield. And I'm not sure how they bridge the gap to where it stops sneaking in absurd errors. I'm not a tech guy.
Ullam iusto et rem quisquam ipsum delectus id. Soluta vel sit animi qui.
Eum omnis facere dicta officia fugit animi. Quo officiis consequatur quo ut enim voluptatem.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...