AI and Excel
How long do you expect until we strictly interact with excel through a LLM prompt (like the current Claude and ChatGPT excel plug ins). Sure, AI makes mistakes and hallucinates but its getting better and better? Obviously this is a real estate/finance forum and and not for tech experts but always fun to speculate.
I'm only a student so apologies if I'm way off the mark but I reckon expressing intent about what you want to do on Excel and AI doing it can't be too far off at all. Wouldn't be surprised if this becomes very common among smaller firms as they don't have as much bureaucracy as bigger ones. They will probably adopt this sort of stuff quite quickly if it saves them a couple hours -- within 12-24 months probably. Wouldn't be out of the question to make a pretty reliable industry grade wrapper for Excel given what's already here.
I've spent the last two months in Claude building a new multifamily pro forma using bits and pieces of all the pro formas I've used in the past.
Is it an incredible tool that gets better with every new version? Yes
Is it still highly flawed and needs to be fully guided by a human? Yes
It still makes a ton of mistakes when pulling data, wiring cells and making assumptions. Unless you 100% know how the underwriting is supposed to work and what it is supposed to do at an institutional level, you'll have errors in the model that you will not be aware of that will throw the returns of a deal off. Sometimes by a little, sometimes by a lot.
I've spent probably $500 - $600 in extra tokens to fully build, audit, break, build, audit, break and I'm still going - even though I have Claude log and remember the work. I'm 95% of the way done, but every little tweak can make a difference.
But its a great tool for a "quick pass" on deals by simply uploading the broker OM and making your own quick adjustments to see if a deal is worth spending more time on.
My opinion is that AI is similar to a calculator - it needs someone competent to operate in order for it to produce the right output. AI will be disruptive tech, but I doubt it will be more impactful than the assembly line, the wheel, the airplane personal computer, calculators or even excel for that matter.
I bet a ton of errors (logic, calculations, data) already exist in the average model. My guess is Claude already makes less errors than a human, we just continue to make errors in prompting and Claude is getting better at recognizing our intent even when direction is poor. Claude is only as smart as the direction given and it’s leaning to manage up.
Agreed - it still gets things like circ refs, formatting, organization, inputs, and optionality incorrect without strong context and the user already having a clear vision of what and how they want to accomplish something.
It is great at building out/reworking complex formulas once you nail down the intent and features, it’s pretty terrible if you ask it to build from scratch even with relatively strong prompts/context.
It is great for multi tasking while Im working on other things however, big timesaver and faster and safer than my analyst once you explain vision and align on plan to have it execute in background.
These are the most obvious retarded AI written posts I have ever seen.
Expedita maiores voluptatem molestiae quis fugiat. Sapiente adipisci quam odio mollitia. Voluptas illo ut quod et tempora esse quis. Repudiandae magni saepe molestiae corporis dignissimos. Optio accusantium sapiente doloremque.
Quia voluptates quam eum consequatur eum ea amet. Explicabo qui dolorem porro inventore modi pariatur laudantium. Eligendi perspiciatis veritatis facere esse dolor dolor.
Esse vel natus perferendis sed voluptatibus ipsa. Laborum quos dolor a.
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...