ChatGPT Use Cases in PE
Any PE associates leveraging ChatGPT on a daily basis? What aspects of the job are you (partially) automating or leveraging ChatGPT / other generative AI software for?
Assuming there are a lot of interesting use cases that could become the norm, especially as we develop a greater understanding of using these tools more effectively
Comments ( 20 )
Bump
b
Finding directional ideas on risks & mitigants
Writing parts of term sheets is the best use case I've found so far.
have been using this to write out all investment committee memos
Doesn't this imply you feed the bot some sensitive info though? Wouldn't that be a compliance breach?
Hey consultant bro, we don't do compliance in PE 😎
fuck compliance bro
Used to summarise research transcripts for market understanding
we use it for summary slides and only if the numbers are checked before publishing. some numbers are off.
also, like stated above, good for ideas you didn't have yourself. the chat bot is good with coming up with lists on all kind of things.
In less than five years you will have a button-like deal in a box. It will take minimum implementation effort to integrate with firm model, comps , and PPT templates, but ChatGPT will be able to spread some basic comps , hook up the financials in the data room to the model, cut revenue cube data, etc. The only reason we are not there yet is that ChatGPT does not have an accuracy guarantee; I won't bore you with the details but this would require a basic rearchitecting of LLMs which is why I say five years not sooner. Just my thoughts from being in the space.
what does cut revenue cube data mean?
ChatGPT:
In a private equity context, "cut revenue cube data" may refer to a process of analyzing revenue data for a company by segmenting it into smaller units, often based on customer, product, or geographical location. This allows private equity firms to identify areas of the business that are performing well or poorly and make strategic decisions to improve the company's revenue growth and profitability.
The term "cube" in this context refers to the three-dimensional nature of the revenue data, as it can be analyzed across different dimensions or segments. "Cutting" the revenue cube data refers to the process of segmenting or slicing the data to examine it from different angles.
Overall, the goal of cutting revenue cube data in private equity is to gain deeper insights into the company's revenue streams and identify opportunities to optimize performance and drive value.
only a mere prospect but I believe it refers to revenue that is segmented as such: think of an x y z axis for revenue segmentation where the axes are
[ geography ] x [ segment ] x [ product ]; or
[ geography ] x [ segment ] x [ revenue by contract type ]; or
[ geography ] x [ product ] x [ revenue by contract type ]; or
[ geography ] x [ segment ] x [ distribution type ]; or
[ geography ] x [ segment ] x [ end market ]
or some other combination ...
Disagree. ChatGPT spits out inaccurate info because it's trained on outdated web-scrapped content that contains doubtful sources. And that's because the trainers of ChatGPT don't have access to the non-public info ( internal financial reports, exec meeting memos, legal documents etc ). But now everyone has seen the power of GPT, they'll start to feed it with non-public corporate data, which will improve accuracy significantly to 99%.
I believe in one or two years, there will be vendors who specialize in this service. So they would purchase GPT model from openAI, and then train GPT on client-specific, non-public datasets. For example, if their client is a law firm, they'll feed GPT with clients' proprietary legal datasets ( past cases, documents etc ), and GPT will learn them and eventually be qualified to perform an associate job. Same goes for PE .
The most scary part is that GPT is an active agent who's able to learn on whatever info that is fed to it. So it's not like a hard-coded search engine, but more like a human who's able to learn. If it's properly coached with high quality proprietary datasets from PE firms, it's gonna be scarily good
Don't disagree with much of this except you won't have accuracy guarantees until it is rearchitected. Packaging existing LLMs to easily train them on new datasets to be delivered through software (kind of like what Twilio did with comms APIs) is probably a 1-2 year problem like you mention, but rearchitecting them to have accuracy guarantees is a fundamentally different issue that =/= training them on specific data sets. I also don't think you should conflate training ChatGPT on a data set with feeding it a data set for analysis.
Following
To answer the question asked about how I use it on a daily basis:
- I shove my raw notes from a call into it, and ask for the top 10-20 most salient bullet points, and it spits them out. I obviously edit them if I got more context or if I understand it differently, but it's a fantastic 80%+ jumping off point
- If there is something I don't know how to model out, I'll ask about it and have it walk me through how it impacts all 3 statements , which has been a life saver for someone like me from a consulting and non-business undergrad background
- When I get a CIM that's crazy jargon-filled or absolutely packed with banker-language such that I can't decipher what the business does, I'll upload the summary and a few other important points and have it summarize what the business does
- Asking for risks in a certain industry
- Lesser extent, but I've been asking for some VBA code so I can do some things faster recently
- If I have a stream of thought that I want to get cleared up/bucketed, I'll just type everything I'm thinking out and then ask it to bucket out what's going on. It's never right, but it's right enough that I can work from there
Overall I think it's a good starting point, it take unorganized matter and organizes it to some degree, but from there I can organize better/faster and the leverage that I get when I only have to think about the ANSWER vs thinking about the PROCESS is super helpful.
Repellat quis repudiandae omnis ipsum rem. Tempora qui laudantium molestiae. Ipsum ad et esse sint. Nisi est dolores et repellat. Labore qui dolores facere odit. Consectetur doloremque facere rerum debitis repudiandae necessitatibus.
Rem est voluptatem quia numquam iste. Reprehenderit quasi sed cupiditate. Rerum autem omnis et ea. Non officiis aut itaque illo.
Nemo quisquam laborum iusto ratione dolorem ea incidunt. Magni quam vel ut dolorum quia. Et ut voluptates commodi sint est recusandae. Ex atque quos repellendus ullam dicta dolores.
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...