Automating Workflows: CVNA Proxy Analysis
Good morning everyone, today I am once again back to shoving AI generated content at you, hopefully of such quality that you cannot tell! Today’s example is a run through of Carvana’s 2025 proxy report. I’m fairly sure my work can be fully automated and generalized to any stock, so WIP! Unfortunately tables and such don’t embed well in Substack, so if you’re interested give it a download below. Always open to feedback!
Going through a 76 page proxy by hand can be a quite lengthy process (it took me basically a whole workday to manually write a sample output). My goal is to essentially automate the vast majority of this so the imagine workflow could be something like:
You read through the ~2 page summary that covers the basics of compensation, insider transactions, any key terms, etc.
Any additional information is split into defined sections you can easily look further into.
For example we can section out all outstanding equity awards, realized awards from 2024, the total value of all outstanding equity awards (if for whatever reason CVNA was going to get acquired?), the share structure, etc. I want to empower users to be able to ping around to whatever section they need a double click on, while having powerful high level overviews as well.
The plan is to make everything I do as modular as possible which has a plethora of benefits
If you want info/data it’s much easier to find
I can allow you to subscribe to or follow only specific modules you’re interested in. Don’t care about share structure but want an email update on equity awards and targets? I got you (soon tm)
Hallucinations near 0? As far as I know the document has 0 incorrect values despite being ~30 pages of AI generated content. Trying to 1 shot the report or interacting via extended chat windows has a MUCH higher risk in my experience.
If you find a hallucination, let me know!
Comparisons across competitors/similar companies can be done on a relatively like for like basis
We can turn stale documents into living documents. We could say update insider ownership with SEC filings, regenerate any sections of the report that are impacted, etc.
It’d be really cool if I can basically get it to the point that anything you want to know about a proxy, or about recent filings, or about 10-K’s, is being automatically generated. Then just keep moving the needle towards investor presentations, expert calls, alternative data, etc.
If I can compress becoming an expert on a company from a month+ to a couple days of reading, that opens up a ton of possibilities to find even more neat angles of optimization, or just time to hang out at the beach.
Anyways, I’ll likely end up adding some more stuff that’s visual and shows historical comps for a variety of metrics over the coming weeks as I hammer out version 0’s of everything. I’m likely leaning towards the end UI for these being via the Clarity website or email notifs since Substack doesn’t play nice with dynamic charts. The Substack would then maybe act as a “Best Hits” in terms of where I’m seeing data moves and such.
Brief CVNA/KMX/CDLX Note:
Currently travelling, will do my best to automate some Twitter delivery of free CVNA/KMX/CDLX data in the next week or so, keep an eye out.
None of it will be written by me, so doing a small public stress test of the consistency. Cheers.