But exploiting this dynamic led to today's high level of exposure to software over time that will be tough to deal with. Perhaps we see more mez type capital come in as a solution to address parts of this maturity wall as 1L lenders underwrite lower LTVs, but I am not sure that bid is there...

while "true leverage" represented by debt/UFCF or debt/EBITDA-capex was not as egregious vs. broadly syndicated loans given higher capital intensity in the overall lev loans index (BCRED often highlighted this discrepancy in "true leverage" vs. public markets in their marketing materials).

(whereas high headline leverage would constrain access to public markets as a result of CCC ratings/limited appetite for ~6-7+ handle headline leverage)

Given capital light nature of software, there was a kind of "arb" where private credit was able to underwrite high "headline leverage" represented by debt/EBITDA multiples for software companies

There is also coverage of some interesting finance-specific workflows using general purpose tools such as NotebookLM and Opus 4.5. Check these write ups out if you haven’t already!

The site already includes in-depth reviews of platforms such as Hudson Labs, Quartr, Shortcut AI, Ramp Sheets, edmundSEC, roic. ai – all posted over the past couple of months.

Been surprised by the reach of the email list, which now includes subscribers from Norway, Brazil, Sweden, New Zealand and China!

www.buysideaireviews.com/p/ai-workflo...

Here's an example of how you can use multi-transcript analysis within Quartr for idea generation and industry theme tracking (note: this is not a sponsored post - I just find it very cool that this is free!).

It is also the only free service I know of that offers built-in multi-transcript analysis capabilities, which are very useful for idea idea generation and industry theme tracking.

I spent the most time trying out Quartr, which probably has the best mobile app functionality for transcript analysis out of any service I’ve tried (paid or free).

But Claude models still have trouble with PDF data retrieval and Excel tasks.

The hype train is going to roll even faster and louder as we approach the Anthropic IPO, so it's important to stay objective.

Stay tuned for more simple workflows like this. I am certainly “non-tech” and the workflows I cover will largely be skewed to simplicity + utility vs. technological complexity.

Once the prompt is set up, it is something that’s easy to run quarterly to keep a pulse on the direct lending market without waiting for external research.

The task involved pulling new loan data from 1,000+ positions in BDC loan portfolios, providing granular insights into direct lending loan pricing trends as soon as BDC 10-Q/K’s are released (see Exhibit above).

The first post featured using Claude Opus 4.5 and NotebookLM on a data extraction task that would not be worth the time for me without AI.