A message from Octus: Coverage fell from 5.0x to 3.3x. See which CLO managers still have cushion

US BSL net leverage has climbed to close to 5.0x since 2021. Interest coverage has fallen from 5.0x in 2022 to roughly 3.3x. Cushion is shrinking, and manager selection carries more weight.

The Octus CLO Benchmarking brief puts numbers behind the comparison. It measures five notable 2025-vintage deals against market average on spreads, leverage and coverage. It ranks the top 15 US and top 15 European managers, then isolates B3 or lower exposure to show which ones hold up under stress. Use it as a working comparison, then rebuild the rankings around your own peer set.

Inside the brief:

Find out where your managers land and if you want a custom analysis reach out to us at [email protected].

Welcome back!

This is a good one.

We’re talking about the early innings debate on whether AI rollups in PE & SMBs are going to succeed or not.

My expectation for this piece is that this is not a 12-0 verdict, but just beginning to scratch the surface of something that’s going to be evolving dramatically year-over-year; but more like quarter-over-quarter. I’ve talked to many ppl with heated opinions on this and hope to talk to many more!

A lot of this discussion got kicked off by Greg Isenberg, a software engineer and host of the Startup Ideas Podcast. He called buying SMBs a "massive $5T opportunity” which naturally sparked a lot of thoughts from SMB & LMM PE professionals.

The takeaway seems to be that AI in SMBs/LMM is just a tool in the toolbox as opposed to the end all be all. Aka you can’t run an Agent-only business in the blue collar world. For SMBs, these businesses are so thinly staffed to begin with, that the real operating leverage and AI-related synergies are likely to come from deployment across larger organizations. However, there seems to be some outliers and some firms deploying the AI rollup playbook at scale, so let’s get right into it.

Also, we published our State of AI in Finance report yesterday - check it out. We surveyed Buyside professionals and Bankers to see what AI they love and use and had some great insights. It’s available for every active Buyside Hub user.

With 1.8mm views, this article set FinTwit alight a couple weeks ago.

Greg sized up $5T worth of American small businesses that need to change hands by 2035 as a massive opportunity due to impending retirements. Greg has a holdco himself and thinks AI agents are good enough to actually run these businesses; typing numbers in, chasing documents, and writing status updates.

He develops agent files that agents can work within - involving creating drafts, sending things for review, and following specific documentation.

From listening to Greg’s video on this, I thought he has a generally pretty naive view about just having a great GM and then plugging in agents in the background and everything else being fine. He also says these deals “are too small for the big funds”….not true…

Candidly, seeing "Profit 5-10% → 30%-40%” looks like LARPing as an actual operator.

But what’s the goal with Greg’s agent-led approach? That the human becomes the accountability layer; moving from a contributor role to a signoff/managerial role.

When talking to other SMB owners, I kept hearing something similar to what Greg is saying. You need to build mutli-step workflows for agents that STOP at a key point. That key point might be before buying something or sending out a quote. AI can build a draft quote, set a schedule, send automated outreach, and do data pulls and analysis; but it might stop before sending out the quote or replying to the 2nd or 3rd exchange that a customer has. That might be where you add a layer of personalization.

But it does seem like Silicon Valley is all in on the “AI Rollup” strategy. Greg said he also received a lot of DMs saying to “stop talking about it publicly”….meaning people are finding success with this..

He calls out Thrive Holdings and General Catalyst as big funds already deploying AI enhancements at scale. Thrive Holdings, backed by Josh Kushner, is not a firm new to you all - we talked about Thrive Eternal in depth as a strategy focused on acquiring longstanding, everlasting, and scarce institutions like sports teams. Another focus Kushner has are these AI-integrated rollups and Thrive has gone out and acquired 50 accounting firms; buying majority stakes, combining and centralizing back office, marketing, and recruiting functions, and then integrating AI for data mapping and memo writing.

Thrive’s platform is Crete Professional Alliance, an AI-powered Accounting rollup play. Apparently, one accountant backed by Thrive went from 180 hours to 15 hours of work. This allows for 1) more time taking on more clients and work and 2) more client-facing and relationship-driven work.

The next group is Long Lake, “a permanent-capital holding company transforming the services economy with applied AI”, backed by General Catalyst and other VCs. General Catalyst set aside $1.5B for AI rollups - including backing Long Lake. For their ideal structure, the Cash purchase mix is 60%-70% and then the rest is tied to the seller.

For Long Lake CEO Alex Taubman (yes, that Taubman family) he constantly draws comparisons to being an AI version of Danaher or Constellation.

Long Lake has completed 40 acquisitions and has built a proprietary AI platform, Nexus, to deploy within companies. They’ve claimed they’ve doubled their EBITDA in less than top years without conducting a single layoff. And they’re finding Nexus is saving the average employee 8.3+ hours a week. Nexus is vertical agnostic so it can be deployed across the varying industries that Long Lake is looking at.

Within Long Lake is an HOA business and they’ve claimed top-performing HOA managers are able to reclaim 65 hours of work a week. They’ve completed 20+ HOA management deals. The goal is to have AI handle administrative and repetitive tasks so there’s more focus on serving clients and doing more for existing and new folks. One of the key processes that is getting sped up is handling tenant requests and vendor asks, turning vendor-sourcing into 45-60 minute emails into a 5-10 minute edit of a pre-drafted email.

Long Lake is an AI holding company that bought Amex GBT for $6.3B. Amex GBT is a software and services business for travel, expense, and meetings & events. The plan is to turbocharge existing relationships with airlines, hotels, etc and then disrupt the largely manual processes of bookings, changes, cancellations, and disruptions. As you can imagine, there is a lot of disruption and high-volume work that comes off that can be taken on by the right AI-driven processes. The deal is levered at “mid-9x” from the start (Actually lower, the agencies always over-calc leverage). The financing includes a $1.5B TLB (S+400) and $1B in secured notes.

I listened to the Alex Taubman podcast on David Senra’s channel, and the Long Lake CEO seems extremely bulled up and confident about the opportunity. They’ve also said that there’s been growing headcount, no non-performance driven layoffs, and as a result that drives a better work life environment plus stronger retention.

Regardless of what happens, this is the first, massive AI-LBO and is the case study of what happens with AI-enabled take privates from here.

And notably too, both Thrive and Long Lake are of the belief they’ll create more economic value deploying their model internally than selling it as a SaaS.

Not to pick on Greg, but his lens is from a software engineer instead of the lens of someone who talks to small business and rural business owners. That’s why a lot of LMM PE/SMB FinTwit got extremely riled up.

Everyone that’s looked at a disorganized, LMM/SMB deal knows that the amount of variables that could turn against you far exceed the casually worded risks & mitigants that might appear on an IC memo piece of paper.

Leverage is a double-edged sword: Let’s start with a semi-positive view. Jesse, self-reportedly worth several hundreds million, came to the view that Greg is right, but leverage is a double-edged sword and a game he doesn’t want to play.. Given his wealth, he seemingly doesn’t have to.

Eric Pacifici of SMB Attorney called it a good thesis, and says that as an attorney serving small business buyers, pretty much everyone is coming to market with some sort of AI driven thesis. The benefits can be very real from a reducing administrative labor and accelerating/improving sales processes lens, Still, he has some critiques:

From there, comes equipment breaking, employees leaving, odd seller behavior, unexpected capex, and constant fires.

What Eric gets to ultimately is that AI is just a tool - not the entire business. LMM PE chief John Caple said “Roll ups are easy to describe and really hard to execute! AI does not change that....”

Held together by duct tape: This is great from Brent here, the reality is a lot of these small businesses are loosely held together by the Seller/Key Man.

Extreme risks: Back in a detailed 2023 piece I called SMB acquisitions very risky and part of why was because every single risk is amplified in smaller businesses. More concentration risk among suppliers and customers, more key man or key employee risk, higher fixed costs, lower operating leverage, and less cash flow generation to manage the many, differing capital allocation needs that arise.

In rollups like accounting, there is no room for AI hallucination. Likewise with Lawyers. Use AI and mess up big time for your client? The regulatory and reputational damage might be devastating.

Jon Matzner had the best thread on this and pushed back on Greg’s thread because of the innate volatility of SMBs. There’s still key things like repair trucks, fixing a Google My Business listing, and withdrawing a salary because of a reinvestment you need to make. Matzner thinks you will have to retain people and add a bunch of AI-related expenses before recognizing any synergy.

This opportunity set is not overlooked. Multiples are going up, there is efficiency in the market: One of the mistakes Greg makes is describing the opportunity set as overlooked….Many of these businesses get mass PE or search inbound. This opportunity of buying small businesses is already well known. There’s already a ton of buyer interest. And this interest drives EV/EBITDA multiples higher. Most Sellers have already received inbound and many SMBs try to sell but are unable to because of poor financials and fundamentals.

Terrible deployment drives both customer and employee churn: I was talking to someone aware of an SMB that had a nightmare AI-enabled rollout. The SMB deployed an AI answering service for their home services-based business and it was a nightmare. It led to customers hanging up. And it led to key employee churn in the business; who then went out and tried to compete against the business. That’s the other problem with SMBs….make sure there’s a real non-compete - because jaded employees and even the Seller might try to compete against you! To caveat, this didn’t sound like an indictment of AI - just that it was deployed poorly AND maybe in the wrong part of the business? As much as companies gravitate towards AI replacing customer support, there is definitely a level where we’re all on the phone or in a chatroom and just want to reach a real agent already. This is a nightmare anecdote though, because there was revenue loss, talent loss, and remediating things AI got wrong from a job context standpoint created more busywork and callbacks.

But you know what a core critique of the silver hair tsunami is? The fact that a lot of these businesses will probably just shut down after the owner calls it quits. And that’s necessarily a market inefficiency; it’s coming from the fact that the Company is small and the Owner did all the heavy lifting, had all the relationships, and the team is kinda done as well.

It’s still a people business: Regardless of what you do with AI, at the end of the day, it’s still a people business: HVAC for example, is a simple business - people have a problem and you quote the price. But everything else in between could fall apart with bad service or a key technician leaving.

Distraction/Shiny Object Risk: Greg laid out what the typical week may look like for a one-person holdco and I think it’s good that there’s a framework that captures key KPIs like margin, human productivity, how often fixes are coming up, retention/renewal numbers, etc. It seems like there’s a framework in place iterate, learn, and make sure the wheels aren’t falling off.

But there’s a big problem and a level of naivety when you look at Thursday and Friday. Yes, that says “Next Business”. If you’re a one-man holdco, you checked in Monday-Wednesday, now it’s time to buy another business and “learn a new industry”.

That’s where things could go wrong. Chasing the next thing & adding a layer of distraction will ultimately mean something will slip through the cracks or not get the focus it deserves. This type of schedule might pied-piper business owners into thinking this is a breeze and a chill role once you have AI agents set up.

My advice to Greg would be to hire an Associate or teammate if you’re going to be adding a ton of SMB portfolio companies to the mix. Because it’s hard to imagine how bad a week could get if one of these businesses experiences even a mildly bad week.

I would say there’s two types of success stories at the SMB/LMM end so far: 1) Very small companies standing up high-margin businesses and 2) Operators leveraging AI-powered software and driving solid margin expansion.

Re: Anecdote 1 - One SMB CEO that is very tech-enabled reached out to me and says he leveraged various AI, coupled with strong understanding of Meta & Google campaigns, and the ability to spend heavily on key initial growth capex needs, has allowed them to marry a durable blue-collar service with strong digital marketing expertise. The key point - a lot of this would’ve required a team of designers, developers, meta buyers, and more folks, but AI has cut the need for a lot of this semi-expensive consulting work that would eat up the budgets of SMBs.

Re: Anecdote 2 - Service Titan might be at the forefront of this. Earlier this year, Service Titan sold off materially during the first wave of the SaaSpocolypse. But the commentary coming from the company recently has been extremely interesting. They’ve been finding that customers who are leveraging AI are capturing significant margin expansion. Some ServiceTitan Max customers are dramatically improving call-booking and job-closing.

One customer apparently saw 1,200bps margin expansion (18% to 30% margins) from leveraging the AI suite. What this resulted in was the office staff declining from seven to two and technician compensation increasing. Another customer saw invoice preparation fall from 30 minutes to less than 5 minutes, decreasing time to payment by 70% to <5 days. There were a lot more anecdotes like that in this blog. This might be one of the reasons to stay bullish software. If customers benefit at scale, then the software providers may be sticky and can increase pricing. Of course, commoditization and moat fears remain.

AI-driven improvements are probably going to have to be backend or middle-office oriented at the moment, as AI integration among forward facing customers who dislike AI might have adverse consequences. But the reality is, there’s a big “email job” component about a lot roles that should be going away. Sometimes it really feels like it takes forever to craft a thoughtful email, so being able to draw on context faster and create drafts should speed up an annoyingly long bottleneck. Sometimes I feel like my emails and written content would go faster if I just spoke to an AI and it transcribed and turned the thoughts racing into my head faster. And that’s as someone who types pretty fast. As long as it’s not a slop grenade, AI should be dramatically speeding up email communication.

Candidly though, I do think AI will result in mass layoffs over time. Pre-AI, I thought there was at least 10%-15% bloat in a lot of organizations. It’s funny, one of Logan Roy’s last scenes in Succession has him hovering over an employee typing away at a time-consuming, long-winded email and Logan jokingly telling the employee not to over-exert himself. There’s a reason so many white collar roles are called email jobs, because all a lot of people are doing is sending emails (primarily internally) and sitting in overfilled zoom meetings. Eventually, AI will further bifurcate the producers within the company and expose the people who do less or do busy work, allowing for pretty straight forward labor reductions due to a clear lack of productivity. The other obvious reason too, is that with software moats under pressure and knowledge work losing steam and value; a lot of firms will have to downsize or consolidate due to lack of competitive value prop compared to commoditized intelligence.

What’s important though, is that you can’t just slap AI on a hard business and solve it. And having a level of naivety plus a lack of operating experience may end in tears.

I didn’t mean to pick on Greg as a software engineer (there’s a lot to learn from his videos), but rather lay out how remote SaaS and gritty blue collar worlds are two vastly different things.

AI deployment seems quite real and is going okay so far; but it doesn’t change the same business and execution risks that have always existed.

That’s all for this edition.

Until next time.

-Harry

Upgrade to HYH Premium: Get access to my recruiting materials, restructuring writeups, and more. Join Hundreds of Premium readers:

Reminder, we published our State of AI in Finance report yesterday - check it out.