How private equity firm THL is winning with AI implementation

In this episode, Alex Rawlings is joined by Alex Sabel, Vice President at THL Partners, to explore how the firm is embracing AI across its operations and portfolio. Alex leads THL’s research function and AI strategy, bringing a data-driven, quant mindset to private equity.

He shares how THL uses AI internally to improve investment workflows, how portfolio companies are deploying AI to drive product innovation and efficiency, and what he’s learned from both successes and failures. Whether you’re exploring AI adoption or refining your strategy, this episode offers clear, practical insights from the front lines.

⏱ Timestamps

00:00 – Intro to Alex Sabel & THL’s AI Focus
AI at the firm, portfolio, and personal level.

00:55 – Alex’s Background
From public markets to PE, building THL’s research and AI function.

03:12 – Common Mistakes in AI Adoption
Why firms must focus on data foundations first.

04:35 – Real-Time Insight vs. Info Overload
The value of surfacing insights at the right time.

06:02 – Investing in the Global Compute Ecosystem
THL’s thesis: look beyond GPUs to second-order AI winners.

09:19 – THL’s Three-Pronged AI Strategy
Invest in AI, use it internally, and support portfolio adoption.

11:42 – How THL Supports Portfolio Innovation
Roundtables, tech summits, and a vendor ecosystem to foster AI experimentation.

13:36 – Case Studies: Binder, FourKites & Sentria
Examples of AI-driven product innovation and operational efficiency.

17:57 – When AI Fails
Why THL embraces “fail fast” and knows when to build vs. buy.

22:12 – Deciding When to Use AI
Break problems into first principles—GenAI isn’t always the answer.

25:26 – What’s Blown Alex’s Mind
AI organizing and standardizing messy, fragmented enterprise data.

27:45 – Daily Tools & Tips
Coding assistants and using multiple LLMs (ChatGPT, Claude, Gemini, Groq) for varied insights.

31:06 – What to Read
Short term: WSJ, FT, Economist.
Deep dives: SemiAnalysis and top Substacks.

32:31 – Connect with Alex Sabel
Open to collaborating on AI, compute infrastructure, and emerging vendors.

Raw Selection partners with Private Equity firms and their portfolio companies to secure exceptional executive talent. We focus on de-risking executive recruitment through meticulous search and selection processes, ensuring top-tier performance and long-term success.

🔗 Connect with Alex Rawlings on LinkedIn: https://www.linkedin.com/in/alexrawlings/
🌐 Visit Raw Selection: www.raw-selection.com

Transcript

00:00
Welcome back to the Raw Selection Private Equity Podcast. Joining us today is Alex Sable at THL Partners, a private equity firm based out of Boston  in the US. Today we are doing a deep dive into how THL and Alex are driving the initiative on artificial intelligence, both leveraging this at the private equity firm level and how they're using it,  and of course, their portfolio company.

00:28
Alex is going to talk about how he uses it personally, additionally as well, where you're going to dive into how you can leverage  AI. I have taken a ton of notes from this conversation and I would recommend you have a pen ready to do the same.  Alex, if you can share a brief insight into you, please.  Yeah, happy to and thanks for having me on,  Alex. So  my name is Alex Sable.  I am a vice president at THL Partners on our technology investing team.

00:55
And I actually lead the research effort at THL more broadly, which  I can talk about. maybe if I think about my background, I took a pretty  nontraditional path into private equity. studied finance and math in undergrad. And instead of going the traditional banking to PE route, I actually joined the buy side straight out of college,  focused in the public markets. I was at a large asset manager.

01:20
in Boston covering US equities and then alternatives for a little while on their  top down macro focused um fund to funds group. And what was really unique about that role is on the alternative side, I covered essentially everything that wasn't traditional stocks and bonds. So think, you know, currencies, commodities, options and futures. I even covered catastrophe bonds for a little bit,  which is a very obscure part of the market. And I feel like that experience really shaped

01:48
me as an investor, especially in those formidable early years, it forced me to be really data driven, combine kind of the quantitative analysis with the fundamental research  and get comfortable forming views on things that didn't necessarily have a playbook or, you know, a textbook or well-trodden rules to fall back on. And following that experience, I was really fortunate enough to join THL to build essentially a brand new research function. And like I said, I sit on our tech team, but I wear a lot of hats at the firm more broadly.

02:18
I focus mostly on emerging technology today, specifically AI, thinking about how THL invests in it, how we help portfolio companies implement it and drive value both operationally and in the product. And I would say even more so, how do we apply AI, machine learning, ultimately to do our jobs better? THL more broadly has been around for a very long time. And you can imagine we have a lot of data to work on. I think...

02:47
about me as maybe an investor, very quantitative. like going deep on topic, becoming a subject matter expert,  but ultimately using data and technology to try to  generate differentiated insight. Perfect. Thank you very much for that insight. What  are the mistakes that you see the private equity firms or portfolio companies making and what would you suggest to correct them?

03:12
There's obviously a huge focus  on  AI, data, machine learning at the private equity level. There's a ton of vendors out there that are focused on trying to solve some of these pain points. They've talked to a lot of other sponsors on what they're trying to do in the market. And uh I feel like there's a lot of commonalities among how people are approaching it. It all starts with  data and data strategy.  Where does your data sit? How are you aggregating it, et cetera?

03:38
And I feel like a lot of people maybe focus more on the outcomes first versus a lot of the foundational steps that you have to do. It's not necessarily glamorous or fun to go in and really understand, you how your data is structured or where it's coming from or how people are ultimately using it. But those are the things that I feel like THL more broadly has done a really good job at in the first couple of years  of using this technology uh really closely in the diligence process is really focused on.

04:07
You know, how do we fundamentally create a good foundation for some of the data, some of the strategy to actually build some of these tools internally and ultimately transform some of the workflows that we have on the investment side to do our jobs both more efficiently with the same resources, but also make better decisions. You know, it's obviously very competitive in the middle market where THL plays. And so a lot of these tools can help surface, you know, potential insights or

04:35
uncover things that we've either learned in the past or might be available publicly or through one of our vendors in just the right moment. I keep coming back this with some of our work is, it is really about being able to deliver the insight exactly at the right time versus you kind of have a deluge of information in a lot of these organizations. And while making all of that information searchable is really helpful in certain times.

05:03
It is, know, when you're on a specific deal or looking at a specific opportunity, being able to surface that right insight or that right workflow uh at the right time. uh That's what separates, I think, the  positive implementations from some of the ones that might not be as positive in the long run. Okay. Well, let's dive into the world of AI. think  AI is a difficult one. think, you  if we just look at the public markets at the moment, guys like Palantir.

05:33
Um, uh,  I'm going to forget guys. I can't even forgot Nvidia's name. Then I was like the company, one of largest, well, the largest revenue business in the world. Jensen Huang, just forgetting about him. Obviously I'm too big for that now.  Um, joke, obviously the, these guys are all obviously pile in their, their capital into AI and obviously doing lots of revenue shares and lots of different bits and pieces from there. What's your, what's your kind of personal take on what's happening in the AI world if you're happy to share it?

06:02
Yeah, it's really great question.  It's an area that we've spent a lot of time thinking about. If you look at some of our focus areas as a firm,  we've committed capital against some of these themes around the global compute ecosystem. We have two companies in the portfolio today, Brooks Automation, which makes some of the uh weaver handling technology, the robots that actually transfer wafers from  one area of a clean room to another, and then AMI, which actually makes some of the foundational firmware.

06:32
that actually powers on servers, laptops, embedded devices uh across the world. And we've actually spent a lot of time thinking about this thesis in global compute more broadly. And I keep  thinking about uh some of the market movements.  there was this saying of like, or the public market sometimes overestimate the short term and then underestimate  the long term.  I think that

06:57
ah When you're building a thesis in this area of the market, uh we try to take a really long-term view at it because at the end of the day, as buyout investors, you don't have the luxury of having liquidity like a lot of these public market investors. And when we make an investment,  we're going to be in it for a long time.  There's things that we want to accomplish. And then on the other side of that, you're potentially selling to someone else who doesn't have liquidity. so the things that get me really excited about maybe the global compute ecosystem more broadly is really the shift from

07:27
maybe chips to more of these integrated system level implementations, rack scale, AI, et cetera. And I feel like if you look at the public markets, you know, there's a lot of focus on some of the first order beneficiaries, GPUs, hyperscalers, model providers, know, valuations are getting bit up. They're moving very quickly. They're very well capitalized and it's really hard in the middle market to even be involved or have an edge or see anything that's actionable at that level.

07:56
And  for us, thinking about a long duration  investment cycle like infrastructure, uh we're focused on some of those players that might be second or third degree beneficiaries where they'll endure through multiple cycles. Think about the IP or the critical embedded software, things that are capital light, they're mission critical, but they'll exist  beyond the cycle that we're going through right now.

08:24
And I think where we've spent a lot of time is kind of the convergence of a few themes, you know, architecture transitions from  x86 to arm.  have, you know, custom silicon programs. have uh very fragmented  supply chains, very heterogeneous data centers, and all of these contribute to uh creating a lot of value in the IP, the tools, the materials and the services that actually design and optimize these systems. m

08:51
That's to us where I think there's going to be a lot of interest long-term and  potentially a lot of uh value and opportunities for middle market  private equity firms that are focused on more of the silicon and semiconductor uh end markets, if that makes sense. does. Thank you very much for the insights there. So let's have a look at  specifically with  THL. What are you guys doing and how are you leveraging the firm either from a sourcing and due diligence?

09:19
or at the private equity firm level when it comes to AI? Yeah, great question. And I think it's probably worth maybe doing a little on THL more broadly because it'll help tell the story of  why we're so focused on AI and data, both in the firm more broadly at the management company level than in the portfolio. So  as I mentioned, THL has been around for a while. We were actually founded in the 70s, um which  by default makes us one of the oldest private equity firms on the street.

09:49
We're investing out of three verticals,  technology, financial services, and healthcare. We're currently in the ninth flagship fund, um which is about five and a half billion. And I think the unique thing about THL is we take a really integrated approach to our investment process where we have deep sector knowledge. We call them identified sector opportunities where essentially we're building a multi-year thesis in a specific end market that we feel

10:18
uh has a long duration investment cycle to it and is very durable. And then we pair that with actually an in-house operating team that we call our strategic resources group. So think of this as an uh embedded operators that have come from C-level positions or come from strategy consultants that help us effectively execute value creation plans that we've identified during diligence according to

10:45
go to market, M &A, tech and product, talent, et cetera. And this team has uh very fast over time.  And we take a very integrated approach as you're thinking about the underwriting uh process. And so as we think about the AI strategy for THL, as I mentioned before, kind of three pronged, how to invest in it, how to use it internally, and how to use it in the portfolio. On the portfolio side,  myself  and uh

11:12
two  members of our strategic uh resources group focused on tech implementation, sort of co-lead this AI strategy uh in the portfolio. we try to, I think  the ethos or guiding principles really, how do we foster innovation at the portfolio company level to help management teams maybe identify  or prioritize some of these product and operational uh opportunities  without being so prescriptive to say, you should be doing this or you should be doing that.

11:42
And in order to do this, you know, we typically have a few angles of support. One is because we're bio investors, we have a portfolio of about 40 companies in our, um, you know, across our funds and we do, I think a relatively good job at fostering the innovation actually, uh, across companies. And so the best value that we've received from that is you have companies in completely different verticals, right? You'll have, you know, healthcare company talking to a technology company about.

12:11
the same pain point of implementing AI in their business. And that to me drives a ton of value that cross pollination. And so we'll actually host quarterly round tables. We'll do a yearly in-person tech summit focused on  workshops on use cases, implementations, new and emerging vendors. We have a list of  vendors and  third party integrators that can help our portfolio companies actually implement these things where we're providing them with the resources and the advice.

12:40
And it's kind of up to them to go and execute some of that. there's a collaboration, a level of collaboration that we typically go through with them. And then the last thing I would say is on the portfolio side is we have a number of what I would say like challenges where in 2024, the market pretty quickly identified coding assistance as low-hanging fruit for AI more broadly in portfolio companies.

13:06
What we did was challenged companies to actually implement uh these coding assistants and report back some of the metrics and some of the learnings that they had when they did this. And making it more of kind of a competition versus a uh direction really helped foster some of that innovation get buy-in. And a lot of people talk about the success, but there's a lot of change management that has to go on when you're doing something  at that scale of  hundreds, if not thousands of developers that are effectively transforming how they work.

13:36
Right. And so it was a special learning for us and we published a white paper on that uh in mid 2024 that documented some of the successes around productivity that we've seen more broadly, but we've been really impressed with some of the companies that are actually implementing some of those AI features in the product and operationally. And I'm happy  to go in depth, um you know, if you have the time on some of the companies that I feel like we've been really successful with uh in fostering some of that innovation.

14:04
Everyone in the podcast that's listening in is going, Alex, we definitely have time. Let's dive in.  So happy to document some of the examples that come to mind on companies that have implemented things really well. Like I said, we have a pretty diverse portfolio and  the interesting thing about middle market is you have companies that are up and down the maturity spectrum as it relates to technology.  each company  has a different opportunity to implement uh AI differently.

14:34
As I mentioned before, we split it between product and operationally focused and some  of the product facing implementations that I feel like have been, you're really successful and really impressed with those teams. We have a company called Binder in our tech  vertical. They are a digital asset management provider.  They're actually based in Europe as well. And they launched a suite of agents targeted at assisting uh users with end to end transformation  of

15:03
some of their digital assets. So for an example, would be  a  customer has a suite of  photos as it relates to  some of their uh marketing content. And Binder's uh enrichment agent would effectively help them  add metadata to those pictures at scale. And that's sold actually as an additional module to the traditional Binder uh GAM product that they have. And  the customer feedback has been relatively positive  on that.

15:32
uh Another company that I feel like has done a really good job in the product side is a company called Forekites,  which is also in our uh tech vertical. They are a supply chain visibility platform. uh And they launched a, again, a suite of agents that helps companies uh actually drill down into some of the supply chain visibility attributes, like, you know, trying to prevent disruptions, scheduling, uh optimizing things like yards or visibility.

16:02
ah And so you're seeing a trend in our portfolio of  how do get agents into the product? And then how do you offer that uh implementation as another skew or an add-on to the existing product to make your product even better than it was? ah On the flip side of that,  you talk about revenue generating opportunities. There's also operational improvements. And so  we have a number of companies that have actually used AI to make their businesses more efficient, ah which is obviously interesting to us as

16:32
private equity investors. There's a company in our healthcare vertical called Sentria Healthcare. And this is a really, really interesting company because they are uh applied behavior analysis for therapy for children with autism. And so you can imagine, you know, there's a large care component to their offering, which requires  a lot of human capital to actually make that engine run. And so they were doing a ton of interviews with

17:01
um people  that  could become behavioral technicians for them. And the scale at which they're actually doing these interviews was enormous. And so they were able to implement a voice agent that automated some of that high volume candidate  recruiting to actually do the interview with the prospective candidate almost immediately,  summarize it, document the outcomes, et cetera. So uh they've seen a lot of success in that.

17:28
uh super impressed with how quickly the team stood up that operation, fostered the innovation internally, and then got that out into production, into people's hands to actually use it, uh which is something we're really proud of in the portfolio. Well, something you've tried from... So  I'll be open. We've recently tried a project where  we have a lot of a manual process around  lead generation coming into the business and what we call hiring triggers.

17:57
THL raises a new fund,  comes into us, comes into an inbox.  So we tried to AI some of that coming through and then some of the  Google  spiders is the best kind of notion.  But  it turned out that it just couldn't work well enough because we're specialists, because we're niche. It just ended up pulling lots of different things in and we keyworded it through, but there's so many different press releases and  private equity loves like the same press release a million different times.  That's the point of getting the word out.

18:26
So it just kind of made a mess. What is something that you guys have had a crack at either P level or portfolio level and you've just had to pull back because either it just doesn't work, the technology is not there yet, or maybe  the juice isn't worth the squeeze. Yeah, really great question. The  short of it is there's a lot of those examples and it kind of fits the mantra  of  our internal AI strategy where

18:52
Because we've been around for a while, can imagine THL sits on just an enormous trove of information. Past deals, diligence materials, things that we've learned through portfolio companies, financials, et cetera. And all of this is documented  somewhere or another. was actually in uh file cabinets on site for a while. It became digitized oh over time, et cetera. And so we've had a lot of these use cases come through where  we've identified a pain point.

19:22
We've tried to solve it and then we've evaluated if it worked or not. And it's actually funny to think back two or three years ago when we first started really getting involved in some of these advanced analytics.  What we found,  this was two years ago, was the models just weren't good enough for the use cases.  fast forward two years, a lot of that stuff actually has changed, right? Where you have uh some of these reasoning models or models with larger context windows.

19:51
We, our mantra is kind of to fail fast with a lot of these use cases where we try something,  we get feedback from either the team here or cross-functional team across the firm, and we try to figure out if it worked or not. And then  we go back to the drawing board because I think what the special thing is about a lot of these use cases is there's a hundred ways to solve it with AI. Like if you actually go down to the architecture layer,  there's a hundred different ways you can implement the model or have the data.

20:18
available or the output organized. And so, yeah, we had a few use cases where either the model didn't work, the data wasn't ready, or the use case just wasn't mature enough. And I really think the advice that we've given to the team and some of our portfolio companies is try to fail fast and try to revisit the ones that have failed fast over time. And I think to your point, there is definitely a difference that we've seen in our business of by-verse builds where some of the things that we aspire to do

20:47
it's just better to go and find a vendor that is offering that off the shelf. we have a service internally that we've onboarded that actually monitors some of the third party publications for us, some of the press releases, et cetera. And we just get an automated email. And at one point in time, we tried to do some of this internally. to your point, the juice just wasn't worth the squeeze, right? And so we just went buy in that scenario. But I think there's benefit also to the build side of the

21:16
equation, if there is a use case that is achievable internally where you can start to build some IP or some workflow or some data assets that can be used over time, that to me, I think is where the  puck is headed.  For lack of a better analogy in the private markets of how do you actually retain more of that workflow uh to get that edge long term? So long winded way of saying like, we've definitely seen those use cases over time and  we try to revisit them. We try to fail fast. We try to instill

21:45
this feedback loop into our processes. Okay. Interesting. I appreciate  the insight. And yes,  it's a difficult one with, you know, is knowing what you've got, knowing how to prepare for that. When you, when we, if we just track back to like when you make a decision on when to AI something, yeah, again, I'll frame for us, hopefully for all the listeners that are fairly new to it, given uh your advanced expertise.

22:12
So what we've done is we've built a workflow like a customer journey workflow,  a candidate journey workflow. So we know exactly what goes in every single touch point. And then we'll begin to think what can we do here that just takes out manual nature, takes out friction, takes out problems. How is it, that's how we've approached it. How is it that you've kind of looked at that, whether it be portfolio or PFA? Yeah,  it's, it's a question that we've, we've answered a few times in our portfolio in our, in our business. There was a...

22:42
bias coming out of  early 2023, where people wanted to use large language models for every single use case. And you're kind of holding a gen AI hammer and you're looking for nails to hit it with. And people realize gen AI is really, really good uh at certain things, know, creative outputs, synthesizing data, extracting, formatting stuff, et cetera. There are some things that we've learned over time AI is really bad for, right? Because these systems are

23:12
probabilistic in nature, hallucinations are just inherent to the design. And I think people struggled with that at first, where you hear the buzzword, you see all this hype in the media and what other companies are doing, which may be 100 % honest or might just be gen-AI marketing. And I think our notion has evolved over time to say, OK, when you think about the use case, break it down into first principles. Can this be solved with something

23:40
way simpler, right? Is this a data visualization problem? Is this a machine learning problem? Is this a Gen.ai problem? And there's definitely a big decision tree that you should go through of, can this be solved in an easier manner? Because oftentimes,  we saw people trying to implement AI or Gen.ai for something that was  not Gen.ai, right? It's like you could just solve this with  a few lines of code  and someone hitting a button every day or  whatever. um

24:10
And I think the systems are  be evolving to uh accommodate that, right? So on the large language model side, because of these agent architectures that are evolving, they're actually, the whole premise of agents is that you're calling tools that can actually commit uh more  deterministic workflows, right? Where the LLM is like your orchestrator, and then the tool is like your actual executor of that.  that's like the whole premise of  why you should think about

24:39
tool beforehand, but I totally sympathize with some of the commentary there. We struggled with it at first because it's new and exciting and it's the first major tech change that we've seen in some years that really is garnering the level of hype that it is  akin to mobile or the internet. so there's a time and a place for it. The systems are getting better and um it really does require thinking about if that's the right way to solve  the problem.

25:08
so embedded in this space. What's, and it doesn't have to be at your private entity firm or the portfolio company with THL or any of the  investments. What's something you've seen that you just think, wow, that just blows my mind when it comes to artificial intelligence?

25:26
It's a really, really good question. I  saw a few things recently from one of the third parties that we're working with that  really, really surprised me with just, I would say the scale of intelligence that you can now, that's now democratized, right, in like the palm of your hand where they're effectively looking at, you know, one of our portfolio companies is trying to organize.

25:50
very, very complicated data, right? It's distributed across systems, it's across tables, it's across schemas. um It's hidden in places that only  the person that's working on that project knows where it is and effectively using AI to consolidate all of that data. A, making it consistent and standardized across all those em environments. And then B, actually explaining and architecting what the data should look like.

26:19
in the future of like, if you do want one consistent view into customer A, here's how you do it. Here's how you should join the data. Here's what this column means, what this table means, why I chose this and creating this almost ontology of all the data that relates to like this customer or this site. That was the point in time where I took a step back and said, wow, this, really is, um, you know, something that was never achievable in the past unless you threw

26:48
an enormous amount of man hours at it, know, teams of people, months, if not years of actually staring into this data and doing it. The skill at which you're able to operate, like the intelligence you're able to distribute now um is so shocking and it's only getting better. It's really, really hard in our business to keep track of everything going on because every single week there's a new model release or a new technology, a new funding round, a new company that comes up and

27:17
The way that we've approached it is you just need to be on the ground level. You need to be building with this stuff. You need to be talking to portfolio companies. You need to see what people are talking about. Because otherwise you fall behind. And if you woke up from a coma that you were in for six months, the world looks very different from some of these technologies that people are actually implementing. And I don't see the pace slowing down anytime soon.

27:45
Everyone loves a simple hack, a multi-vitamin you can pop in and feel better. The vitamin D because I live in the UK and all it does is rain. What is something that you're using that's fairly simple beyond hey, use a chat GPT prompt. What is something you just use in your everyday life that you just go, you know what guys, this just makes life a lot simpler, whether it's work, whether it's personal from an AI perspective. That's a good question. I get asked this actually.

28:14
A fair amount. the first thing that I would say, um because we do a lot of the implementation in-house, um absolute game changer is, I would say, any of the coding assistants. like  an absolute night and day difference. Pick your assistant cursor,  GitHub copilot, um any of the ones from some of the other  shops. Absolute game changers, you're looking to actually implement some of this stuff at scale.

28:42
Um, the other thing is, you know, people are maybe a little bit surprised by this, but, um, I try to use as many model providers as I can in daily life, right? Where, you know, have chat GBT on my phone. also have Claude, Jeb and I, Croc. Cause I feel like there are different personas and there's different training used, especially in the, the, you know, post-training steps, the reinforcement learning that actually will influence how these models react to some of the things about your life. And so.

29:11
I definitely have personal uh usage of this that uh is really, really helpful, just outside of the normal bounds of  OpenChat GBT. I feel like most people do that uh that are in  tech or finance, et cetera, but actually using all of them and paying for uh the pro level version,  you don't really get as much as you need from the free version. need the all you can eat one. That's like, that unlocks the.

29:39
the level of intelligence and the level of uh activity that you actually need with these models. So I definitely recommend that whatever model provider you prefer to get  the most premium one you can. Who's your favorite one? Is it ChaiGBT or is it who's your go-to? It depends on the use case, which is a bad answer. uh I do really like uh the Claude  models. uh I also do like Gemini, I Rock.

30:08
I like OpenAI, but I feel like I use them for different things. And  way that we've approached it internally is actually we try to be um tech agnostic. So  we have all the model providers available internally for people to use. And so we have a proprietary system that we use and this  system has every model provider. um And actually having that diversity of thought in the output does help, especially as you're doing stuff that's like really, really involved. um But we try to do that because

30:37
Next week, there could be another model that we want to add and we don't want to be locked into, you know, one vendor or one platform, especially as things are moving so fast. Okay. Perfect. So what do you watch, read, listen to that you recommend others should check out, Alex? um I  read a lot.  try to, um I separate it out. Obviously you need to be,  I separate between kind of short term, long term, and then like,

31:06
I guess reading for pleasure. uh More of the short-term stuff is probably what everyone is reading, whether it's the journal, Financial Times, The Economist, just to stay abreast with those sort of things. think for the uh longer-term or more strategic thinking, as it relates specifically to AI, there are surprisingly a ton of really, really great independent blogs out there where uh it's either someone's ex  account

31:34
or it's someone's substack or someone's medium or their own platform. There are so many good AI specific or infrastructure specific ones out there. One that we've found a lot of use internally and have worked with that team are the semi analysis folks. They have a publication that is just phenomenal as it relates to understanding AI and also underneath the hood what

32:01
what actually goes on behind the scenes. They do an incredible job. I feel like if you can make your way through some of their pieces, you'll be in really good shape to capitalize on some of these trends long term. Perfect. If anybody wishes to pick your brain on AI or reach out to you post this podcast, how best to do so, please. Yeah, you can reach out either via email or through the THL thread lines. We're more than happy to talk at any point if there people that are

32:31
um in this space, whether it's SEMI's global compute ecosystem or AI more broadly, uh like I said before, I feel like having these conversations are super helpful. uh And we've built up a  relatively good network around these and are always looking for people with interesting or differentiated views or  companies or vendors that um could be relevant for us as we're  continuing to trot on in the middle market here, but always open um to chat.

33:01
Perfect. Well, thank you very much for coming on to the Private Exe... Just bit, my teeth in. Thank you very much for coming on to the Private Exe podcast and sharing everything Artificial Intelligence. Yeah, thanks Alex. Really appreciate the time today. And thank you to all our listeners. Till the next time, keep smashing it.