The biggest AI opportunities may be hiding where ChatGPT and Claude reach their limits. Parag Vaish, cofounder of NextNow, joins host Greg Hawver to explain how private equity firms can move beyond off-the-shelf AI and build custom capabilities that create a real competitive edge. Drawing on a career spanning Tesla, Google, and Evolv Technology, Parag shares how NextNow approaches AI through first-principles problem-solving and a focus on measurable business outcomes. One example: a playground equipment portfolio company multiplied its leads more than tenfold from an unlikely data source. He also makes the case for why opportunity creation should come before job replacement. Tune in for a practical playbook on turning AI into real EBITDA gains.

Connect and Learn More

☑️ Parag Vaish | LinkedIn

☑️ NextNow | LinkedIn | Instagram

☑️ Greg Hawver | LinkedIn

☑️ McGuireWoods | LinkedIn | Facebook | Instagram | X

☑️ Subscribe Apple Podcasts | Spotify | Amazon Music

This podcast was recorded and is being made available by McGuireWoods for informational purposes only. By accessing this podcast, you acknowledge that McGuireWoods makes no warranty, guarantee, or representation as to the accuracy or sufficiency of the information featured in the podcast. The views, information, or opinions expressed during this podcast series are solely those of the individuals involved and do not necessarily reflect those of McGuireWoods. This podcast should not be used as a substitute for competent legal advice from a licensed professional attorney in your state and should not be construed as an offer to make or consider any investment or course of action.

Voice Over:

You are listening to Deal by Deal, a McGuireWoods podcast. Deal by Deal invites you to conversations with experienced independent sponsors and other private equity professionals. Join McGuireWoods partners, Greg Hawver and Jeff Brooker, as they explore middle market private equity M&A to provide you with timely insights and relevant takeaways.

Greg Hawver:

Hello, and welcome to Deal by Deal, a podcast for independent sponsors and other investors in the lower half of the middle market. My name's Greg Hawver. I'm a partner in the Chicago office of McGuireWoods. I focus on M&A, and I'm excited for this episode. I'm joined today by Parag Vaish of NextNow. Some of our episodes, we talk about meaty legal topics like QSBS or Hart-Scott-Rodino filings or things like that. For this episode, I'm excited because we're going to talk about AI and we're going to talk about how Parag helps private equity investors and others use AI to really take their portfolio companies to the next level.

So really excited Parag took time out of his schedule to join us today. But why don't we kick off, Parag, with just an intro to you and your super interesting career that led you to this point?

Parag Vaish:

I went into a role as the head of digital product and design for Tesla. That is a role that is as large and as stressful as it sounds. It probably shaved five years off my lifespan, but it was well worth it to get that chance to work side by side with Elon and Jon McNeill, who was the president of the company. He was my manager. The window of time was 2017 and '18 when the Model 3 was just coming out of production hell, as many people know. We were figuring out how to scale the digital assets of the company to compliment the physical development of the vehicle among solar and other product lines. After that, I went to Google. I got a very rare opportunity, which was effectively a one-line sentence job description, which was, build whatever you want for the next two years.

That creates the ability for people to get to where they're trying to go or to be able to scan people in environments where there previously wasn't the possibility of doing a TSA style prosecution. That has all led me to create NextNow. But I'll pause there, Greg, because that's the background side. That's the storied career. Happy to answer any more questions about it or go deep on some of those stories.

Greg Hawver:

Yeah, it's super interesting. When you mentioned that you were at Tesla during the era of rolling out the Model 3 and then at Google in the era of really the precursor to AI, just how interesting of moments in time to be there. I'm curious, I know the culture of big law firms and private equity investors, but maybe compare the culture of Tesla at that time to Google at that time. Were they worlds apart or lots of similarities? I know we're off on a tangent here from middle market investing.

Parag Vaish:

Yeah, it's worlds apart is the right answer. Tesla moves with incredible speed. There are oftentimes project timelines in Elon's mind that are one day or one week. There's not really other timelines. A really, really big project is one week. Most things are one day.

Greg Hawver:

I hope my clients aren't listening to this, that everything is either one day or one week. No, I'm just kidding.

Parag Vaish:

The opposite being true at Google, which takes a very methodical approach, oftentimes it verifies that what is being built does not cross lines that might be interpreted as antitrust violations or being viewed as a monopoly. Those type of things tend to slow the company down a bit. The group I was in had the ability to rapidly develop and experiment and break things. We did. It had some resemblance of, but overall Google has a little bit of a different pace. Things have changed a bit, I think in the most recent time as I keep in touch with my colleagues. It is that embracement in how liberal you can be in the use of technology is common to both.

If I were to juxtapose Tesla and Google to private equity, I would say it's on the most extreme perspective or end of the spectrum of using technology, that's Tesla and Google. Private equity oftentimes has a very conservative approach. I'm observing that difference that you're calling out, Greg, and have the spectrums in the technology world, and now I'm seeing it in the private equity side as well.

Greg Hawver:

Yeah, that's a good segue to tell us about the clients you're working with now and what you're helping them implement.

Parag Vaish:

Absolutely. We work with middle market private equity firms, and there's really two functions that we play for them. Maybe it's worth me giving a brief overview of what NextNow is in the context of what we do for them. NextNow is an AI product studio that builds discrete and specialized AI capabilities that cannot be delivered through off-the-shelf technologies. If you can do it through Claude and ChatGPT, that is not something we would build for you. We look for high value, very discreet capabilities that give you the 10X results. That's what we strive for. We focus on the metric first, verify that what we're trying to achieve is a worthwhile outcome, and then go get it with extreme creativity and extreme rate of pace.

Now, to your question, private equity firms have engaged us in two fundamental ways. There's the internal operations of a private equity firm becoming more efficient. Much of that is document-based, connecting into data and different systems, moving at the rate the world should work and move, which is revealing the right opportunities to the right people and having AI give you a remarkably high starting point. Take, for example, one where you have CIMs that are passed around quite liberally. They're verbose, sometimes hundred pages, many charts and tables. Imagine if you can have a distilled version of that, that is edited by human, verified the data is accurate and served up to you on a platter for you to do what you do next with it.

When you compare that, and some people on this call, in this podcast might hear it and say, "Well, I do that through Claude today." The answer is yes, you do. You might experience about a 95% accuracy rate. The challenge is you may not know what the 5% is that is inaccurate. Well, that's where we come in. We have developed the capability to be more than 99% accurate because of the verification, the variety of AI models being employed at the same time to assess a document. That's the one side of how we work with private equity firms, is internal operations, highly accurate, remarkable efficiency gains by connecting systems that currently don't connect together and delivering you things that you may not have thought were possible before.

The other side is the portfolio companies. EBITDA is king. We focus on top line revenue improvement or cost reduction to realize that EBITDA gain. There are so many different stories that I can describe. I'll illustrate one that's very good example of how a private equity firm can utilize AI for a remarkable advantage or a portfolio company. One of the PE firms we work with has an investment in a company that sells playground equipment, very simple business, jungle gyms, swing sets, those kind of things. They sell it to K through 12 schools across the country, but parks and rec and all those things as well. So private equity firm asked us, "What can you do for this company?"

We saw that they were generating about 80 leads per year through a very traditional marketing means. This meant conferences where superintendents or facility managers gather, and they attend those conferences and have a booth where those people walk by and you scan the badge and that's considered a marketing qualified lead, so they're getting 80 per year. We're a fresh pair of eyes on the problem. We don't know much about K through 12, let alone playground equipment. But we looked at that problem statement and found that their target audience being K through 12 schools, how might you know when they're the market besides them walking by your booth?

And we found is that after COVID, 8,000 of those 13,000 school districts stream their board meetings on different video platforms, YouTube being one of those. One click off of every YouTube video is the transcript of the video. And we take all those transcripts for all 8,000 schools and scan for questions or answers that relate to their interest in playground equipment. Something like, are they expanding their campus that might warrant playground equipment? Is there express displeasure? Are they taking out funding for? You name it, you can see how you can get to a dozen questions. And we deliver to our customer a 50-word summary whenever it's a yes, with a link, the timestamp, map it to the sales territory, and the sales rep and drop it into the CRM system of choice.

So now this is taking a company that previously was thinking myopically about how they did lead generation and saying there's a broader world and a broader set of data that is available to you for the taking. They are getting 22 leads per week delivered every Monday morning from the prior seven days of those 8,000 school district board meetings. Mathematically, that results in over a thousand leads per year. So 80 per year versus a thousand is the 10X that we were seeking that we delivered on.

Greg Hawver:

That's incredible. It's such a creative idea as well to go onto YouTube and find those meetings. A question I have is, how do you get to that creativity and to that great idea? I imagine it's a collaboration between you, maybe the PE investor and existing management, and you might have to battle against some inertia or some resistance to AI. How are you able to come up with these great ideas?

Parag Vaish:

Ultimately, the background I described at the start of this podcast has probably a lot to do with it. You mentioned working at Tesla and how is that against Google? There's a principle at Tesla around taking first principle approach to problem solving. If you apply that, if you take Elon's words about first principles, which he's using everywhere, in Starlink and other parts of SpaceX and so on. The first principle here is my customers, the school boards and where are they present, and how can you understand their interest and intent? That leads to then the solution or leads to the opportunity to find the solution.

If you take that approach, which is well written about in many books and folks have spoken about first principles, and you truly look at the right metric to go solve for at a company, these opportunities are sitting there for the taking. It takes a little bit of that breadth of understanding of different industries and the drivers of a business to then reveal those opportunities. That's where we propose things, run experiments very quickly, show results. Private equity recipients or their portfolios, they are oftentimes just wowed by the output of what's possible. That's where a little bit of the back and forth then results in the improvement of the idea.

Greg Hawver:

That's really interesting. Taking a step way back, when I think about why I like to work in what I call the lower half of the middle market where my clients are buying businesses typically from founders and really taking them to the next level, it's because there's real progress and real changes that my clients can make day one after closing. It's funny, a decade ago it related to professionalizing the management team, accounting books and records and sales, et cetera. Now it just seems like you have a whole nother tool that should just be a almost standard practice day one, how can we improve this in the business we just bought?

It's obviously going to vary business to business, but if I'm thinking about as an investor, my playbook, what are some broad themes that you see that can be improved with AI that aren't quite as specific as the school board example, but just general concepts AI can improve with across businesses generally?

Parag Vaish:

Yeah, I think the income statement is a very good place to start, and that's where you can rank your top expenses and you can, as an approach, look at each line item. We ask the question very directly, if this number on the expense side of the income statement is 50% lower than what it is today, would it be worth it to you? That's a very good approach to then finding cost reduction opportunities to then say, okay, yes, it's worth it at half that number because it changes your margins, you might become a rule of 40 company, whatever the result is. But you've now focused on the right area to then reveal the opportunities. Same on the revenue side.

That's one approach you can take. Another approach is you look at a business and say, where is their time spent by human capital, by labor? That could be in the form of processing deals or scanning, understanding RFPs, responding to RFPs, whatever it is. But if you see where you're putting people time and anything that those people are doing that are related to documents or responses, meetings, there's something there that can be had that can improve it, which is either, and I like to favor the idea of if you put 10X more leads or improvements there, you'll get an improvement in your business. You could also realize some potential cost savings, or scale.

You have five people doing this and they're supporting 25 customers. Well, how can those five people support a hundred customers in the presence of AI? Those are the philosophies and principles that I would use, but I think the income statement's a very good place to start. One thing to add on, Greg, is that as a private equity firm makes an investment from taking a founder-led business and then having ownership over it, there's also the flip side, which is what is the AI risk to this business? Maybe you don't make the investment because you spot this chance that this business could be fundamentally disrupted in the presence of AI.

That is another way to use capabilities and skills of AI to make those assessments where that downside risk is. Hopefully these are a bunch of different ways to get to the answer.

Greg Hawver:

You're having conversations even with potential buyers that might be looking at an industry and saying, "Is there an existential risk to this industry before I make the buy?"

Parag Vaish:

That's absolutely right. It's both sides, and maybe there's an opportunity as well because you see that they're doing 22% margin, but in the presence of AI, you can get up to 38%. You can do all this analysis before making the investment, be much more informed.

Greg Hawver:

Right. I'd love to get your view on it. This is a little bit of a touchy subject, so maybe we can cut this out if we need to. But the fear was a couple years ago that AI is going to eradicate huge swaths of jobs, white collar jobs, entry jobs, et cetera. I think that maybe the data is more of a mixed message, and at law firms, I think we're one potential area where AI was going to come and take all of everyone's job and my job included, and we're not seeing that at the moment. When you're implementing AI with portcos, how do you see total job replacement versus if I have eight things that are part of my job, AI is doing three of them or five of them really well and I'm still doing the other five to eight, et cetera?

Parag Vaish:

The way I think about it is opportunity creation before I think about job replacement. The data supports that there's tremendous amount of opportunity creation and new jobs and roles being created today and new companies and individuals who are becoming founders that previously didn't think it was possible. There's all kinds of growth stories there. But specific to a portfolio company, the essence of my message here is, if you could do more with the same, then the business is healthy. The reason I say the same is that the people that you have that in your question are the ones at risk of being sustained in their role, those are the ones who are effectively training the AI to do what it is is the outcome we're trying to achieve.

If those people feel like they're going to be replaced, human nature does kick in at times where they try to disrupt the project from being successful. It's upon the founder and the owner of the business to ensure that those people are assisting the AI to do the work such that you can get 10X more opportunities in front of them or more business, or they will benefit from doing so, because the project or the execution of that technology will just go better, versus if you have folks who are constantly worried, they'll find ways to sabotage it, and things don't go as smoothly as you'd expect.

Greg Hawver:

That makes sense. We're looking at it similar way here at McGuireWoods in that it's increasing the opportunity, it's making us faster and better at certain aspects of our job, which frees us up to take on even more work and expand the pie, really. To me, I've heard the term jagged frontier as it relates to AI sometimes. There are certain things that it does really well and certain things that you think it might, but it actually doesn't do a great job as far as, again, thinking about the menu of 20 things that I do as a lawyer on a daily basis or an engineer does on a daily basis. I guess a question is, what has surprised you as far as an AI function where it just did fantastic and maybe what's been a disappointment where you thought it might do an incredible job, but actually it was a bit disappointing?

Parag Vaish:

Inherent in your question is an interesting thing that I think a lot of people are not able to get over. They'll go and use Claude or ChatGPT to try to achieve something, and it might be a miss or doesn't deliver what they're expecting and they stop there and say, "Well, AI is not there yet and that's a disappointment." That's precisely where we start NextNow. The capabilities of ChatGPT and Claude are phenomenal and they should widely be utilized by private equity firms, no doubt about it. There are limitations as well, and that limitation spot is where the remarkable opportunity is. I would ask everyone to look at it and say, "Well, it didn't do what I wanted it to do." The most likely outcome or reason why is that you needed multiple models to be employed to be able to do the end result you're trying to get to.

That's what we construct, are the series of models. In the pleasant surprise category, I'll describe one story where we had a portfolio company of private equity, they are rather tech savvy, utilize Claude, pounded on it many times over to try to get the ability to analyze blueprints. This is blueprints of a new construction building, and they were trying to figure out where their product would be deployed in these new buildings. It's a multi-day, call it 10-day process to analyze those under the current capabilities of their people. They tried Claude and it didn't work for them.

Greg Hawver:

Yeah. Just a little bit more detail. Claude, we're talking they were using a large language model and they were just using it off the shelf, that subscription. You come into a company, you're building a harness on top of an LLM, or maybe just go into a little bit more tech details for those that are interested in it.

Parag Vaish:

Sure. You're right, the description is off the shelf. You open Claude, you pay $20 a month or some capacity level. Essentially, what they were trying to do is, like I said, analyze blueprints, new construction buildings for where their product would be deployed. You can imagine that scenario, you upload a PDF, you type out a series of questions, you say, "Where are these 25 keywords present in this document?" That's the scenario. It did not deliver what they are presently doing using human talent to assess those documents. Now you have a gap and you say, "Well, AI is not there yet." The reality is that there are different models from the large language model companies that do certain things better than others, and so you don't know that.

It's a very hidden behind the curtain type of a thing as to which model does what really well. Just as a simplified example, certain models can scan through text incredibly well, but those models may not do image evaluation very well. They don't scan through the image of the diagram that's on the same page. Then vice versa, some will analyze images incredibly well, but they'll miss on text. As a novice person in private equity, you most likely don't know which model does what well, and you're certainly not going to jump to the stage where you're going to have one model do something really well and then flip to the other one, do the other thing really well and combine the two outputs at the end.

That's a bit too much effort and very technically savvy people might be able to do it. What we do, as you said, like a harness, we'll look at what model does what well and employ those in the scenario to achieve the end output. In this particular example, the ease with which we can evaluate those blueprints and do what their current present team does represented a 10X decrease in the rate at which they can evaluate an opportunity and figure out if they're going to bid on that project at all. We're able to analyze both the text and the images and get to that end result. We say the pleasant surprise is your initial question versus the what didn't achieve meet the mark. I now am a believer that anything is possible. I truly believe that anything is possible.

We limited our creativity of what to go after, and the way that you get to the creativity is by getting to the number that you're trying to influence. It's combining two of your questions there of, what are the mechanics? Then I do believe the range is nearly unlimited if you can understand that these models do different things well, and then work with somebody who can do that.

Greg Hawver:

No, that's fantastic. Look, Parag, I really appreciate your time today. I work with and think about AI quite a bit, at least for an M&A lawyer, and so this has been really fun. But before we go though, for our listeners who are primarily, again, buyers of middle market companies and operators of them post-closing, anything we didn't cover today that is critical as they think about AI and how they're going to use it over the next couple of years?

Parag Vaish:

Ultimately, the AI strategy of a company has to be more than using the off-the-shelf tools, because the off-the-shelf tools will give a company the ability to meet parity with their competitors. What I mean by that is, one day Claude and ChatGPT will be just as ubiquitous as Excel and Word and PowerPoint and Outlook and so on. And no one thinks of those four things as a competitive advantage these days. And so the ability to use Claude and ChatGPT are becoming ubiquitous. You might think you do it better than others, but you'll never know. You use Outlook better than others. You never know. So the world is broader than what those off-the-shelf tools offer, in particular in private equity, because you're opportunistic to seize improvement on EBITDA and to get a return on your investment as quickly as possible.

The way you get outsized returns and differentiate from the competitor companies in that space is by doing unique custom AI executions that you cannot do from the off-the-shelf tools. That's where differentiation is created, that's where improvement in EBITDA happens, and you get the end result that you're seeking in the trade that you're in.

Greg Hawver:

Such a good point. Well, thanks again, Parag. Where can people find you and NextNow and learn more?

Parag Vaish:

Nextnow.ai, the phrase there is we are employing the next technologies now on your behalf. LinkedIn is a great way. Parag Vaish. Our website has all of our details. We have case studies and video demonstrations of things we've built that can exhibit more than what I've shared here, Greg, or my email address, [email protected].

Greg Hawver:

Well, great. This is really fun. Thanks again, Parag.

Parag Vaish:

Of course. Thank you.

Voice Over:

Thank you for joining us on this episode of Deal by Deal, a McGuireWoods podcast. To learn more about today's discussion and our commitment to the independent sponsor community, please visit our website at mcguirewoods.com. We look forward to hearing from you. This podcast was recorded and is being made available by McGuireWoods for informational purposes only. By accessing this podcast, you acknowledge that McGuireWoods makes no warranty, guarantee, or representation as to the accuracy or sufficiency of the information featured in the podcast. The views, information, or opinions expressed during this podcast series are solely those of the individuals involved and do not necessarily reflect those of McGuireWoods.

This podcast should not be used as a substitute for competent legal advice from a licensed professional attorney in your state and should not be construed as an offer to make or consider any investment or course of action.