In short
How AI creates value in the US lower middle market (sub-$100M revenue) and what it means for private equity—especially the gap between GenAI pilots and production, the role of “human-in-the-loop,” and how deal sourcing/portfolio value creation will evolve.
Guests (backgrounds)
- Matt Fitzpatrick, CEO of Invisible Technologies. Former private equity investor/entrepreneur; led AI capability at McKinsey’s QuantumBlack Labs; now runs Invisible, which builds AI systems using a data platform plus an expert marketplace.
- Mark Porat, “visionary behind General Magic,” later worked across telecom and foundational tech (intelligent agents, cloud, iOS/Android-era ideas). Now Chief AI Advisor at Access Holdings.
Key claims
- Most AI projects don’t reach production (only ~5–8% succeed); lower middle market failures stem from unclear ownership and lack of “process scaffolding.”
- Human validation remains essential for regulated/critical decisions; humans become more important in oversaturated sales channels.
- Deal sourcing info will democratize; alpha shifts to process/data integration and operational/transformational playbooks.
Notable examples
- Swiss Gear inventory forecasting: Invisible combined ~270 data tables to improve inventory accuracy by ~60%.
- Supplier onboarding: AI must track process stages (emails/documents) to establish baselines and enable validation.
- Pet-care call handling: AI answers/daisy-captures leads into CRM to prevent missed calls.
Written by AI. May contain mistakes. Listen to the episode to check what was said.
Chapters
Tap a time to open that second in VOThe Impact of AI on the Lower Middle Market
0:25 to 1:44
Discussion on how AI is transforming the lower middle market and the potential it holds.
“Today, we're stepping into the next frontier.”
Introducing Matt Fitzpatrick
1:44 to 2:43
Introduction to guest Matt Fitzpatrick and his background in AI and business.
“He went over to the dark side to professional services and consulting, leading the mighty McKinsey's AI capability in quantum black labs.”
Invisible Technologies: Company Overview
2:43 to 4:27
Matt discusses the operations and expertise of Invisible Technologies.
“Let's start with a little bit of quick fire on who Invisible is.”
Challenges and Opportunities in AI Implementation
4:27 to 11:00
Matt explains the challenges firms face when operationalizing AI and shares insights on overcoming them.
“And then while I was McKinsey over the course of a decade, I was one of the early builders of a lot of their engineering infrastructure.”
Case Study: Swiss Gear's Inventory Forecasting
11:00 to 12:53
Discussion on how Invisible helped Swiss Gear improve inventory forecasting and accuracy.
“In the context of things not working, let's talk about some examples of things working.”
The Importance of Human Expertise in AI
12:53 to 14:01
Exploration of why combining human expertise with AI is crucial for success in private equity.
“And that fits exactly in the sweet spot.”
Human Process and AI Integration
14:01 to 17:20
Learn how understanding human processes is crucial for effective AI implementation.
“People don't actually understand the current baseline.”
AI in Customer Service Dynamics
17:21 to 19:16
Explore how AI can improve customer service while preserving human connections.
“And so most of the people in that store are doing exactly what they're saying.”
Future of Deal Sourcing with AI
19:17 to 22:33
Understand the potential future impacts of AI on sourcing and deal processes.
“And we're going to soon turn our attention to Mark, who's the ultimate visionary.”
Operational Efficiency in Private Equity
22:34 to 24:14
Discover how AI can drive operational efficiencies in private equity firms.
“And I think the middle market more than ever will be where you'll see real bifurcation there.”
Show all 22 chapters
Leveraging AI for Investment Professionals
24:15 to 28:06
Learn how AI can transform the way investment professionals gather and analyze information.
“So let's close out with a couple of tips.”
The Evolving Landscape of Software Engineering
28:06 to 29:48
Explore how modern engineering is democratizing software development in business.
“Even to get a baseline level would probably take you years.”
The Future of AI: Possibilities and Responsibilities
31:09 to 36:27
Dive into the potential of AI to revolutionize human capabilities and its implications.
“And with it, the capability that's changing our lives.”
Balancing Utopia and Dystopia in AI Development
36:27 to 42:04
Examine the contrasting visions of AI's impact on society, both positive and negative.
“And it'll be able to solve equations, for example, that make fusion happen.”
The Dual Nature of AI's Promise and Threat
42:04 to 43:34
Explore the dual potential of AI to create both opportunities and challenges.
“The extinction thesis is brilliantly laid out, and when you read it, it does not sound like wacko science fiction.”
Preparing for AI Disruption in Business
43:34 to 45:56
Learn how private equity and business leaders can prepare for AI changes.
“which after some of what we've talked about feels somewhat trivial.”
Integrating AI into Personal and Professional Life
45:56 to 48:33
Understand the importance of integrating AI into personal and business practices.
“records, you've opened up your real-time communication, you know, this conversation we're having today, you know, just take the MP4 and drop it in and say, make that part of me.”
Competitive Advantages in an AI-Driven Market
48:33 to 53:14
Discover how creativity and strategic thinking become essential for competitive edge.
“That's basically what it, and that reduces operating expense, which increases margin, increased satisfaction for the customer, and everything's good.”
Reflections on Inspiration and Impact
53:14 to 56:00
Engage in a personal discussion about the influence of documentaries and visionary leadership.
“I first heard of you when I watched the General Magic trailer.”
The Legacy of General Magic
56:00 to 58:48
Discover the significant impact of General Magic and its innovators on modern smartphones.
“And the others were very, very young, really young people.”
Dinner with Internet Pioneers
58:48 to 1:01:26
Explore the insights shared during a dinner with Tim Berners-Lee and the reflections on the web's creation.
“I was going to ask you, if you could have dinner with anyone, who would it be with?”
The Future of Technology and Discontinuity
1:01:26 to 1:02:52
Learn about the current state of technology and its potential for future exponential growth.
“And people who saw the world after, dot, dot, dot, created trillions of dollars of value.”
Transcript
Automatic transcript. May contain errors.0:00This is Why We Like It, a quick dive into the industries capturing our attention in the lower middle market in the United States. We're asking the questions that unlock why and where we see value in the lower middle market. I'm Sam, and each episode I'll share why our sector stands out. We're going to bring in the voices who know it best, and we'll point to where we think the future's moving. Welcome to Why We Like It. Today, we're stepping into the next frontier. Where AI is, it's not just transforming how we work, it's totally redefining who wins in the workplace. The lower middle market in the United States contains over 30 million companies.
0:42That's 99.9 % of all businesses in North America. And that's a huge amount of efficiency that AI could help with. And the private equity players know it. But it's not that easy. 91 % of middle market firms are already using Gen.AI. Half say it's saving them hours and hours in analytics. It's saving them time in deal flow. It's saving them time in IT. And many people say that it's cutting diligence costs. 70 % is the number thrown out there. But behind the numbers, the truth is so much starker. 92 % hit major challenges. 70 % had to bring in outside help. And only a quarter professed to have really integrated AI.
1:29So the stakes are high. AI can now surface nearly 200 targets, companies to invest in in the time it takes a human to find one. If you're not adapting, you're probably already behind. Today I'm joined by two people, two friends, who've seen what happens when technology meets revolution. matt fitzpatrick ceo of invisible and the iconic mark porat the visionary behind general magic let's get into it matt fitzpatrick's done it all he's been an investor uh early on in his career in private equity he's been an innovator and an entrepreneur building a company that he sold to Alibaba. He went over to the dark side to professional services and consulting, leading the mighty McKinsey's AI capability in quantum black labs.
2:27And now he's the CEO of arguably the most exciting AI company on the planet, Invisible Technologies. Matt, thank you for spending time. When you're with me, you're not with clients. I appreciate it. And And yeah, it's great to see you. Thank you for having me. Let's start with a little bit of quick fire on who Invisible is. We're going to intersperse them with some questions about Matt. I won't give him much time to react. So hopefully we get all sorts of honest truths out of this moment. Matt, when was Invisible founded? I think we had our 10-year anniversary recently. So quite a lot. HQ? uh we don't have a formal hq uh we have now offices across new york san francisco london dc uh paris holland they're in the process of opening austin and seattle as well awesome i'm so glad you didn't say decentralized um employees uh we're at about 450 now and is there a broader employee bench that you have yeah so one of our core platforms is our expert marketplace so um think of this as we have the you know our formal full-time team and then we have a lot of our our core business historically has been the fine-tuning and training of uh models for all the large language model builders and so to serve that we have what's called our expert marketplace where we bring in experts on all different really obscure topics on earth whether that's phds in computational biology or physics or tax language experts in French, as an example.
4:03And so rough math on the marketplace, we have about 20 ,000 active agents as of now, about 750 ,000 applications per year to get into the marketplace. And so we're adding about a thousand different contractors per week on different topics. So you can think of that as the enormous pool of experts that support all the AI we do that do that very flexibly on whatever expertise set you could imagine. three words to describe invisible data expertise process and the reason i say those three is you think about the core of what we do uh we have neuron which is our data platform that brings together structured unstructured data uh the expert marketplace i just mentioned which brings expertise on a topic on earth and then uh on the back of that we build lots of different custom software and then the fine tuning of agents and models which we do via our platform is atomic and axon and so what we end up is by piecing together human expertise and fragmented data we can build very custom different applications for individual institutions awesome let's get to know matt a little bit matt what's home for you uh new york trebecca awesome also glad you didn't say decentralized how many companies have you started um years ago i i co-founded an e-commerce uh retail in Southeast Asia.
5:19And then while I was McKinsey over the course of a decade, I was one of the early builders of a lot of their engineering infrastructure. So when I started the firm, I had less than 100 engineers by the time I left, we had about 7 ,000. And I built and started all of the engineering infrastructure and financial services and expanded to lead Quantum Black Labs, which is the firm's global tech development group. That makes perfect sense while you're in the seat. How many languages do you speak? one just english french horrendously but i wouldn't even count i'll go with one makes me feel better um what is matt's favorite food in case someone's trying to impress him
6:00fish okay good to know that one's in my back pocket for when i need to need to win you over um let's get stuck into it we'll uh we'll go to the the regular way questions now let's start with your background tell us a little bit about how you got to invisible um it's pretty unusual for anyone to to end up running a company of this exciting nature having not founded it uh but here we are how did that happen so i started doing ai in the enterprise somewhere around 10 years ago um in the early days i think by the time i left mckenzie had about 7 000 engineers and I ever saw about a thousand of that.
6:41When I started at the firm, probably less than 100 engineers total. And so, you know, it was somewhere around 2015, started to realize that the consulting profession was going to evolve quite a bit of what we need to do is actually technology delivery. So starting to hire a team in the early days that was doing everything from like data warehouse transformation work to custom application builds of software. Then eventually, I would say machine learning in the early days, different type of AI. And that process, we scaled very materially over the course of the last decade. And so I think the interesting dynamic as I got to know Francis, the founder of the last three years, was that I had spent as much time as probably anyone in the last decade getting to know all the problems, challenges, and ways to be successful in building AI in the enterprise and in the middle market, by the way.
7:31And Invisible had reached its incredible threshold largely as a mix of kind of AI training of the model builders in the beginnings of an enterprise business. And the desire to continue that and fully build out the enterprise business, I became, I think, a uniquely suited person he came to add. When did you meet Francis? I met Francis at an organization called Dialogue. It's kind of a discussion group where we talk about all different random topics, different philosophy to military history. And Francis and I bonded over a whole host of different topics that were completely non-work-related about three years before I took it.
8:06Amazing. I love that. Shows the power of community. And you and I, we're going to be in the mountains in Davos soon. It's another place where those kind of collisions happen. You've said many times that AI projects that get underway actually never make it into production. And in the lower middle market context, let's categorize that as firms under 100 million of revenue. that also happens a lot but what are some of the most common pitfalls that you see when firms try to operationalize AI yeah so the stat i've referenced is somewhere between depending on the stat you look at five and eight percent of AI projects make the production right now so the vast majority are effectively prototype science project experimentation that don't don't reach production and i think the challenges the enterprise and the middle market face are different.
9:00I think in the enterprise case, the big challenge is you usually have really fragmented, large existing application footprints. So you might have 60 ERP systems and HR database. You have all this fragmented legacy systems footprint, which means your data as a consequence of that is very fragmented. And so most of the challenges in the enterprise context has been, how do I bring all that data together? How do I realign and redesign my processes to be clean sheet, to be to optimally use AI? And then I think most importantly, how do I test and validate, statistically validate Gen AI, which is a lot harder than machine learning.
9:37So if you take an example in a machine learning model, like a pricing model, you can backtest, you know, an ML model versus a traditional price being set by a human. And you can say, how do I prove that I would have been more accurate over a six month basis if I'd use an ML model? And you can prove that out. Gen AI is much harder to do that. So if you think about something like generating a supplier onboarding agreement using text generation. That's actually harder to say. What does good look like? How do I validate it? So the enterprise has had this suite of challenges around, I would say, fragmented systems, data, business process redesign, and the definition of good.
10:11The lower middle market, I think, has actually been a different set of challenges. Usually an organization like that will have, at most, one IT person, a very lean technology function. And so they just have had a struggle to find who should own transformations they're going to make. Meaning, if you're a leanly staffed organization that does one or two things really well, it's mostly kind of business executives running it, you're not necessarily configured to set up and implement AI. And so you end up trying to use a lot of different tools. The problem is, and I think this is something we say often, AI is not set.
10:45So it's hard to go, you know, buy a tool off the shelf that will actually do use Gen.AI in a way that's helpful. And so the process of how do I implement that? How do I resource it? Who leads it? Has been, I think, the bigger challenge in the low and middle market. In the context of things not working, let's talk about some examples of things working. And I love Invisible's branding. I love the name. It's so powerful. it points to AI working seamlessly in the background. Can you share an example of where Invisible has quietly transformed a business process in a way that would really resonate with a private equity investor and audience?
11:24Yeah. So I'll give one recent example. Swiss Gear, it's like Swiss Army, the luggage brand. We worked in the U.S. around inventory forecasting. And this is a problem that many mid-market businesses struggle with, which is six months, let's say I have a long lead time in my inventory and employment might be three months, might be six months. I need to be able to accurately forecast revenue for a pretty large mix of SKUs. And so my challenge is, let's say I place 50 orders for those SKUs. In some cases, I undershoot the forecast, meaning I think revenues will be higher and ends up lower. In some cases, I think it's a little bit lower and end up higher.
11:57And you end up with a lot of different challenges as a result of that. You either miss out on revenue because you don't have enough inventory, or you overorder inventory and end up with a lot of excess stock. And so that challenge is probably something that vast majority of businesses in America struggle with. And the solution to it is not just a basic ML model. It's starting to gather lots and lots of different data across social media, listening, understanding what inputs you can make to improve the forecast, how you can validate the forecast, how you can display it in a way the business user can look at it.
12:30And so I think the exciting output for us was over a course of a couple of months, we were able to bring together, I think it was 270 different data tables to bring that forecast together and improve their inventory accuracy by about 60%. Wow. Yeah, it's outcomes like that that really moved the dial for asset owners. And at Access, we spend a lot of time thinking through what do our set initiatives for B2B or B2C services businesses look like in our organic vertical. And that fits exactly in the sweet spot. We spent an hour earlier today talking through a very similar initiative within our value creation division.
13:15Let's talk for a moment about exactly that, human beings being in the loop. Invisible is a massive advocate for not replacing humans, but combining humans with human expertise with AI capability. What do you think the human in the loop is such an important element for firms in private agency? Yeah, look, I mean, I think the first challenge I think the vast majority of middle market companies face this in some form is you often don't have a good baseline for how accurate, costly, timely a process is today. So if you take something like supplier onboarding is another area we spend a lot of time.
13:54If I ask most companies, like, what is your supplier onboarding timeline? How many people spend time on it? Where do you have accuracy issues? People don't actually understand the current baseline. And you actually can't build a current baseline unless you actually are able to build process scaffolding around the current humans doing the work. So like tracking emails, looking at the individual stages of a process like onboarding, which documents are consistent. So I think the first thing, the first part of this that's important is you actually need to know how the human process works today. That's one of the most painful parts of it.
14:27But I think once you've started to add AI throughout that process, there's a couple reasons humans are really essential. One is validation. I think that, you know, the current media obsession with like how often AGI will determine all of this stuff, I do think in the enterprise context is overhyped in some ways. And that even if you have a situation where an AI agent can make that decision, particularly for things like regulated entities, payments, etc., you're going to want human validations. Like I can think of, I built years ago, the multifamily appraisal engine for a real estate player that had to do real estate pricing.
15:04And for a whole host of reasons, the regulated entities, they absolutely needed a human to check off in every single case. This is the end appraisal. This is what the human says. And so the question is, how do you give that human the ability to do that a lot faster, more efficiently, more accurate? But you're always going to have a human one. I think the second part of it is the vast majority of functional work, humans are still going to be better at for a long time. Like as an example, sales. Everyone will currently tell you that there's a bunch of AI sales agents that are going to play sense.
15:31Completely not true. I mean, simplest reason for any mid-cap business, every mid-cap business that's sending 50 different email chat agents and 50 different LinkedIn chat agents has already saturated those channels massively. So every out-of-the-box sales agent, everyone's now moving to like, what's a human connection that might make me interested in talking to a seller? So I actually think in the age of oversaturated email inboxes and oversaturated LinkedIn inboxes, human sellers become more important than ever. But the question is, how do I give them a lot more information to make decisions, to do the sales process in an informed way, to think about pricing consistently across what they do?
16:06And so I actually think for the vast majority of these processes, the human in the loop will be incredibly important. You'll just take out a lot of the data gathering process that the current humans are wasting a lot of time on. The human in the loop part, It's what keeps it engaging and human, ultimately. I think it's very difficult today as we go through various different hype cycles for a human to buy from AI. But a good example of that would be a pet care business. The store team are running around looking after the pets and caring for them. And the calls are coming into the front of the house and they're missing calls.
16:51no one's going to call up a pet care business get dealt with by an ai call center and hand their dog over uh to to a machine but but it does prevent you uh it allows you to do daisy capture it prevents you from missing the lead and it puts it into to the crm and into the the system yeah in my mind that's a really good example so so let's take i i don't work with any pet companies but i'll take that as an interesting example of how i would think about that from an standpoint. Most retailers like a pet care business have some amount of revenue per store that's generated. They have some fixed costs.
17:28And so most of the people in that store are doing exactly what they're saying. They're working with pets. They're working with people. They're doing very human-centric things, which is the right decision for how to spend time. They don't have the capacity to staff 25 people on a help desk to resolve calls everyone that calls in. They don't have enough people to field every question or ongoing. And so So what ends up happening, right, is that a lot of what suffers is kind of the after service part of this. So like, you know, if you call your cable company, you might end up on hold for quite a bit of time, as an example, right?
17:58And so in my mind, a lot of what AI does is actually it doesn't change any of that human element. It should allow you to not be waiting on hold, to not be waiting for answers, to get the information you need in a much more intelligent way. And a lot of that relies on for that to be effective. that help desk has to have all the information about you as a customer all the prior interaction history all the conversational data it's a perfect use case for gen ai it's just one that is hard to implement effectively and i think that's why folks struggle with it yeah and and to be a material upgrade from the press one if you want press two if you want press three uh yeah i mean there is no uh no segment of society that people are more unhappy with on an MPS basis than contact center.
18:41So we are, you know, we're entering the contact center space. We've, we're pretty excited with some stuff we're doing there around sentiment analysis, around proving that if you think of any series of calls, let's say you have 10 different topics somebody might call about, you know, there's five of those that actually an AI agent can provide a superior experience for with way more, with way better informational access, train more consistently. And that's why like right now, people I think generally are very frustrated from a customer experience standpoint about how much time they spend waiting on calls or waiting on hold for contact centers.
19:14And I think that's one of the things I can fix. So let's do a bit of stargazing for a moment. And we're going to soon turn our attention to Mark, who's the ultimate visionary. So you've got your work cut out. But if we imagine the lower middle market in 10 years' time, and maybe we hit AGI, superintelligence, who knows? What do you think is one surprising way AI could fundamentally change the way firms source deals or create value? I'm going to unfortunately take the counter to that, which is I know that the current conventional wisdom is that deal sourcing will be majorly changed. I think you will have better information on the set of companies available.
20:03My unfortunate view on that, though, is that information will become broadly publicly available. Because where I think about competitive advantage versus broad market access is any data that can be built up from third-party sources, somebody will come along and they will democratize access to that. And so I think there will be a series of companies that will spin up and they will build private company repositories like the Capital IQ or anything else that people will have broad access to. And so I'll go back to, I actually think in that case, it will all be a human networking will determine the access.
20:33I think it will be better information and the same human networking, and none of that will change at all. And look, I understand there will be parts of the deal process that I do think become a lot more efficient. I think investment committee memos, 70 % of that can probably be auto-generated. I think probably that whole exercise is a lot of paperwork that is pretty repeatable. Like credit memos for commercial credit are another one where we just generate a lot of paperwork right now as a society that I just don't think needs to be generated for the core sets of theses you actually need. So I think the actual core alpha generation from sourcing will still be entirely human.
21:09I think the investment memo generation that supports that will have a lot of GNI in it. I do think some of the stuff like the ability to build a financial forecast in Excel using GNI, I think you're going to see some really interesting kind of text to forecast work that will happen. Like I think it's very possible that two years from now, you'll be able to say give me six percent revenue growth uh margin expansion by two basis points and a reduction in opex by one by you know five percent and excel model will just be spit up to do that from text like i think that sort of thing will be material in the way you resource firms but i actually think the core functionality of what do i source how do i think about what to invest in won't change it's all of the again the the data gathering and process scaffolding i think will change materially.
21:58So it's an industry question then of like what creates the alpha generation, I think probably pretty similar things. I think real estate is a very similar example of this. Like the core function of what properties do I buy and how do I build relationships to do that will stay the same. The process of like how do I gather information from those properties? How do I bring that together and look at a list of master assets gets way more efficient. Like right now, that's way too clunky and manual. And so it's the support functions that get easier. I do think, though, the area that gets much more interesting is the way the portfolio companies evolve.
22:32And I think you are going to see a really big bifurcation in the firms that really embrace this versus the firms that kind of talk about it but don't. And I think the middle market more than ever will be where you'll see real bifurcation there. I think that right now, across the landscape, I think a lot of private firms are dipping a toe into this. They're interested in it. They're doing pilots, but I think a lot have struggled to figure out, like, what are the four or five, you know, things I do repeatedly to create value, like, you know, selling motions or contact centers. And I think a couple of firms will figure that out.
23:05And so I think the interesting thing, if you thought about an efficient market where, you know, information becomes more readily available, then you have some networking and then you have a bidding price. if some firms are able to bid to a price that's a little bit different, like if I know that I can add, you know, a couple points of revenue based on better contact center operations, you're going to see some firms actually be able to bid based on operational alpha. I think that has not been the case society-wide to date, but I think you will see that over the next 10 years. Yeah, I guess it's no different assessing synergies as a strategy.
23:40yeah yeah i think i think that's the case and i think you'll see private equity firms that develop playbooks of things they do repeatedly and you know you've seen some of this already in some sector based you know like if you're a consumer focused focus fund you probably have a strategy of how you add value in e-commerce optimization for example um i think you will see that with ai but it's been interesting to me the variance in uh level of interest i would say there are some firms that i can tell i do believe will succeed just the amount of time resourcing and focus they're committing. And then there are some that are kind of like, well, this is what we do.
24:10We're having to focus on it. And I think that'll be interesting to see in the next couple of years. So let's close out with a couple of tips. I'm going to take you back to the earliest stages of your career. Imagine you were an investment professional today in private equity. What is the one way, if you could only use it one way, you would use AI today? Interesting question.
24:39there's subtext there for the listeners whenever anyone says interesting question they're like stupid question and that wasn't in the notes no no no no i actually do have an answer for this so i think the most powerful thing about about ai and jai now as an investment professional is when i was an associate um your ability to get up to speed on the industry was a function of maybe doing some calls from expert networks, trying to buy some really expensive initiating coverage reports. And you were kind of, you know, piecing together very painfully information to get basic knowledge in order. Like I'll give you an example.
25:21I had to do, I remember I had to do a, we were looking at a company in the pumps and valves space. And I spent like five hours doing Google searches trying to understand like the different types of pumps and valves. And what I'd say now is your ability to take within 30 minutes and get up to speed on almost any sector on Earth is limitless. Like you can go as deep down the rabbit hole on pumps and valves or semiconductors or whatever topic, and you can learn at a totally different pace. I actually think that's going to be one of the really interesting changes in society and our educational system in some ways is people who are intellectually curious will be able to expand their knowledge bases at multiples of people that are not intellectually curious.
26:04And so you can learn almost any topic on earth in a week if you put your mind to it. And so I think the really powerful thing, if I got to, I'd say, after a week, I could probably come up to speed on industries in basic form. I think you can get to really interesting levels of sector expertise really quickly now. and i think that the um ambitious and intellectually curious associates will do that i think you will have your classic power to law dynamics of certain people who really dig in and learn whole spaces in the times that others are you know i think you've had people that did one thing for only 10 years and became experts in it which is great but i think for people that want to be generalists want to do a lot more and cover a lot more ground you can cover almost infinite ground if you're intellectually curious enough now.
26:52Completely agree with you. And it's almost like we have Kevin McAllister, founder of Access with us. He always talks about being intellectually curious. And it's why we built NOAA Research and AI-driven research capability that you pick an essential service, we can go crazy deep on it. We can build, we can market map. We can build one-to-own lists and build in-depth theses. And we work with other organizations and independent sponsors to support their work there as well. But that is a great application. And if you were back to your entrepreneurial days and you were building a company in the lower middle market, what is your single use of AI for that?
Read the full transcript
27:37Would that be learning too? Yeah, I think that there's a bunch of differences between starting a company now versus 10 years ago versus 20 years ago. So let's say that, let me take three different time crosses. 15, 20 years ago, if you wanted to start a software company, you literally needed like mainframe computer access. I mean, it was painful. It was expensive. You probably needed an HR department, a finance department. Your ability to start an initial company is expensive, time consuming. Even to get a baseline level would probably take you years. You know, as of seven, eight years ago, your ability to spin up a company, you can spin up a cloud environment pretty quickly.
28:15You can get Gusto payroll. You can get all the different components of the back office spun together really quickly. You can set up Salesforce. You can actually build scaffolding to build a business really quickly as of 70 years ago. But your obstacle back then would have been engineering predominantly. So you would have, you know, tried to figure out how you resource initial products. I think the most interesting difference now is, particularly with a lot of coding applications, with a lot of the multiplier effects of that, I think you're going to see the democratization of actually software engineering in some interesting ways.
28:51I'll give you an example. Think about, you know, when you were in university and you wanted to build a website. Like, you probably either had to pay somebody a pretty material amount to do a website, or you might have had to use Squarespace and something like that. The reality is now, you will be able to go text a website, you already can in a lot of ways, incredibly quickly. And so the barriers to stand up, one of the stats I heard recently was why combinators class this time and last are more than double the revenue of any prior vintage, just because of the speed of their engineering, right?
29:26So I think you have the ability to stand things up at a very different pace now. And so that actually has implications for companies in the middle market as well. Their ability to use modern engineering to take advantage of software development should be very different if they have a tech first mindset. Love that. Matt, thank you. It's been great to have you. Thank you for your partnership. And we're going to move on to Mark now. But we really appreciate you being on Why We Like It, and we'll see you soon. Thank you for having me. And really excited to be honest with Mark, who has done many amazing things in his career.
30:05Huge respect for that. Okay, my next guest. Not only is he a great friend, which I feel very fortunate, but we are joined today by the iconic innovator, one of the greats. someone who and i am going to make him blush has quite literally changed the world we live in he had a vision that everyone would walk around the world with a phone a smartphone in their pocket he transformed telecommunications created the blueprint for a whole host of technologies from intelligent agents to cloud to ios and android that came thereafter and since then And he's gone on to do things in the climate space, in artificial intelligence, in quantum computing, and a whole host of other things.
30:55He's also, as of just a couple weeks ago, the chief AI advisor at Access Holdings, helping us innovate the lower middle market to really make a difference for the backbone of America. Mark, thank you for joining us. It is my great pleasure, Sam. Really my great pleasure. so i mean i don't even know where to begin on the stats on ai we're going to talk big vision i want to blow everyone's brains with the scale of the opportunity the rate of change but just this morning when i was looking at the opening of the stock market nvidia is at 4.4 trillion dollars of market cap today fueling the growth in the ai agenda that's bigger than japan's gdp it's bigger on the UK's GDP, which is probably declining, but it continues to grow exponentially.
31:50And with it, the capability that's changing our lives. One of the areas I want to start though, is if we fast forward 10 years. So to talk to our audience, whether you're running a business in the lower middle market, or you're an investor, think about in 10 years time, what your company, what your platform might look like, but also think of the context it exists in. Mark, in 10 years' time, what does the world look like? Have we hit superintelligence? What is superintelligence? I would love to double down on what you said to encourage people viewing this, our guests, to do just that. Give yourself the permission to go 10 years, as Sam requested, into the future and give yourself permission to see it vividly.
32:40Even if you don't, even if it's kind of like foggy, just give yourself the permission because that's where the beginning of vision is, is going out there into the near future and having a very clear grasp, conviction in the moment that you've seen something. And then turn around and see where we came from. So, context. So, yes, AI has just been on the scene since the chat GPT moment for 20 minutes in decade time. immediately. They're now talking about the next wave, which is artificial, which is general intelligence, not just AI, but AGI. And then the next level beyond that is when super intelligence comes.
33:26And Sam, because you asked, out there in the future, there is an intelligence, intelligences that have been created that now are smarter, more capable, and more valuable than any human being in economic terms that can do things, than any human being. So imagine, you know, is that a subspecies? What is that? We've never seen anything like it. We've never been in an environment where we are not the alpha, top of the apex of the food chain. We are always the smartest. Well, we're about to create something that is synthetic, doesn't have all the human properties, the humane human properties. But nonetheless, it is smarter than us.
34:18It is doing fantastic things for us. Yes, it's creating abundance, productivity, unbelievable science that we haven't even dreamt of. it's made signs such that we can have therapeutics you know drugs, prescriptions that basically solve things solve problems, it's just out there and at the same time it is really can be really pernicious if this thing goes rogue so that's in the future that's the 10 year future, we will be grappling with the most amazing profound amplifier of the human intellect since language and i've been through what four or five major technological revolutions including the ones you mentioned and this is this is 10x that if not 100x in importance so in a world where and if i understand correctly mark you're saying we could ask super intelligence to cure cancer We could ask it to solve things that human beings haven't been able to solve through science, study, research, pattern recognition, and so on.
35:30And it would be able to do that for us. It would, yes. In a brief word, yes. This is how you get there.
35:43The superintelligence basically understands, in your case, understands biology. the one we talked about of curing cancer. It understands biology profoundly at a quantum level. And what that means is that its ability to see what is going on with expression of, you know, where cancer comes is seen by superintelligence, not seen by classical intelligence. First, it can create materials that don't exist in nature, never existed in nature, that are profound. It can finally create the compute alongside, in 10 years, by the way, there will be actually quantum computers, primitive ones. But that combination is incredibly potent.
36:29And it'll be able to solve equations, for example, that make fusion happen. What happens when fusion happens to energy? What happens to water desalination? What happens to food? What happens to everything that is dependent on energy? We've been talking about fusion for 30 years. We've been talking about hyper-personalized medicines for 20 years since RNA and DNA became something you could program with CRISPR. So we've been talking about this is now the capability that will make it happen on the science side. On the human side, running a business, creating and inventing, co-inventing, co-creating, which is something that you coined, which I really love.
37:14All of those things will amplify the human dimension in intellect and creativity because it lets you do that. Now, the final thing to know about superintelligence is that once it's given a purpose, it then learns from its own environment how to modify that purpose, how to create subgoals, how to create strategies. And it will continue to learn and evolve, and that's when superintelligence happens, is when we're not programming it, it is now able to program itself. It's able to realize what does it need and go out into the environment, the knowledge environment, and find ways to learn about what it needs to know and then generate the code to execute on that.
38:00So think of the lovely things that could happen and think of the not so lovely things that could happen in that context. Yeah, I mean, there are two parallel worlds here, one that's dystopian and one that's utopian, one where human beings are freed of the grind, freed of disease, freed of poverty, freed of struggle, but yet another world that looks more like Spielberg's Ready Player One, where we're all having to live in a digital world just to escape the reality of disparity. Which do you see? Thank you for that question. You see the background here. That's not a digital background. That's where I live and work.
38:40And this is San Francisco. And many of the AI companies, most of the big ones, are within a five-minute walking of where I am. So this is a bubble of AI. Very famous neighbors, right? We have famous neighbors. We have all kinds of incredible people. Blue Bottle Cafe down there. You walk in. It used to be empty because San Francisco went down. You walk in there, you can't get a seat at the table, and everyone is talking AI. I mean, everyone. This is a coffee shop. So, in any event, in these bubble circles, what you said, what you articulated, Sam, is exactly right. There is a really robust live debate, which is now global, and it's not just here, of are we tending to the, as you said, to the utopian, a place of abundance, a place of incredible progress in things that we want to have the defined high quality of life.
39:38And on the other side, the utopian side, on the dystopian side, what they talk about is not abundance, so they talk about extinction. That's the ultra, ultra, you know, sort of the extreme view of dystopian. And do you know how many people are on that extreme view? A lot. What does Peter Diamandis, Peter Diamandis wrote my favorite book, Abundance. If you haven't read it, you must. It's an incredibly positive spin about the trajectory of humankind. He's a friend of yours. What does Peter say? Well, Peter, as you said, says abundance. By the way, he has a podcast that's brilliant. He's prolific at that called Moonshots.
40:20It's well worth getting in the habit of seeing it if you're so inclined. By abundance, what he means is that everything that we want and need can be produced at a fraction of today's cost, possibly, and distributed widely to everyone. That's the superabundance thesis. is that Eric Schmid was on with Peter and said the economists that Eric, and Eric is extremely well-connected, the ex-chairman and CEO of Google, the economists he's speaking with are talking productivity gains that the world has never seen. Not 7%, 12%, 14%, which we have seen, GDP growth, China in its heyday, but 20%, 30 % productivity.
41:11Imagine, we've never seen anything like it, which means that gets you into a zone where the abundance is so abundant that you can have the equivalent of universal basic income. You can have free cheap water. You know, you can have things that are fundamental in life, as well as personalized education, personalized doctor, personalized trainer, personalized art teacher. the best talent worldwide can be made available to you. And that's an abundance. That's an abundance place. And that's where Peter lives. Peter, we should package his hormones, whatever. He is such a positive human being. He kind of almost lives in a state of just positivity and joy.
41:57So love listening to him. At the same time, and this is an important answer to your question, point of view of your question, which I think you meant, the other side is as articulate. The extinction thesis is brilliantly laid out, and when you read it, it does not sound like wacko science fiction. It sounds like it could be inevitable. So I am asking your viewers, your audience, to do something that's really hard, intellectually and emotionally. Stand in the possibility that both of those could be true at the same time. In other words, the promise of AI will create scenarios in medicine, let's say, or in education.
42:40Pick a field that are amazing and at exactly the same time, scary. And they can both be true at the same time. So our goal as kind of people, sentient people, not semi-sentient, is what do we do as a society to tilt in the direction of positive and tilt away from the direction of the negative? What can we do? What can we do with AI, with large language models? You know, NVIDIA is four trillion and change. It might go to six trillion. It is an astonishing, explosive, volcanic growth of power, AI power. So how do we, and we've never seen anything like it, as I mentioned, so how do we prepare? That's the open question that's here.
43:33Yeah, and let's double-click on that in the context of private equity and building companies, which after some of what we've talked about feels somewhat trivial. Who knows what the world will look like in 10 years' time. But how do people prepare now? Whether you're the founder of a company, a lower middle market business in the United States, or you're an asset owner and you're a private equity investor, how should you be thinking about not waking up in three, four, five years' time and being decimated by the competition? which could happen AI this stuff is we think in terms of discontinuity where something is going private equity firm a fund something's going along just fine and AI begins to really hit in a way that works as opposed to and there's a dot dot dot discontinuity to the upside there's a dot dot dot continuity to the downside in private equity it's an IQ test which side would you like to be on?
44:44The discontinuity is happening. So it's a question of choice. Before I get to kind of the real answer to what you're looking for, which is using AI for tactical things, operating efficiency, using it for strategic, create new revenue, or using it for transformational discontinuity level stuff, which is complete reinvention of the business. Before I get to that, there is a predicate. There's a precedent that something has to happen before that. before we introduce new services and new this and new that, each of us, each of you, if not already done and not already on the journey, is to deeply integrate AI into your personal life.
45:30Because that has juice. Use it. The invitation can only be done if you use ChatGBT at this time because it needs memory. others are not offering memory, is to create a digital twin of you that knows you intimately. Because you've permissioned it, you've given it documents, you've given it things that some of your viewers are going to say, no, you've given it health records, financial records, you've opened up your real-time communication, you know, this conversation we're having today, you know, just take the MP4 and drop it in and say, make that part of me. That, when you get to that point, whether you want to talk to the me, who knows you intimately, knows you incredibly intimately, if you let it.
46:21This is where some people in the audience are saying, no, no, no. But the ones that say yes, the value that I have, I've had a me for two years. I can only compare it to, do you remember before there was a Google? How did you learn? How did you know anything? Then it was Google. and then there was the chat GPT moment and the chat GPT moment from where we were before which is kind of where we were before is like this when you hear to me going from just using chat GPT to the amazingness experience is the same leap I've been living in it for the last two years I invite you to do it it's not hard to do it and learn whatever it is that you're that you need to know whether you know from the trivial to the profound um and people use it for both some people use it just as a google search you know augmented no have conversations remember we used to say prompt engineering and everybody else didn't no that's 2024 have conversations because you have an you know you have an intelligent person who knows you you're you're me i call my me mia by the way i've given her legend and then i know exactly who she is where she came and hi Mia and off we go I'm having a problem with my dishwasher how do I reset it to the most profound questions of all Mia can go find it so once you have that once you have that in your skills in your awareness of you know what AI means on a personal level then you can begin to understand some of the tactical strategic and discontinuity or transformative impact in the business.
48:12So if you're there, if you're a ninja power user of AI in your personal life, awesome. If you create an AMI, awesomer. Or go create a AMI and you'll see what AI can do. And then we can come back and talk about what are the tactical implications. And those are mostly about doing things we're doing today smarter, cheaper, faster. That's basically what it, and that reduces operating expense, which increases margin, increased satisfaction for the customer, and everything's good. So if we fast forward, Mark, does that mean that when it comes to sourcing companies and the proliferation of real-time data available for everyone or trying to find arbitrage in businesses, the leveling of the playing field is totally eroded.
49:15There is no competitive edge on that stuff in five years' time because everyone has access to the same tools, technology, data sets. What becomes the edge in the industry? So to just touch briefly on the other things, we talked about operating expense and efficiency. Strategic is massive revenue acceleration for a whole host of reasons. And transforming is inventing a new business around your business. The expression around here is if you don't cannibalize yourself, your competitor will. So completely transform your business. In five years' time or so, yes, people will have these tools and they'll be able to do the operating expense, cheaper, faster, better, increase revenue and transform.
50:08So what's the edge for a private equity firm? It's the creativity to match a company and see a company and say, oh, that is an operating expense, low-hanging fruit. And efficiency, which means customer service is going to go through the roof because we're able to do that. Or they can see, the private equity firm can see some strategic ideas here that massively increase revenue and income. They see that. Or, top of the heap is, we can take that company with a discontinuity and create a new category and crush it. and you know historically you know whoever creates a new category gets 80 of the market cap and you know the enterprise value out of that new category that's transformative that is that is the content of the conversation in the future private equity circles where the investment committee lives that's what you're talking about and if someone has a tactical operational idea a strategic idea or a transformative idea, everyone's going to go to their smartphone or their computer and say, what do you think of this idea?
51:26And I just thought it was something fun. Do it across, so everybody uses a different large language model. You ask Gemini, and you ask Anthropoc Claude, and you ask JGBT, and you ask Grok, okay, why not, of what do you think of this idea? And rather than dispatching a committee to go study it per month, four answers come back within a minute, which could be stupid answers. They could be hallucination answers. They could be dead wrong. But if you train your professional me, if you create one, the investment committee and the CEO will be using, in my prediction, if they create a me, that me, that professional me will know how the CEO is thinking, to how the senior managing director partners are thinking.
52:20They know where they're coming from. They know their prejudice, their blind spots, their strengths, their genius. They kind of get a holistic view. So that's the extension of the personal me is to create a professional me. And the way they'll be exercising differentiation and competitive dominance is the creativity of that team, augmented by AI, to converse, to debate, to bring in data, facts, bring in hypotheses, look for torpedoes, red team, blue team, do all the things that an executive team needs to do in coming to an answer or to a point of view. And given the same exact tools in two private equity firms with exactly the same profile across the street from each other, those teams will come up with different ideas.
53:09And that's the differentiation. And if your transformative idea is better than the one across the street, boom, you start getting a flood of lps that want to be in your funds i've got so many questions for you and you've already tolerated a lot of mine uh the last uh few weeks and months i'm not going to go any further on the private equity side because that's some of the access secret source that we're working on uh but i want to ask some fun personal questions uh to to close out so people really get to know the icon that is Mark. I first heard of you when I watched the General Magic trailer.
53:47I watched it on an airplane. It changed my life and it led to me meeting you. I sought you out and here we are, friends, partners, doing all sorts of exciting things together. That documentary changed my life. What documentary changed yours? Oh my goodness. That's a great question. What was it about General Magic that changed yours? And I've actually heard that from others. Is the ability to, you should speak for yourself, but others have said, it's the ability to see something, a culture, an inventive place, extraordinary people, and that's possible. It's possible to get that level of, you know, kind of genius and fun and purpose all in one room.
54:37So it gave them the ability to go think in that way. That's one. There are many others. Mine are connected to a dark place, unfortunately, because of my background is there's a Holocaust family, you know, tragic thing. And from World War II. and that does take me to films and documentaries but films that explore that I don't want to go into it but there's that which other documentary I don't know I'll pick a one let's touch on general magic for me there are two things that stood out and have shaped the way I think the way I lead the way I live one is the ability to take people on a journey that is your vision that feels like their vision the the cult philosophy that you built which was just a beautiful thing you feel the energy in the documentary I wasn't there but I feel like I was and the second thing is contrary to popular belief success doesn't mean you build a unicorn success doesn't mean you need to look like nvidia um you know we talked you and i talked about the butterfly effect where um actions in one place that may be seemingly not successful at the end or not have longevity change the world we live in and that's what general magic did um to tell me uh i'm gonna ask some some specific questions but you assembled an all-star team that spun out of apple um who was the greatest unsung hero of that bunch uh bowser was a rabbit that ran around um our children our kids the ones that came in as not knowing much um uh where tony tony fidel is one no others i'll just talk about Tony.
56:48And the others were very, very young, really young people. It's like, you know, don't build a company with anybody over 30. It's a little harsh, but these are really young people. Two engineers sat next to each other, essentially 10 feet apart. Tony Fidel, Andy Rubin. Tony came in not knowing anything. I mean, I think it might have been his first job. he pounded the door down because he heard that the Mac team the Mac team had assembled and were doing something amazing which was a secret and it was a smartphone so he pounded the door down, came in, knew nothing eventually after the general magic experience where he was working on smartphones he went to Apple and worked on the iPhone team and created the iPhone with first the iPod then the iPhone big surprise, unbelievable The other one is Andy Rubin, who sat next to him, they were friends, who did the same thing.
57:44He was inspired by the vision, the passion, the sheer IQ points that were floating around. I can't imagine how many IQ points. And then went off after General Magic and created Android. And so these two guys sitting 10 feet apart were responsible for what is now 95 % to 98 % of all the smartphone operating systems in the world. I came out of 20 feet, two cubicles, they're 10 by 10 cubicles, we were generous in those days. 20 feet by 10 feet. So in those 200 square feet, you know, a butterfly, you know, that was a chrysalis. A butterfly came out of that and flapped its wings and 12 years later, Steve, with some of the magicians, we call them general magic, the magicians, went on stage and, you know, and introduced iPhone 1.
58:43And the rest is history of what's been going on. Unbelievable. I was going to ask you, if you could have dinner with anyone, who would it be with? You've had dinner with many amazing people. In fact, you had dinner with the author of this book, Tim Berners-Litton, recently. So maybe I'll ask a question about that. How was dinner bringing together Mark Porat and Sir Tim Berners-Lee in one sitting? Astonishing. So Sam has been friends with Sir Tim and his wife for a while, a long time. So coincidence that we knew each other. Tim Berners-Lee invented the World Wide Web. As simple as that. He invented it.
59:32the fathers of the internet, which is the internet without the web, there were four fathers. One of them, Vint Cerf, by the way, was my first boss. I worked on, as a graduate student, I worked on the internet under Vint Cerf, who created one of the forefathers. And at that time, there were 14 computers. That was the internet. So having dinner with Tim, who created the web, was mind-blowing. I mean, this is single-handedly, wrote the entire spec that became the World Wide Web. And he just wrote a book, and I went to the dinner party that you cited. Tim said something to me that just brought tears to my eyes.
1:00:22He said, out of the blue, I was at CERN, I was doing physics, and I heard about General Magic. I should have joined General Magic. And I started, you know, tears going. Why? Because the World Wide Web was just, the internet was just washing over the industry. And we scrambled to make a internet native experience. Agent technology, agentics, which is what it's called now. Agent technology and smartphones and these things, you know, watches. Which, by the way, is being made in Apple by Kevin. Kevin Lynch, who worked on the same thing, Smartphone. He's now doing the watch. So where we need to be is more people like Tim Berners-Lee, who have the audacity to break out and do something.
1:01:19That book is one of the best examples of a discontinuity. There was the world before, and then there was the world after. And people who saw the world after, dot, dot, dot, created trillions of dollars of value. Maybe tens of trillions, but certainly trillions of dollars of value out of that awareness. That's what happens when you see something on the horizon. Tim is a perfect example of that. And by the way, after the World Wide Web, in terms of revenue and impact, it didn't do anything for 10 years. It was just kind of lobbying around, and then it went exponential. That's how technology is. so we're now you haven't so you you know watching this going oh my god I've missed the train no we're now in the early exuberance stage where everybody's throwing money at it 90 % fail rate and that's okay because because the experimentation has begun that's a discontinuity in 10 years it'll have it'll climb that same exponential maybe less, maybe five years because it's growing so quickly and multi-trillion dollar market cap companies that do not even exist today there may be a couple of people across the street there the next NVIDIA or the next Apple or the next Google will hatch and that's where we're headed Mark, I'm going to leave it there That's a perfect place to start.
1:02:58And thank you for sharing that with our audience. Thank you for sharing it with me. And thank you for being a great partner on all of this stuff. I couldn't be more excited about doing it with you. And you give me a level of comfort that this crazy world we're going into isn't going to be quite so scary after all. So thank you for everything you've done. And thanks for being here. Thank you for being so gracious. And on it, I really appreciate the chance to speak with you, Sam. Thanks for joining Why We Like It. Don't miss an episode. Be sure to subscribe. Expect big perspectives, big profiles, and big potential.
1:03:37Don't miss out.
From the publisher
Private equity’s future won’t be defined by capital; it will be defined by capability.
In the debut episode of Why We Like It, the Access Holdings podcast exploring the ideas, industries, and inflection points shaping the lower middle market, host Sam Tidswell-Norrish sits down with two of the most respected voices in artificial intelligence: Marc Porat (technology pioneer and former co-founder of General Magic) and Matt Fitzpatrick (CEO of Invisible Technologies, former Global Leader, McKinsey QuantumBlack).
Together, they unpack what AI really means for essential industries and value creation in the lower middle market.
They cover:
- Why AI is accelerating faster than most industries are prepared for
- How superintelligence may reshape private markets and value creation
- Why essential service industries are uniquely positioned to benefit
- Practical, near-term applications operators should care about now
- How Access Holdings is embedding AI into its operating model: The Access Edge
For founders, operators, and investors, the message is clear:
The firms that combine data, systems, and learning velocity will win.




