Jan 24, 2025 · 1h 6m · my-first-million
I run a $180M+ company...here's how I'm using AI on a daily basis
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode, entrepreneur Andrew Wilkinson joins Sean Puri and Sam Parr to discuss the transformative power of artificial intelligence, sharing concrete automation workflows and debating the future of software moats. They also explore macro AI investments, investor psychology, and the critical importance of avoiding overly clever financial mistakes.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 53.5% of the talking time here. How this is scored →
The hosts as informed peer 5.5 Guest teaching 4.5 Guest disagreement 1.6 The hosts pushing back 2.4
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 28:40 Wilkinson declaring vertical software commoditized
Wilkinson forcefully challenges standard venture assumptions by arguing that even defensible niche software like funeral home management can now be replicated in 20 minutes.
Hardest push from the hosts ▶ 31:20 Shaan defending enterprise software moatsShaan directly pushes back against Wilkinson's thesis that software values will collapse, detailing enterprise procurement, security compliance, and liability defensibility.
Biggest teaching moment ▶ 18:25 Wilkinson demonstrating Claude financial restructuringWilkinson surprises the hosts by showing how ingesting raw CSV general ledgers into Claude identified structural holding company moves saving $100,000 in cash taxes.
The host holds their own ▶ 32:20 Shaan demonstrating software market expansion thesisShaan counters pessimistic software forecasts with real examples of personal app creation on Replit, proving how AI unlocks non-coder software markets.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Andrew Wilkinson on the Magnitude of the AI Shift | 4 | 5 | 1 | 2 | Andrew Wilkinson sets an enthusiastic vision for AI agents replacing human labor and administrative work. Hosts Shaan and Sam ask practical probing questions, with Sam sharing an anecdote about confusing an AI demo with an actual person. | |
| HubSpot Sponsor Break: 200 AI Business Ideas | 3 | 6 | 1 | 1 | After an initial sponsor read, Wilkinson educates the hosts on using tools like Lindy, Howie, and Fixer to automate administrative tasks such as email drafting and restaurant bookings. The hosts largely listen and react with astonishment. | |
| Sean Puri's AI Task Evaluation Framework and Edge Cases | 7 | 3 | 1 | 3 | Shaan takes control of the dialogue to lay out his practical framework for evaluating AI in business operations rather than abstract model benchmarks. Wilkinson listens and asks clarifying questions about specific applications. | |
| Enterprise Forecasting, Meeting Summaries, and Life Planning | 6 | 4 | 1 | 1 | Shaan and Wilkinson build on each other's points about meeting summary tools like Fathom and Otter. Wilkinson shares a personal example of planning a relationship offsite with AI, while Shaan details operational time savings. | |
| Custom Agents, Financial Claude Hacks, and Virtual Coworkers | 5 | 6 | 1 | 2 | Wilkinson details sophisticated use cases of feeding complex financial CSVs to Claude to identify tax optimizations and replace back-office hires. Sam interjects with practical critiques about brittle Zapier workflows breaking in production. | |
| Mainstream Adoption and Everyday Family AI Use Cases | 5 | 2 | 0 | 1 | Sam asks whether mainstream non-technical family members use AI, prompting Shaan to share anecdotes about his mother and sister using ChatGPT for drafting and medical test interpretation. The dynamic is purely collaborative storytelling. | |
| Why Traditional Software May Become a Worse Business | 4 | 7 | 4 | 2 | Wilkinson presents a contrarian investment thesis arguing that vertical software margins and moats will erode as no-code AI tools make custom software replication trivial. Sam and Shaan listen intently to the Constellation Software comparison. | |
| Software Moats, Market Expansion, and No-Code Creation | 8 | 4 | 3 | 6 | Shaan pushes back constructively on Wilkinson's thesis, arguing that enterprise procurement, security liability, and habit moats protect established software while market size actually expands. Wilkinson concedes that user experience and support create nuance. | |
| Larry Ellison's Vision for AI and Healthcare Solutions | 5 | 6 | 2 | 4 | Wilkinson pitches his asymmetric public market AI thesis around Iris Energy transitioning Bitcoin mining infrastructure to AI compute. Shaan and Sam question the assumptions around cluster capacity and debt while teasing Wilkinson's sales pitch. | |
| The Charisma Discount and Avoiding Overcomplicated Theses | 7 | 3 | 2 | 3 | Shaan articulates the 'charisma discount' heuristic to avoid being seduced by overly persuasive investment theses or clever content creators. Wilkinson and Sam agree, reflecting on past temptations to overcomplicate simple investments. | |
| The Cost of Being Too Clever: Personal Investing Mistakes | 6 | 3 | 1 | 1 | The group candidly discusses costly investing mistakes caused by getting too clever rather than taking obvious winners. Shaan shares a crypto tax-loss harvesting misstep and Wilkinson recounts turning down Slack equity for agency cash. |
Statements from this episode (25)
Wilkinson: Rolling Out Existing AI and Self-Driving Would Eliminate 20% of Jobs
“So if we just paused AI, we shut down all the AI labs and all we focused on was just rollout of what already exists. I think probably Between self-driving cars and AI agents, 20% of all current jobs are gone.”
Wilkinson: AI agent replaced assistant for pre-meeting research briefs
“Like I have a, I have an agent that preps me for every meeting. So, 30 minutes before every meeting, I get a text and it says, hey, you're meeting Sean and Sam. You were emailing them about this topic. You're going to invest in their company. And here's two bi…”
Wilkinson: Fyxer reads Gmail, prioritizes emails, and drafts responses
“I use fixer F Y X E R. It's this British startup. It basically reads every single email that comes into your Gmail. It sorts it based on whether you need to respond to it or not. And then it drafts responses.”
Puri: Evaluates latest AI models every 3-6 months on internal business tasks
“I have my own little eval, which is in my businesses, I basically picked a few tasks that I'm like today, humans in my companies do these tasks. And about every sort of three to six months, we go back and we run through it again. We said, take the latest model…”
Wilkinson: AI product photography will solve last-mile issues in 6-12 months
“There's definitely like a last mile problem, but it'll be there in the next six to 12 months, I think.”
Puri: Fathom is far better at meeting notes than human assistants
“Not only did it replace the human, it's way better than my assistant could do it. Cause she'd be trying to keep up and it's just was not going to be good enough.”
Wilkinson: Claude Saved $100K Annually in Taxes Across Holding Companies
“I actually saved a 100,000 dollars a year of cash taxes by moving an investment from one hold code to another thanks to Claude”
Wilkinson: Hiring an AI-First CFO to Build Financial Automations
“We're hiring like an AI first CFO, basically somebody where their job is basically just to build automations, financial automations.”
Wilkinson: Indistinguishable 4K Video AI Coworkers on Zoom Within 24-36 Months
“I think that within 24 to 36 months, these will be four K video people that are indistinguishable, indistinguishable from a human that you'll talk to on zoom.”
Puri: ChatGPT Explains Medical Test Results Better Than Doctors
“And we just upload all of the screenshots into chat GPT. We're like, what does this mean? And it gives us beautiful, perfect bedside manner explanations, more patient, more thorough than the doctor. You could ask follow-up questions at your own leisure.”
Wilkinson: Software will commoditize as AI eliminates coding barriers
“And so my belief is basically that as more and more people don't need to learn how to code and can build software, software will become more and more commoditized. There'll be way more competition and it'll be a way harder world for software businesses.”
Wilkinson: Underwriting software deals conservatively, shifting focus to communities
“Personally, when I'm underwriting software deals now, I am being incredibly conservative. And honestly, I'm much more focused on communities and social networks as a result. We still look at software when there's like a hardware component or there's some kind …”
Puri: Enterprise and habit-driven software will resist AI commoditization
“So you have software where there's already habits. Those will be safe software where the value is the people or the data. Those will be safe. Software where the value is in the enterprise relationship, the service that's on top of it, or the like fine tuning a…”
Puri: AI software creation will expand the total software market
“And so I think the overall market for software gets bigger because you can create way more stuff. So, so that's the other counterpoint to like software being bad.”
Wilkinson: Traditional stock photo businesses are doomed by AI generation
“Two years ago, you know, stock photos generated by AI were terrible. Now I would not want to be in the stock photo business because it's perfect and it's only getting better.”
Parr: HubSpot canceled millions in booked ads after acquiring The Hustle
“Dude, when we sold the company to HubSpot, one of the reasons we went into the year with millions and millions of dollars booked of advertising, and we were probably going to do twenty million dollars the next year. And they were like, yeah, just cancel all th…”
Wilkinson: Iris Energy generates $500M annual EBITDA at $2.4B valuation
“And right now they're making somewhere in the range of five hundred million dollars of EBITDA annually doing that. Now the stock trades at about a 2.4 billion dollar valuation.”
Wilkinson reveals $1 million personal stock position in Iris Energy
“I do own the stock. But I think it's interesting. ... I only bought like a million dollars, but I look at it as like a small, it's a very small position.”
Puri: Apply a 'charisma discount' to persuasive speakers and amplify deadpan ones
“I apply basically a discount factor. It's like a charisma discount. So the more persuasive somebody is about everything, the more I have to discount to some degree. It doesn't mean I just like throw it away. Right. But I have to be careful not to get overly ex…”
Puri: Content creator incentives compromise advice from online health influencers
“This is my fear of all, all YouTube doctors, by the way. And I like your man. I like it to you. I like all the, I like them, but also they're content creators now first. And so you, again, you have to apply the discount factor, which is that Their job is every…”
Wilkinson: Passed on investing in Bench Accounting over low-margin services concerns
“My first designer was this guy named Adam Saint... He started this company called bench. And so he comes to me and he's like, Hey, we're doing, you know, basically AI accounting 10 years early... But I looked at it and I was like, oh, it's a services business …”
Puri: Passed on early Coinbase round to buy Bitcoin directly
“For example, I had the opportunity to invest in Coinbase, not super early, but like, let's say, series A or B, somewhere there. And I remember doing the math and being like, okay, I believe in Bitcoin, should I invest in Coinbase, the exchange, or buy Bitcoin?…”
Wilkinson: MetaLab declined equity for designing Slack at a $20M valuation
“Metalab back in 20, I guess it was 2012 or 20 13, we designed Slack. And at the time Slack was actually a failed gaming startup and Stuart Butterfield was basically trying to do something with the remains of it and trying to build, build a chat tool. And so He…”
Parr: Passed on TrueMed because founder had too much going on
“My most recent one is Justin Mayer's TrueMed. He was raising for it at a, I forget what it was, reasonable valuation. Justin's my friend. I think Justin's amazing. He had, he, I thought he had too much going on and I was like, no way are you going to focus on …”