Real talk from a technical founder building AI-powered businesses
Published October 08, 2025 • Based on Founder Reality Episode 27
Also available on: Apple Podcasts • Spotify • YouTube
Last week, Yale University released a report analyzing 33 months of labor data since ChatGPT launched. Their conclusion? There's no proof AI is displacing jobs.
Then Stanford dropped their own report a month ago with a completely different conclusion: AI is destroying junior jobs, with entry-level software developer employment down 20% from 2022 peaks.
So which one is right?
Both of them. And understanding why reveals something critical about what's actually happening in the job market right now.
Why Yale Says Nothing Has Changed
Yale looked at the macro picture - total jobs in the economy. They measured job exposure to AI across different occupations and tracked unemployment data alongside actual AI usage from tools like ChatGPT and Claude.
From their view, nothing has really changed. Jobs aren't just disappearing. America is a huge economy, and the total number of jobs hasn't collapsed over the past three years.
And honestly, they're not wrong. Think about what we just got last week - Sora 2, Claude 4.5 Sonnet. These are impressive, but they're still not good enough to replace most jobs.
Claude 4.5 Sonnet is the first model that could successfully recreate the Claude.ai interface when researchers asked it to. Every previous version failed. That's impressive progress, but we're still far from AI that can actually replace entire job functions.
People I've talked to at Google, Meta, and other big firms say they're spending less time coding from scratch. AI is taking away the early-stage grunt work. But they're spending more time on high-level thinking, architectural design, and optimization.
The jobs still exist. Companies aren't laying off predominantly because of AI.
Why Stanford Says Junior Jobs Are Disappearing
But Stanford looked at something different. They tracked individual people early in their careers - recent graduates trying to get their first jobs.
They worked with ADP, a major payroll provider, to get anonymized data on who actually gets hired for entry-level positions.
And on this micro level, there's a definite shift.
Companies are asking themselves: "Why do we need to hire junior analysts to do work that ChatGPT can do 60-70% of? We can just pay our mid-level people a bit more (or keep them at the same salary) and they can handle this with AI assistance."
Entry-level customer service jobs show similar trends. The positions haven't disappeared completely, but hiring has slowed dramatically.
The Pattern I'm Seeing Firsthand
I graduated from University of Waterloo. It used to be incredibly easy for co-op students from Waterloo to get hired by Google, Meta, and other big tech companies. This was back in 2021, pre-ChatGPT.
Now it's getting tougher. Junior roles are disappearing. Sure, the best people from the schools are still getting hired. But students who need more time to learn, who haven't adapted to AI yet, or who simply don't have work experience when they graduate - they're struggling.
And most students don't have work experience. That's normal. Schools are supposed to provide education, not guarantee jobs. But the gap between graduation and employment is widening for anyone who isn't at the very top.
Even Starbucks Is Cutting Office Jobs
Last week, Starbucks had a massive round of corporate layoffs. They let go of 900 corporate staff and paused almost every new hire posting.
Most of the positions cut were office jobs. They closed locations too, but the corporate staff cuts were significant.
This isn't just a tech problem. It's happening across industries.
What's Happening at SimpleDirect
We haven't laid off anyone for AI reasons. We did have a small layoff in early 2023, but that was purely operational - we didn't have enough money to pay people, so we had to narrow scope.
Our team shrunk to 3-4 people then. Now we're at 5. In the past two years, we've only added one role.
And honestly? We've had very few turnovers since 2023. Here's why:
For roles that can be done by AI or contractors (Fiverr, Upwork, LinkedIn freelancers), we use those. Design stuff, marketing stuff - we do it on demand with contractors paid by task.
For stuff that can be automated, we use AI. It gives us speed. It's not perfect, but it's good enough to avoid hiring for more roles.
There's also an internal reason: I don't want more people to manage. Two years ago I had 14 people on the team including myself, and it was a nightmare. The time spent managing everyone was insane.
So yes, we're not hiring junior staff right now. But we're a small company.
The World Economic Forum Numbers
The World Economic Forum projects that AI will create 1 million jobs and displace 9 million jobs.
That's a net positive on paper. But if you're one of the 9 million people who gets laid off, that's not exactly comforting.
It basically means people who know how to use AI will take the jobs. People who can't use AI won't.
That's terrifying.
The Youth Unemployment Crisis Nobody's Talking About
In several countries, youth unemployment has hit record highs. Canada is at a 30-year high of 15%.
This isn't entirely AI's fault. But in a future where graduates can't get white-collar jobs, how do we sell the value of universities and higher education?
What happens when the only jobs available for young people are in service industries that aren't as heavily impacted by AI?
AI Still Isn't Good Enough (But It's Getting There)
Here's the reality: AI technology isn't at a place where it can replace any single person I've worked with or any single job function I see.
Even with AI-enabled customer support, I still have to do heavy lifting - talking to customers, understanding nuances. AI just isn't there yet.
But I can see it having the potential to get there in the next few months or years.
That's what makes this so urgent.
If You're Young: Make Yourself AI-Compatible
If you're listening to this and you're young, make yourself more AI-compatible to be hired.
I'm not saying you need to be an expert on everything. But learning how to learn is critical.
If I have no idea how to use Lovable (a tool that builds websites with AI), I can probably figure it out in half an hour by reading documentation and using AI tools to help me understand.
That makes me more employable than someone who needs more time to learn.
There's a productivity gap forming in developed countries, especially in the G7. Young people need to step up and use the free resources available:
- Perplexity (free)
- ChatGPT (free tier)
- Claude (free tier)
- Google Gemini (free tier)
All the information is readily available online. It's not difficult to learn if you're really desperate to have a job.
The ConvertKit Story Nobody Tells
Let me tell you about ConvertKit. Everyone's heard of them - if you create content, you've probably seen their email opt-in forms.
The founder Nathan Barry took the company from $5K to $36M ARR. Impressive, right?
But nobody talks about the messy middle. Let me walk you through the actual timeline:
2013: Nathan launched ConvertKit with $5K of his own money. He set a very public goal: hit $5,000 MRR in six months.
He failed. He only hit $2,000 MRR.
2014: It got worse. Revenue dropped to $1,300 per month. He was ready to shut it down.
But here's the key decision: He put in $50,000 more of his own money, hired a full-time developer, and doubled down.
2015 March: Two years after launching, he finally hit that $5,000 MRR he was supposed to hit in six months.
2015 December: Nine months later, they hit $97,000 MRR. That's 19x growth.
How?
The Things That Don't Scale
Nathan did something crazy: concierge migrations.
Their competitor was Mailchimp. Customers would say "We don't understand ConvertKit. We don't know how to migrate."
Instead of writing blog posts or help articles, Nathan said: "We'll help you. Personally. On Zoom calls. We'll move everything for you."
Think about the time investment. Hundreds of hours spent personally helping customers switch from a competing platform.
That's the perfect example of doing things that don't scale.
They also tapped into affiliate marketing. Pat Flynn did one promotion and they got $5,000 more MRR in a single month.
2016: They hosted 150+ webinars. Not sales pitches - actual education on how to use email marketing and how to be a creator.
They closed the year at $600,000 MRR. Six times growth from the webinar strategy alone.
Think about doing 150+ webinars in a year. That's commitment.
2023: Over $36 million ARR, 49,000+ customers. They stayed laser-focused on creators the entire time. They never added new customer segments. Twelve years focused on one group.
And they're bootstrapped. They never raised venture capital.
Three Lessons From ConvertKit
1. Don't give up too easily.
Nathan failed publicly and didn't stand down. Even when revenue dropped in year two, he doubled down with $50K more.
I've questioned myself multiple times whether to shut down SimpleDirect. I'm glad I didn't. We grew from a few thousand MRR to tens of thousands. And even though we're sunsetting SimpleDirect Financing, it had nothing to do with the product - it was external forces.
2. Do things that don't scale.
As bootstrap founders, we think about building websites that convert, writing blog posts, going all-in on social media. We expect traffic to come by itself because of stories we've heard online.
That's not what happens. You have to sell yourself. You have to get the word out there.
Nathan spent hundreds or thousands of hours helping individual customers switch platforms. That goes against every startup orthodoxy about scale, but it worked.
3. Bootstrap is about depth, not breadth.
ConvertKit didn't try to beat Mailchimp by being Mailchimp. They found that creators were being underserved and ignored. They obsessed over that specific group for 12 years.
They never wavered. Not when things didn't work. Not even when things worked really well and they could have expanded to new segments.
That focus made them $36M ARR without venture capital.
But 2025 Isn't 2013
How Nathan started ConvertKit shouldn't be exactly how you start a company today.
In my free ebook "De-Risk Your Startup," I talk about a new approach: Use consulting first, leverage AI to deliver services at scale, charge high premiums, build an initial bank of money, then adapt that experience into actual products and scale.
Consulting first, empowered by AI, making money, turning that experience into products.
We have more capabilities now than 12 years ago. Zoom didn't exist. Google Meet didn't exist. AI didn't exist.
You don't have to put $50K on the line like Nathan did. There are more resources, more compute, more tools you can use as a solopreneur without additional capital.
Three Founder Tweets That Hit Different
I spent the past week curating content from founders you probably haven't heard of. Not the super influencers like Naval or Sam Altman. Real founders in the trenches.
Tweet 1: Obsession
From Slash Remish Nudie:
"Startups are not about safety. They're about placing bets so bold that they scare you. Most founders quit when things get hard, but successful ones share one trait: Not luck, not connections, not brilliance - obsession. Pick a problem you cannot stop thinking about. Go all in when logic says stop. Stay in the game longer than anyone else."
This hit me hard.
I had a conversation with a friend running a content business. He asked about scaling tools and whether to hire a team. I told him: You know exactly what audience you're serving and what problems they have. Be obsessed focusing on that. You don't need to hire anyone yet.
Uber is the perfect example of this.
When Uber first started in San Francisco, the taxi union threatened them with criminal reporting, hundreds of thousands of dollars in fines every day, and potential jail time.
Travis Kalanick said "Go F yourself" and kept going.
The logical choice was to comply, stand down, shut down the business. But Travis didn't. Look at Uber today.
I had a similar moment. I was signing a partnership with a big bank - a multimillion dollar deal. But something felt wrong. It took them 3.5 months just to review our IP. The weekly meetings, the hustle, the time investment.
Logic said take the deal. I ended the discussions and didn't sign.
Not logical, but the right call.
Tweet 2: Emotional Detachment
From Gorov Saying 03:
"The best founders I know are not the ones who never feel fear or excitement. They're the ones who don't let either control them. Detach self-worth from business outcomes. Write down emotions before making big calls. Remember: the company is not me. It's something I'm building."
I feel this deeply.
A lot of people say I'm too humble. I admit I'm wrong very publicly and multiple times. When we shut down SimpleDirect Financing in September, I was very transparent about why we're sunsetting something I spent four years building.
I wasn't hiding that I screwed up. That's raw. That's emotional. But I'm not ashamed of it.
Your business is not you.
I used to be very triggered when people said bad things about the company. I'd defend every single comment emotionally.
Eventually I realized: You can't make 100% of customers happy. You can try to make 51% happy. For the rest, figure out why they're not happy and try to fix it.
I did 100% of customer support until recently. I had phone calls where customers told me "This is so disappointing" - not in an email, but raw, over the phone.
It hurt. But I learned to see outcomes as data points, not identity points.
My framework now: Write down your emotions if you're emotional about something. Then be rational about the decision.
Tweet 3: Marketing From Day Zero
From Saeed Barak:
"Advise for first-time founders: Marketing is from day zero. MVP means single killer feature and no more. Bake distribution in the app. Subscriptions more important than one-time payments. Launch in month one, make first sales, start marketing in month one and compound."
I have to admit: I only picked up marketing in recent weeks and months.
This is probably the single most important mistake I made with SimpleDirect Financing.
We focused on business development, distribution partners, affiliate marketing. We spent minimal time on actual marketing because we thought we were targeting blue-collar workers.
That was an excuse. Blue-collar people use Facebook, Yelp, Instagram - everything we use. Their habits are different, but you can still reach them with marketing.
I learned this the hard way. Now we're sunsetting SimpleDirect Financing and launching a new product targeting white-collar workers and startup people. But we don't have any marketing infrastructure to do that.
I took a big L. I should have done marketing from day zero of SimpleDirect Financing.
I've spent the past few weeks doing 20-30 hours per week on marketing, trying to catch up on four years of missed marketing.
It's overwhelming, but that's the way it goes.
If you're launching a new product or you haven't started yet: Bake in marketing, bake in distribution channels, start creating a brand either for yourself or your business.
Don't wait. Start marketing even before your product is born. Maybe 2-3 months ahead of time.
How These Three Connect
You need obsession to get started and keep building when logic says quit. Obsess over your customer segment and their problems - not yourself or your company.
You need emotional detachment so that obsession doesn't destroy you. Your business is separate from you. Outcomes are data points, not identity points.
You need marketing from day zero. Don't make my mistake of waiting four years. Start before you even have a product.
The Bottom Line
Yale and Stanford are both right about AI and jobs. The macro picture looks stable, but the micro picture for young people entering the workforce is brutal.
ConvertKit shows us the messy middle everyone hides - two years of public failure before hitting initial goals, but staying obsessed enough to reach $36M ARR.
Three founder truths emerged this week:
- Obsess over customer problems, not your company
- Detach emotions from outcomes
- Start marketing before you even have a product
Junior jobs are disappearing. The gap is widening. But there's still opportunity if you adapt, use AI, and stay obsessed with solving real problems.
Don't wait for logic to make sense. Go all in on something you can't stop thinking about.
What made you think this week? Tweet me @TheGeorgePu with resources you want discussed in the next "Three Things" episode.
Get my free ebook "De-Risk Your Startup" at founderreality.com - the consulting-first approach to building in 2025.
Subscribe to my weekly newsletter at newsletter.founderreality.com for behind-the-scenes startup stories you don't hear on the podcast.
New episodes Monday/Wednesday/Friday at 9am EST. Real founder lessons from the messy middle.
