5 Hot Takes on AI and AI agents

For anyone tired of the AI echo chamber

Every time I log into LinkedIn or X these days, I’m hit with a flood of AI hype. People with zero coding experience claim they’ve “built a full-scale app in 10 minutes,” others say they’ve quit their jobs to “print money with AI agents” - and it only gets more extreme from there.

I bought into the hype for a while. But after months of hands-on experimentation, I’ve realized most of these voices are charlatans with little real-world experience. In this article, I share five hot takes on AI and AI agents grounded in actual trial, error, and exploration. Anyone not interested in reading the full article can find a summary here.

On another note, I haven’t been posting as consistently, due to a new job that’s taking up a large chunk of my time. I am working on getting my schedule back on track, so I can resume posting every Monday again.

#1 - To build a big business, prioritize problem solving first, AI second

Over the past 18 months, I’ve gone deep into AI and AI agents, especially in the context of GTM. I participated in training programs with CrewAI, Clay, and Gumloop and spent countless hours experimenting with tools like Apify and N8N.

But the deeper I went, the more I realized how vast the landscape really is. Some of my friends were clocking 3+ hours a day on AI tutorials, hackathons, and community sharing.

A pattern began to emerge: the most fervent AI adopters typically evolved into one of two paths:

#1 - AI educators
#2 - AI consultants

Nate Herk is my favorite

There’s nothing wrong with either. Both are in high demand and can be pretty lucrative. However, if your goal is to build a product (or productized service) business, most of your time will be spent solving hard problems and grinding through unsexy work. AI can support that grind - but to build something meaningful, problem-solving should come first and AI education second.

#2 - Leave AI products to engineers / product people

When I first discovered AI-assisted coding (aka “vibe coding”) and spun up a basic app in minutes, I felt unstoppable. Like Lukas Mattson from Succession, I believed that with a bit of analysis, capital, and execution, success would be inevitable.

Me after my first day of vibe coding (circa 14 months ago)

But then, reality set in quickly. While AI helped me build 70% of an app in minutes, the final 30% - the part that made it actually work - required hours of tedious problem-solving. A personal tool I built in 4 hours to convert long-form content into tweets took nearly 100 hours to properly deploy for others.

I also noticed a pattern: friends who dove headfirst into vibe coding without a product or engineering background often ended up building clunky, awkward products. They technically worked but lacked polish, had poor UX, and were often downright ugly. Usable? Sure. Desirable? Not really.

That experience grounded me. Instead of pretending to be a product expert, I’ve focused on using AI and code to enhance my GTM strengths.

#3 - There is no substitute for domain expertise

With a background in sales and marketing, I often spot the BS in AI startups that others miss. A few patterns I’ve noticed:

• Social media AI tools (like Jasper or Taplio) → Are often just ChatGPT wrappers with a UI and a few add-ons. Building a loyal audience still depends on sharing original insights, not repackaged content - and no tool can automate that.

• AI landing page builders (like Durable.co) → Tend to ignore basic marketing principles. The pages they spit out may look fine, but they rarely convert - because the creators don’t actually understand what makes a landing page work.

• AI-powered brand monitoring tools (like AthenaHQ) → are frequently overpriced and underpowered. Many don’t stack up against traditional SEO tools, which are cheaper, offer broader tracking, and have deeper feature sets (e.g. AthenaHQ at $270/month vs. Ahrefs at $108/month with 10x the coverage).

Yes, AI lowers the barrier to building - but if you don’t understand the customer’s real pain or lack domain expertise, you’re better off staying on the sidelines.

A prescient analysis

#4 - Most “AI everything” solutions are BS

Lately, I’ve seen a wave of AI solutions - often from folks with little domain expertise - making bold, sweeping claims like:

“Our AI replaces your entire sales team.”
“Our AI agents automate your whole marketing agency.”

For me, that’s an instant red flag.

Category-defining companies don’t try to conquer the entire market from day one. They start small, win a niche, build traction, and expand gradually. Jeff Bezos didn’t launch the Everything Store - he started with books.

Not a bad guy to learn from

Another red flag: overly complex AI workflows from creators deep into tools like N8N or Make. While both are powerful (I’ve written about N8N here and here), I’ve seen people string together 100+ nodes to simulate entire business functions. It looks impressive, but in practice, these systems are fragile, complex to maintain, and rarely scale beyond the demo phase.

When I see this, I run for the hills

#5 - To win long-term, apply AI to non-sexy problems

AI is evolving at breakneck speed, and many of today’s front-runners won’t survive the next wave. If you’re young, ambitious, and ready to go all-in on bleeding-edge tech, there’s a real shot at something big - but it requires relentless learning and adaptability. There’s also a serious risk: your business could be wiped out overnight by a model update or outpaced by a major player with proprietary data and the resources to train their own models.

If you’re aiming for long-term success, though, the smarter move is to apply AI to an under-served, unsexy problem that’s flying under the radar. And no - that doesn’t mean laundromats, self-storage, or short-term rentals. If it’s all over social media, it’s already too late.

If you hear the idea from her, its past its expiration date

The most interesting problems and overlooked niches often emerge through genuine conversations with sharp individuals in your network. Currently, I’m helping a friend scale a business that combines AI and coaching to help people secure better jobs more quickly. In today’s market, connecting individuals with high-paying opportunities is a challenging and high-value task and one that most AI builders are not rushing to solve.

Conclusion

AI is powerful, but it’s not magic. The loudest voices online often skip the hard parts - problem definition, domain expertise, and thoughtful execution - in favor of flashy demos and exaggerated claims. What actually works is less exciting: solving real problems, applying AI where it helps, and ignoring the hype. If you want to build something that lasts, start with substance. AI can accelerate your path, but it can’t replace the work.

---

If you liked this content, please click the <3 button on Substack so I know which content to double down on.

---

TLDR Summary

This article offers a critical look at AI hype in the growth and tech space. I reflect on my personal experiences and lessons learned from working with AI tools and agents. I also debunk common misconceptions, highlighting that AI can’t replace domain expertise, problem-solving, or the hard work necessary to build lasting businesses. The key takeaway is that AI should be used strategically to enhance existing strengths, not as a shortcut to success.

Key Steps and Insights

1. Prioritize Problem-Solving, Not AI First

   • While AI can support business growth, true success comes from solving real problems, not just using AI tools.

   • Focus on the grind and hard work; AI can enhance, but it shouldn’t be the core driver.

2. AI Products Are Better Left to Engineers and Product Experts

   • AI-assisted coding might seem exciting, but without a deep understanding of product development, AI tools often result in clunky, unfinished products.

   • Successful AI applications require thoughtful problem-solving, not just quick app builds.

3. Domain Expertise is Irreplaceable

   • AI tools, like those for content creation or brand monitoring, often lack the depth needed to truly address customer pain points.

   • A strong domain understanding is essential for creating products or services that genuinely add value.

4. Beware of “AI Everything” Solutions

   • Many AI solutions claiming to replace entire business functions are often oversold.

   • Simpler, niche-focused applications are often more effective than grandiose claims or overly complex workflows.

5. AI Should Tackle Non-Sexy, Under-the-Radar Problems

   • Instead of jumping on the latest trendy AI solution, focus on addressing overlooked or underserved problems.

   • Real, sustainable success often comes from solving tough, less glamorous challenges.

---

Conclusion

AI is a powerful tool, but it’s no magic bullet. The most successful applications of AI come from thoughtful problem definition, leveraging domain expertise, and executing with discipline. The loudest AI voices online often miss the hard work necessary for long-term success. If you want to build something lasting, start with substance and use AI to enhance your existing strengths, not as a shortcut to quick success.