I’m going to tell you something that might surprise you. I’ve built and sold two companies, invested in over 200 startups, including some of the biggest names in AI like Anthropic and OpenAI, and yet, my own attempts at using AI to automate my businesses have been, for the most part, a series of failures.
It’s true. I’ve tried all the fancy tools. The complex workflows. The custom-coded solutions. And you know what I’ve learned? Most of it is a waste of time. The real secret to AI automation isn’t about building some kind of super-intelligent, all-knowing system. It’s about finding simple, almost boring, ways to get AI to do the grunt work for you.
I’m writing this because I see so many founders getting caught up in the hype. They think they need to be AI experts to automate their businesses. They spend countless hours and thousands of dollars on complex solutions that never quite work as advertised. I’m here to tell you that you don’t. I built a multi-million dollar company with a few simple tricks, and I’m going to share my entire playbook with you in this post.
The Allure of the Automated CEO
When I first started experimenting with AI, I had this grand vision. I imagined an “Automated CEO” – a system that could handle everything from scheduling meetings to writing performance reviews. I figured, with my background in tech and my access to the latest and greatest AI tools, how hard could it be?
Famous last words, right? My first big “failure” came when I tried to automate our weekly team meetings at RemoteTeam. The goal was simple: record the meeting, get a transcript, and then have an AI summarize the key decisions and action items. I thought this would be a game-changer. No more tedious note-taking! More time for, you know, actual work.
I tried a half-dozen different services. Some were crazy expensive, promising human-level accuracy. Others were open-source projects that required a PhD in computer science to set up. The result was always the same: a garbled mess of a transcript that was more work to clean up than just taking the notes myself. The AI would confidently misattribute statements, invent action items that were never discussed, and summarize inside jokes as key strategic decisions. It was a disaster.
One time, our lead engineer, a guy with a very thick accent, was explaining a complex technical issue. The AI transcribed his explanation as something about a “herd of angry squirrels attacking the database.” We all had a good laugh, but it was a sobering moment. This wasn't saving us time; it was creating more work and confusion.
My Second Big Failure: The AI Email Bot That Went Rogue
After the meeting notes fiasco, you’d think I would have learned my lesson. But no. My next brilliant idea was to build an AI email assistant. I was drowning in my inbox, and I dreamed of an AI that could automatically categorize emails, draft replies, and even handle scheduling. I spent weeks building a complex system of filters and rules, hooking into the Gmail API and connecting it to a large language model.
For a few days, it was magical. Emails were being sorted, and simple replies were being drafted. I felt like I had unlocked a new level of productivity. Then, it all went horribly wrong.
The bot started getting a little too… creative. It replied to a potential investor with a series of emojis and a GIF of a cat playing a keyboard. It scheduled three different meetings in the same time slot. The final straw was when it sent an email to my entire team, in the middle of the night, with the subject line “A new dawn is upon us” and a body that was just the lyrics to a song from the 80s. My co-founder called me at 3 AM, thinking I had lost my mind. That was the end of the AI email bot.
The Realization: I Was Doing It All Wrong
I was ready to give up on AI entirely. I had wasted so much time and energy on these complex, over-engineered solutions that just didn’t work. But then, I had a realization. The problem wasn’t the AI. The problem was me. I was trying to get the AI to be a strategic thinker, a decision-maker, a replacement for a human. And that’s just not what it’s good at.
AI, at least in its current form, is not a CEO. It’s a very, very smart intern. It can do repetitive, time-consuming tasks with incredible speed and accuracy, but it needs clear instructions and human oversight. Once I understood that, everything changed. I stopped trying to build an “Automated CEO” and started looking for small, boring, repetitive tasks that I could offload to my new “AI intern.”
And that’s when I started to see real results. Here are the five lessons I learned from my failed attempts at AI automation.
Lesson 1: Start Small and Boring
My breakthrough with AI came when I stopped trying to automate complex, strategic tasks and instead focused on the small, boring, and repetitive stuff. The things that no one on my team wanted to do, but that still needed to get done. For example, at RemoteTeam, we had to process a lot of expense reports. It was a manual, soul-crushing process that involved a lot of data entry and receipt checking. I realized that this was the perfect job for my “AI intern.”
I didn’t build a fancy, custom solution. I just used a simple off-the-shelf tool that could extract data from receipts and put it into a spreadsheet. It wasn’t perfect. It made mistakes. But it was 80% accurate, and it saved my team hours of tedious work every week. That was a huge win. And it was the start of a new approach to AI automation for me.
Lesson 2: The 80/20 Rule of AI
This leads me to my second lesson: the 80/20 rule of AI. 80% of the value you’ll get from AI will come from 20% of its features. Most of the fancy, headline-grabbing features of these AI tools are just marketing fluff. They sound cool, but they’re not actually that useful in the real world. The real gold is in the core, boring functionality.
I learned this lesson the hard way. I can’t tell you how many times I’ve been pitched by AI startups with these incredibly complex and sophisticated platforms. They show me demos of AI agents that can do everything from writing code to negotiating contracts. It all looks very impressive. But when I dig in and ask how they’re actually using these tools in their own businesses, it’s always the same story. They’re using them for the simple stuff. Summarizing meeting notes. Drafting emails. Transcribing audio. The 20%.
So my advice is this: ignore the hype. Don’t get distracted by the fancy features. Find a tool that does one or two things really well, and then use the hell out of it.
Lesson 3: Human-in-the-Loop is a Must
This is probably the most important lesson of all. You can’t just “set it and forget it” with AI. You need to have a human in the loop to review the AI’s work and catch its mistakes. Remember, your AI is an intern, not a CEO. And just like a human intern, it’s going to make mistakes. A lot of them.
At RemoteTeam, we had a simple rule: any work done by an AI had to be reviewed by a human before it went out the door. This was especially important for anything customer-facing. We learned this lesson the hard way after my rogue email bot incident. We created a simple two-step review process. The AI would do the initial work, and then a human would review it, make any necessary corrections, and then hit “send.” It added an extra step to the process, but it was worth it. It saved us from countless embarrassing mistakes and protected our brand.
Lesson 4: It's About the Prompts, Stupid
I’m sure you’ve heard this one before, but it’s worth repeating. The quality of your AI’s output is directly proportional to the quality of your input. In other words, it’s all about the prompts. If you give your AI a vague, poorly worded prompt, you’re going to get a vague, poorly worded output. But if you give it a clear, concise, and well-structured prompt, you’re going to be amazed at what it can do.
I’ve spent hundreds of hours experimenting with different prompts. I’ve learned that the best prompts are specific, provide context, and include examples. For example, instead of saying “Write a blog post about AI automation,” I’ll say something like: “Write a 1500-word blog post in the style of Sahin Boydas, a Silicon Valley entrepreneur. The post should be about the common mistakes founders make when trying to automate their businesses with AI. It should include personal anecdotes, specific examples, and a strong, opinionated tone. The target audience is other tech founders. Here are a few examples of my writing style…”
See the difference? The second prompt is much more specific and gives the AI the context it needs to do its best work. It takes a little more effort to write a good prompt, but it’s worth it. It’s the difference between getting a generic, useless piece of content and getting something that’s actually valuable.
Lesson 5: Don't Automate Relationships
My final lesson is a simple one: don’t automate relationships. AI is a powerful tool, but it’s not a replacement for human connection. There are some things that you should just never automate. Things like investor updates, personal emails to your team, or thank you notes to your customers.
I almost learned this lesson the hard way. I was about to send out an automated investor update, drafted by an AI, of course. It was well-written, informative, and hit all the key points. But at the last minute, I had a change of heart. I realized that my investors didn’t just want an update. They wanted to hear from me. They wanted to know what I was thinking, what I was worried about, and what I was excited about. They wanted a human connection.
So I deleted the AI-generated email and wrote a personal update from scratch. It took me an extra hour, but it was worth it. I got a bunch of replies from my investors, not just acknowledging the update, but also offering help and advice. It was a powerful reminder that business is all about relationships. And relationships can’t be automated.
The Takeaway: Be a Practical Pessimist
So, what’s the takeaway from all of this? Should you give up on AI automation? Absolutely not. AI is one of the most powerful tools we have as entrepreneurs. But you need to be smart about how you use it. You need to be a practical pessimist. Assume that the AI is going to make mistakes. Assume that it’s going to be dumber than you think. And assume that it’s going to require a lot of human oversight.
If you approach AI with that mindset, you’ll be much more successful. You’ll avoid the common pitfalls and you’ll be able to unlock the true power of this incredible technology. So go ahead, hire your “AI intern.” Just make sure you keep a close eye on it.
Frequently Asked Questions
How can I apply this thinking to my own situation?
Start by identifying the core principle behind the opinion, not the specific example. Then ask yourself: does this principle apply to my context? If yes, test it in a small, low-risk way before going all in.
Do all experts agree with this view?
No, and that's fine. The best ideas in business are often contrarian. I share my perspective based on my experience and data, but I encourage you to seek out opposing viewpoints and form your own conclusions.
What's the most common pushback you get on this?
People often push back by citing exceptions or edge cases. And they're usually right that exceptions exist. But building a strategy around exceptions rather than patterns is a losing game for most founders.
What experience informs this perspective?
This perspective comes from over a decade of building companies in Silicon Valley, two successful exits (RemoteTeam to Gusto, MovieLaLa to Gfycat), and investing in 200+ startups including Anthropic, OpenAI, and Scale AI. I write about what I've lived.