I remember the exact moment I knew the AI project was doomed. We were six months and about three million dollars in. The team was presenting a new dashboard, a beautiful mess of charts and predictive models meant to revolutionize our sales process. It was supposed to be a crystal ball, telling our sales team exactly who to call, when to call them, and what to say. On paper, it was perfect, a masterpiece of data engineering. In reality, a senior sales director, a guy who had been with the company for 20 years and could sell ice in a blizzard, just stared at it blankly. After a long, painful silence, he squinted and asked, "So, I still just call my top leads from the list I keep in my notebook, right?"
That was it. The fatal flaw. We had built a rocket ship, but nobody had bothered to teach the astronauts how to fly it. All the algorithms, the data pipelines, the brilliant engineering—it was all worthless because we missed the most critical component: the people. We had solved a complex technical problem, but we had completely ignored the human one.
Everyone is scrambling to integrate AI, and most are making a hash of it. They think it’s a technology problem. Buy the right software from a slick vendor, hire a few PhDs with impressive credentials, and boom, you’re living in the future. That’s a fantasy. AI integration is not a tech challenge. It’s a change management challenge. And there are unspoken rules to this game that no one talks about publicly. I’m breaking the silence because I’m tired of seeing good companies waste millions on projects that are destined to fail from the start.
Rule 1: Stop Calling It "Artificial Intelligence"
The first thing I do in any company I advise is ban the term "AI." It’s a terrible, loaded phrase. It immediately conjures images of HAL 9000 or Skynet. It makes people feel threatened, stupid, or both. Your team doesn’t hear "efficiency" and "insight"; they hear "You’re being replaced by a robot." It’s a conversation-stopper.
Call it what it is. Are you implementing a "predictive lead scoring system"? Call it that. Is it an "automated customer support assistant"? Great, say so. Is it a "supply chain optimization engine"? Perfect. Be specific, even boring. Demystify it. Frame it as a tool that helps them do their job better, not a sentient being that’s coming to take it.
When we were building out the systems at RemoteTeam, which was later acquired by Gusto, we never talked about "AI." We built a "smart payroll processor" that flagged potential errors before they happened. We created a "talent suggestion engine" that surfaced qualified candidates from our existing database. These were tools that took the drudgery out of HR and recruiting. Our team wasn’t scared of them; they were excited. The tools helped them get home on time and focus on the human parts of their job, like talking to candidates and helping employees. That’s a benefit people understand and embrace.
Rule 2: Your "AI Champion" Cannot Be the CTO
This is going to sound controversial, but the person leading the charge for AI adoption shouldn’t be your most technical person. I’ve seen it a hundred times. The CTO or a hotshot data scientist gets put in charge. They can explain the intricacies of a recurrent neural network, but they can’t explain the business value to a skeptical marketing manager in a way that connects.
Your champion needs to be a respected operational leader. Someone from sales, or marketing, or finance—a person who feels the pain that the AI is meant to solve. They need to have credibility and trust with the people who will actually use the system. When they stand up and say, "I’ve used this, and it makes my life easier and helps me hit my numbers," people listen. The CTO saying the same thing sounds like a sales pitch for a project they are already invested in.
At one of my portfolio companies, a fintech firm, the AI initiative was going nowhere. It was being run by the head of engineering, a brilliant guy who could make an algorithm sing. But the frontline users, the customer success team, saw it as a black box being forced on them. We switched tracks and put the VP of Customer Success in charge. She was the one dealing with angry customer emails and an overloaded support team. She became the system's biggest advocate because she desperately needed it to work. She translated the tech-speak into real-world benefits: "This will help us answer the top 50 customer questions instantly, so we can spend our time on the really tough problems." Adoption skyrocketed. She didn’t know how the model worked, but she knew exactly what it did for her team and her customers.
Rule 3: Mandate "Day in the Life" Shadowing
Before a single line of code is written, before you even spec out the project, your technical team needs to get out of their chairs and into the trenches. I mandate that the data scientists and engineers spend a full day shadowing the people whose jobs they are about to change. Not a one-hour meeting. A full day.
They need to see the messy reality. The spreadsheets with a million tabs. The manual workarounds people have invented to get their jobs done. The constant interruptions. They need to feel the frustration of the user. This is where the real insights come from. It builds empathy, a critical ingredient that is almost always missing.
I once saw a team build a complex scheduling tool that, on paper, should have saved a logistics company a fortune. But it failed. Why? Because the engineers didn't realize the dispatchers were constantly getting new information over the phone and needed a way to manually override the system on the fly. They had designed a perfect, rigid system for a chaotic, fluid reality. A single day of shadowing would have made that obvious. Empathy isn't a soft skill here; it's a fundamental part of the design process. It ensures you are solving the real problem, not the one you imagined from the comfort of a conference room.
Rule 4: Find the Pain and Start There
Don't start with a moonshot project. I’ve seen founders, fueled by investor hype, try to build an all-knowing AI brain for their company from day one. It never works. It’s too big, too complex, and the ROI is years away. You’ll burn through your cash and your team’s goodwill before you have anything to show for it.
Instead, find the most painful, repetitive, soul-crushing task in your organization and build a solution for that. Is your finance team spending 40 hours a week manually matching invoices to purchase orders? Automate it. Are your junior analysts just copying and pasting data from one system to another? Fix that.
These small, targeted wins are everything. They build momentum. They prove the value of the new tools in a tangible, undeniable way. It’s a simple, powerful loop: find a pain, apply a targeted solution, and then broadcast the success. At MovieLaLa, before we were acquired by Gfycat, we didn’t try to build a model to predict what movie would be a blockbuster. We started by building a simple tool that automated the tagging of movie trailers with genre, actors, and other metadata. It was a small, annoying task that everyone hated. Our little "tag-bot" was a hero. It saved the team hundreds of hours. From there, we earned the trust and the right to tackle bigger, more ambitious problems.
Rule 5: The 80% Solution is Better Than the 100% Failure
Perfectionism is the enemy of progress in AI. Your models will not be perfect. They will make mistakes. If you wait for a system that is 100% accurate, you will never ship anything. The goal is not to build a flawless oracle; it’s to build a tool that is significantly better than the status quo.
I once invested in a company that was building an AI for legal contract review. The founders were obsessed with getting it to 99.99% accuracy. They spent a year and millions of dollars chasing that last percentage point. Meanwhile, their competitor launched a product that was only 85% accurate, but it was 10 times faster than a human paralegal. They marketed it not as a replacement for a lawyer, but as a "super-powered paralegal" that could handle the first pass of a document review in minutes instead of days. They captured the entire market while the first company was still tweaking its model.
Your team needs to understand this. The AI is a co-pilot, not the pilot. It’s there to provide a first draft, to spot things a human might miss, to handle the bulk of the work. The human is still there to provide the final sign-off, the critical thinking, and the common sense. An 80% accurate tool that frees up a skilled employee to focus on high-value tasks is a massive win. A 100% accurate tool that never sees the light of day is a complete failure.
The Real Secret to Success
I’ve invested in over 200 companies, including some of the biggest names in AI like Anthropic, Scale AI, and Hugging Face. I’ve seen this movie play out again and again. The companies that succeed are not the ones with the most complex algorithms or the most data. They are the ones that master the human side of the equation.
They understand that AI is not a magic wand. It’s a tool, and like any tool, it’s only as good as the person wielding it. They focus on communication, on building trust, and on solving real, painful problems. They bring their people along on the journey, instead of trying to drag them, kicking and screaming, into the future.
So forget the hype. Forget the breathless articles about a future run by machines. Focus on the messy, human reality of change. The next time someone on your team starts talking about models and algorithms, stop them and ask, "Who is this for? Have you sat with them? Do you understand their job?" If they can't answer that, you're already on the road to failure. If they can, you just might be onto something big. Don't mess it up.
Frequently Asked Questions
How should I work through this guide?
Don't try to absorb everything in one sitting. Read through once to get the big picture, then go back and work through each section as it becomes relevant to your current challenges. Bookmark it and return to it regularly.
What if I disagree with some of the advice?
Good. That means you're thinking critically, which is exactly what a good founder should do. Take what resonates, test it, and discard what doesn't work for your specific situation. No advice is universal.
How often is this guide updated?
I revisit and update my guides regularly as I learn new things and as the market evolves. The core principles tend to stay stable, but specific tactics and tools get refreshed based on what's working right now.