0:02 A recent study from MIT came out and it reveals a brutal reality. 95% of AI
0:08 initiatives fail to deliver any measurable value. That means only 5% of
0:14 companies are actually succeeding in replacing human work with AI. So why is
0:19 everyone hyping this idea that we can replace human work with AI? See, this
0:25 made me curious cuz I'm one of the few who actually made it work. Over the last
0:30 year, I reduced my own team from 11 people down to just three. And I
0:35 replaced their work with AI. And while doing that, revenue went up, our output
0:40 increased, and margins got better. So, I have to take some blame here cuz I'm one
0:44 of the people who have been hyping this idea. I have published several videos
0:49 showing you how to build AI agents, social media content writers, SEO blog
0:55 writers, and AI support systems. I am actively contributing to this hype. But
1:00 what I've realized is that replacing your team with AI seems to only work for
1:05 a very specific type of company run by a very specific type of business owner.
1:10 And if that's not you, making this move might be a terrible idea. So in this
1:15 video, I want to take you through three checkpoints. These are filters that will
1:20 help you answer the question, should I replace my team with AI? and if so,
1:26 which team members should go first? If you're new to this channel, welcome
1:30 here. I'm Simon Horberg and my team and I run a SAS portfolio of 5 SAS products.
1:35 We used to be a small team, but now we're an even smaller team thanks to AI
1:40 and automation. So, the first checkpoint is all about the DNA of your company. Is
1:47 your business humanentric or process centric? A human ccentric business is
1:53 run by people for people. The focus here is on culture, collaboration, and goal
1:57 alignment. It's the type of company where you want every team member to feel
2:01 heard. You have a flat structure. You involve everyone in key decision-making,
2:05 and you spend a lot of time ensuring everyone is motivated and personally
2:09 aligned with the company mission. Think of a creative agency, a hightouch
2:14 consultancy, or typical startups where culture fit is just as important as the
2:18 actual skills. In these environments, the value often comes from the friction
2:23 and the chaos of human collaboration. If you try to replace this team with AI,
2:27 you will rip the soul out of your company. AI doesn't care about your
2:31 mission. It doesn't need to feel heard and it's terrible at nuanced cultural
2:37 decision-m. On the other hand, we have the process ccentric business. This is a
2:41 business that runs like a machine. It relies on vicks, guidelines, SOPs, and
2:47 checklists. Here a team member is assigned a specific task with clearly
2:52 defined requirements, a deadline and a set of dependencies. Decisions are made
2:56 at the top by the business owner and the team is there to execute. My business
3:01 has always been heavily processentric. Even before AI, the biggest selling
3:05 point I had when hiring was we have no meetings, we have no ceremonies, you can
3:10 work whenever you want as long as stuff gets done. But the flip side of that is
3:15 we never really had much culture. I rarely involved the team in big
3:19 strategic decisions and honestly I didn't require them to be passionate
3:23 about the vision of the company. My team would care only to the extent they
3:27 needed to understand the context of their work. And this is one of the
3:31 reasons I was able to transition to AI so successfully because an AI agent
3:36 loves a processcentric environment. It loves clear instructions. It loves SOPs.
3:42 It loves checklists. It doesn't need a pep talk and it doesn't need a zoom
3:47 meeting to feel included. So look at your business. Are you running a human-
3:52 ccentric or a process centric company cuz if you are running a high culture
3:57 human first organization you cannot replace your team. You can augment them
4:02 but you cannot replace them. If you're running a processcentric company on the
4:05 other hand you are in a much better position. Let's move on to checkpoint
4:10 number two. This checkpoint is all about you, the founder. Are you a leader or
4:16 are you an operator? We're often told that to be a successful
4:21 CEO, you have to be a great leader. You have to be the person who rallies the
4:25 troop. The one with the high empathy who makes everyone feel heard, resolve
4:29 conflicts, and motivates the team to go above and beyond. Think of people like
4:33 Steven Bartlett, Dan Martell, or Gary Vaynerchuk. These are incredible
4:38 leaders. Their superpowers are empathy, communication, and culture building. But
4:42 the idea that you have to be a great leader in order to run a company is not
4:46 true because on the other side of the spectrum, you have the operator. This is
4:51 the executive who gets things done. They are obsessed with systems, data, and
4:56 logic. They are big on systems thinking and often, let's be honest, a little low
5:02 on soft skills. The ultimate example I can think of is Elon Musk. If you listen
5:07 to anyone who's worked close to him, they aren't exactly describing him as a
5:11 great leader, but he is probably the world's best operator. He cares about
5:15 the physics of the problem, the efficiency of the line, and the output
5:20 of the system. In fact, usually a great operator is not a great leader and vice
5:25 versa. This is why natural leaders often hire COOs or project managers to handle
5:30 the systems and why natural operators often hire team leads to handle the
5:35 people. And this part requires a bit of self-awareness and introspection. I'll
5:41 admit I was never a natural-born leader. I consider myself a strong operator. I
5:45 love building systems. I love optimizing workflows. But I've never enjoyed the
5:50 people management side of things. the motivation, the conflict resolution, the
5:55 emotional alignment, all of this. And generally operators like me end up
5:59 running process-centric businesses while leaders end up running human-centric
6:04 businesses. And this is the key. If you are an operator, AI is your dream
6:09 employee. Because AI doesn't need empathy. It needs logic. It doesn't need
6:15 motivation. It needs instructions. If you're a natural leader who thrives on
6:20 personal connection, you will struggle to manage an AI workforce because you
6:24 cannot motivate an AI. You cannot inspire an AI with a vision. You cannot
6:29 take an AI out for a onetoone to resolve a conflict. So look at these last two
6:34 checkpoints. If you're a leader running a human-centric business, you can stop
6:38 watching right now. You should not try to replace your workforce with AI. In
6:42 fact, you should go watch other videos on leadership and how to become an even
6:47 better leader cuz that's where the true ROI is for you. But if you're an
6:51 operator running a processcentric business, you are in the perfect
6:56 position to replace your team with AI. But even if you have the right business
7:00 and the right personality, you still can't just fire everyone randomly. You
7:05 need to know exactly which roles to replace. So let's say you're running a
7:09 processcentric business and you are an operator at heart. Does that mean you
7:14 can just fire your whole team tomorrow? No, absolutely not. Because even inside
7:19 a perfect system, there are certain roles that AI simply cannot touch yet.
7:24 To figure out which roles you can replace and which ones you need to keep,
7:31 I use a framework I call the RIP model. Honestly, no pun intended. That is what
7:36 I call this framework. R stands for repetitive. This is the lowhanging
7:40 fruit. Is the task identical every single time? Copying data from an email
7:45 to a spreadsheet, formatting a monthly report, categorizing support tickets. If
7:50 a human is doing this effectively like a robot, an actual robot will do it
7:55 better, faster, and cheaper. The P stands for predictable. Is there a clear
8:00 right or wrong answer? Write an SQL query that pulls active users from the
8:04 database. There is a correct way to do that. But reply to our biggest
8:08 enterprise client who is threatening to turn because of a server outage. This is
8:14 not predictable. AI can write a polite apology, but it can't read the room. It
8:19 doesn't know the specific client, and it doesn't know that he needs an honest
8:23 phone call to be saved. AI thrives on formulas. It fails on nuance. But here's
8:29 the real trap. Most founders do get the R and the P right, but they completely
8:35 ignore I. I stands for isolated, and this is where 90% of AI implementations
8:40 fail. Can the task be done in a vacuum or does it require tribal knowledge? I
8:46 use AI to write code every single day. And AI is incredible at writing clean,
8:52 highquality code. But AI assumes that your code base makes sense, which let's
8:57 be honest, most code bases don't. That's why it will randomly refactor code,
9:02 remove old functions, or delete an unused table in your database. It
9:05 doesn't know that you have a random Zapier automation set up 2 years ago
9:09 that reads that specific table in order to run your payroll. This is context. It
9:15 is tribal knowledge. And AI is terrible at this. It doesn't know the history of
9:18 your company. It doesn't know the invisible dependencies and it doesn't
9:22 know why you made a weird decision several years ago. If a task requires
9:26 deep integration with other moving parts of the business or deep historical
9:31 context, it's not isolated. And if you try to replace that role with AI, the AI
9:36 won't just fail. It will probably break things and make a big mess. You can only
9:40 give AI a task that requires no prior knowledge about a specific thing or
9:44 where that knowledge can be included in brief terms upon solving the task. So
9:49 here's the rule. If a role handles tasks that are repetitive, predictable, and
9:54 isolated, you can replace it today. If it misses even one of these, you have to
10:00 at the very least loop in a human as a part of that workflow. So instead of
10:04 thinking about replacing people, think about replacing tasks and apply the RIP
10:09 model to figure out which ones. This is how you build a system. And this brings
10:13 us right back to the brutal statistic from MIT. The reason 95% of companies
10:19 fail to get value from AI is not because the technology isn't ready. It's because
10:24 they are treating AI like a magic employee that can just figure things out
10:29 instead of a machine that follows strict instructions. They are trying to replace
10:33 leadership, tribal knowledge, and strategy with a chatbot. The 5% of us
10:39 who are actually succeeding, we aren't doing that. We are acting as operators.
10:44 We are stripping away the repetitive, predictable, isolated tasks and we are
10:48 handing them over to agents so humans can focus on the things that actually
10:52 require a pulse. Now understanding the theory is one thing. Actually sitting
10:57 down and building real AI agents that pass the RIP test is a different story.
11:02 And if you want to see exactly how that's done step by step, I recorded a
11:06 full breakdown where I built a news research agent. It creates a perfect
11:10 isolated workflow that does the work of a human researcher. And you can watch
