Getting AI to work in your business
Why AI Isn't Working in Your Business (and How to Fix It)
On this page
TLDR
It's not that AI won't work for your business. There are four things most business owners don't see: there's no real problem for AI to solve, it's doing a job automation does better, the process only lives in your head, or the system never got finished. Fix those and you get way more done with AI.
You've tried AI in your business. A few chats in Claude, maybe an automation or two. And it hasn't really changed anything. So you start to wonder if AI just doesn't work for a business like yours.
It does. When AI stalls, it's almost never the tools. It's one of the four things below, and every one of them has a fix. Each links to a deeper article if it's the one you're stuck on.
#Why most AI projects fail
Not because AI is bad. It's actually really good at tons of things. It's because nobody decided what problem AI was supposed to solve.
According to RAND, by some estimates more than 80% of AI projects fail. And the number one reason? People don't actually understand the problem they're trying to solve. It's a lot like hiring a new employee and not having a job for them. They show up Monday and ask, "Hey, what do you want me to do?" And you say, "I don't know, just help make things better."
On client calls, I keep hearing people say they have to get AI into their business. So I ask, "Okay, how would you like to use AI?" Want to know the answer almost every single time? "I don't know, but we need it." They see other people in their industry using it and they don't want to fall behind. That right there, my friend, is the mistake. They just have a fear of missing out.
So here's what I do. With every single client, I start with two questions:
- What are you optimizing for? Are you trying to save time, save money or make money? Those are different problems.
- What are the top three things that cost you the most in that area? If it's time, what three tasks take the most time out of your week? If it's money, what are you spending the most on that a system could probably handle?
You don't start with what's exciting or what you saw on a YouTube video. You start with the thing that's actually costing you the most, and you build the solution for that. And almost every time, it's not the exciting stuff.
#Why AI agents aren't the answer for every task
Because AI in the wrong kind of task makes your system less reliable, not more.
Every task in your business is one of two things: clockwork or brainwork. Clockwork runs on logic you've already figured out. It could be simple, like someone signs up for your email list and gets a welcome sequence. Or it could have a bunch of branches: a payment fails, it retries in 3 days, and if it fails again it flags the account. Either way, every path is mapped out, so no one needs to think. It just happens.
Brainwork is the opposite. It takes real judgment. A lead comes into your pipeline, so you look at their information, research them, score them on the answers they gave you, and then decide if they're qualified.
AI is great for brainwork. But for clockwork, automation is way better.
For the last 4 years, I've built the systems for the Forward Event, a 1,000-person live event. This year we're moving everything over to Airtable. Someone on the team found the AI agents built into Airtable and came to me like, "Hey, we should totally use these this year." And a bunch of agents working your event sounds pretty dope. But there's one problem. The Forward Event is all clockwork, and it needs its systems to be reliable. If we put AI agents into that system, it would make everything way less reliable.
Think of it this way. If I want to turn on a light, I flip the switch. Light comes on. That's clockwork. Now imagine I replace that switch with a person whose job is to turn the light on when I walk in the room. Most of the time it's fine. But once in a while, they decide that since the sun's coming through the window, you don't actually need the light on. That would be freaking insane. You don't need a person to turn on your lights. And you don't need an AI agent in your clockwork tasks either.
#Why AI gives wrong answers about your business
Because it's guessing. The task isn't written down, so Claude fills in the blanks on its own.
When a task isn't written down clearly, or it's out of date, a person would usually stop and say, "Hey, I don't know what I'm doing here." Claude doesn't really do that. Sure, in a chat it'll sometimes ask you a question. But when it's running a task on its own, nobody's there to answer, so it guesses and keeps going. Even OpenAI's own research says the way models are trained rewards guessing over saying "I don't know." So you don't just get bad results. You get confidently bad results. And you don't even realize it until something actually goes wrong and blows up.
My wife is an incredible cook, but there's one problem. She uses no written recipes. She just feels it, and everything she makes just works. That's all fine and dandy until I say, "Hey, let me help you with dinner." When I ask how much to use, she'll say, and I quote, "Till it's enough." Or "a dash." And I'm sitting there like, dude, how much is enough? And is it my dash or your dash? So I don't actually help. It usually ends up under-flavored and she has to come back and fix it anyway. The recipe is in her head, not mine. And that's exactly what happens when you try to get Claude to do something in your business without the task written down. You're handing Claude half a recipe.
So you need a blueprint. Think of it a lot like a recipe for how something in your business gets done. Get the process out of your head and into a blueprint. The single fastest way I've found is to record yourself doing the actual task in Loom, or whatever you want. Go step by step and explain your thought process, why you're doing certain things and why you're not doing certain things. That "why not" part actually helps Claude a lot. Then I take the transcript and throw it into a Claude skill I built, and it turns it into a clean step-by-step blueprint. If anything's vague, it asks you clarifying questions, so you don't have to be perfect when you record.
Free Claude skill
Give AI the blueprint it needs to do work without you
Record yourself doing the task once. This skill turns the transcript into a blueprint Claude or Codex can follow.
#Why AI automations never get finished
Because nobody makes you finish them.
When you're doing something yourself, nobody stops you from jumping to the next thing when stuff actually gets hard. Nobody says, "Finish this first."
I know this because I built a junk drawer first. I had five or six builds going in Claude Code at the same time: a client pipeline, onboarding, lead capture, a task tracker, a content system. Want to know how many of those were actually running by the end? One. The content system, because it was the only one I forced myself to finish. Every time the others got hard, I'd watch the next YouTube video and start something else. I'd tell myself, "I'll come back to this." I normally didn't. Every single time I opened up Claude Code, it looked like a junk drawer. Tons of skills half built and never used.
I was sitting there thinking AI and automation just weren't going to work for my business. That wasn't the truth. I just hadn't ever finished anything. When I got that content system done, it kind of slapped me in the face. Finishing is what pays. Back when I was running my YouTube agency, finishing our content system helped us put out over 1,600 videos for our clients. The tools were never the problem. I never stuck with a project long enough for it to actually work.
One system fully built and fully running does way more for your business than five that are half built.
#Questions people ask
Should I hire someone to set up AI or do it myself?
If you want it finished, hiring helps. MIT found that AI tools companies build internally only get deployed about 33% of the time. Companies that bring in outside partners doubled that, at about 67%. Now, that's companies reporting on themselves, 52 of them and mostly big ones, so take it with a grain of salt. But I think the reason is simple. When you bring somebody in and pay them, there's an expectation that things get done.
What should I automate first in my business?
The boring back-end stuff. MIT found about half of AI budgets go to sales and marketing, but the biggest savings came from back-office work like customer service, document processing and cutting agency costs, and it often paid back faster. It's the stuff nobody makes YouTube videos about, but it's what runs your business. Get your foundation strong first, before you throw gasoline on the fire.
