Everybody wants the beach muscles. The big AI tool. The flashy demo. The one you flex in the next board meeting.
Looks great in the mirror. But nobody asks the real question. Do you have the bones and tendons to hold it together?
Muscles on a body with no skeleton are not strong. The first time real weight lands, it folds. Your AI tools are the muscles. And a muscle with nothing solid to pull against just flails.
I have been writing about the Business Body. An AI-native company is built like a body, not a pile of software and tools. It has six parts. A Skeleton, which is your data. Muscles, which are your tools. A Central Nervous System, which moves work across the company. A Brain, which is human judgment. An immune system, which catches mistakes. A Circulatory System, which carries the same meaning to every part.
This is part one. We start where the body starts. Not the muscles everyone wants. The frame nobody thinks about. The Skeleton. Your data. The bones and tendons that hold everything up.
Here is what I see. A CEO buys an AI tool. It hands back garbage. They think the tool is broken. The tool is fine. Their bones are broken.
AI cannot stand on a mess. Ask it about a customer, and it goes looking. If your customer lives in six systems with six definitions, AI has nothing solid to stand on. So it guesses. It sounds sure. It is wrong. Confident and wrong is the worst answer a machine can hand you.
Let me make it real. Ask three of your systems what active means.
Billing says active is paying.
Product says active is logged in this week.
Support says active is not yet churned.
Three answers for one word. Now point AI at that and ask how many active customers you have. Watch it pick one and call it the truth.
Do the same with revenue.
Sales count a signed deal.
Finance counts cash in the bank.
Product counts seats turned on.
Same month, three numbers, all called revenue. Nobody is lying. They are each grabbing a different bone for the same word. You are not just confusing the AI. You are running the company on three versions of the truth.
The Skeleton is your data model. Where your data lives. How it is shaped. What a customer is, what a product is, and what active means are written down once and the same everywhere. And not just the words. How they connect. A customer has products. A product has a plan. A plan has a price. A ticket belongs to a customer. Those connections are the tendons. They tie the bones together so the body holds as one piece, not a bag of loose parts. AI follows the connections you give it. Bad connections, bad answers, every time.
Here is the part that should make you sit up. This thing I keep describing, the things in your business and how they connect, has a name. It is called an ontology.
Palantir built a fortune on it. They define an ontology as a map of your organization. The real things, customers, products, orders, and how they all connect. They call it a digital twin of your company. My colleague DR, also known as Danielle Rios, says it simpler. She runs an AI-first telco software company. An ontology is the nouns and verbs of your business. The nouns are customer, product, plan, and invoice. The verbs are sign up, upgrade, cancel, and bill. Write those down so a machine can read them, and your AI finally has ground to stand on. DR says it best. AI without context is just expensive pattern matching. You can follow her here; she has the goods on AI, just like the AI Boss.
Now the real forehead slap. This is not new. Biologists hit this exact wall in the late 1990s. With thousands of labs, each calling the same gene by a different name, none of their data can talk to each other. So they sat down and built a shared map of what every gene is and how they relate. They called it the Gene Ontology. Almost thirty years ago.
So when I tell you to define your things and write down how they connect, that is not busywork. That is building an ontology. The skeleton has a name.
Want to know why you do not already have one? Your data was built one department at a time. Sales built their view of the customer. Support built theirs. Finance built theirs. Nobody built the company's view, because there was no such thing. Departments are a constructed fallacy. They are walls we put up to fit the work into human heads. Your data inherited those walls. Now AI has to climb every one of them.
Here is the trap smart people fall into. They hear all this and go buy a data warehouse. They dump every system into one big pool and call it solved. It is not solved. A warehouse is where data sits. An ontology is the agreement about what it means. You can put every bone in the body into one bucket, and you still do not have a body. You have a bucket. The work is the agreement, not the storage.
They skip the agreement. They tell themselves the AI is smart enough to sort out the mess. It is not. Smart does not fix structure. The smartest brain in the world still needs a spine to sit on.
You would not hang drywall before the studs are up. But buying tools before you build your ontology is exactly that. Muscles bolted onto nothing. And here is the part nobody wants to hear. No consultant can hand you a skeleton. They can sell you a tool. They can sell you a dashboard. The bones have to be yours, because only you know what your business actually is. Don’t outsource this work!
Here is the proof I trust most. AI handles 82 percent of our support. People hear that and picture a magic bot. The reason it works is the bones. When a customer writes in, the AI already knows who they are, what they bought, which plan they are on, and the shape of their account. One answer for each, the same answer every other system would give. The bot is not smart. The ontology is clean. Put that same bot on a broken skeleton, and it would hand your customers garbage, too.
Here is what changes when the bones are right. You ask one question and get one answer. You add a new tool, and it works on the first day because it grabs the same structure that everything else does. A good skeleton is quiet. You only notice it when it is missing.
Most of you do not run a telco or 24 products. You run a software company or a services company that runs on software. Same skeleton problem. Your customer lives in the CRM. Their usage lives in the product. Their bill lives in billing. Their tickets live in support. Four systems. Four ideas of who the customer is. You never wrote down one. So when you bring AI in, it has nothing to stand on either.
How do you start? Not with a giant project. Pick the handful of things your whole business runs on. Customer. Product. Plan. Price. Status. For each one, write down what it means in plain words. Then pick one system to be the home of that answer. Every other system points to it. You do not need every bone on day one. You need the few that carry the most weight.
Here is your figure-it-out moment. Pull up your systems and find one thing. A customer. Count how many places define it, and how many of those definitions agree. If the answer is more than one place and less than full agreement, you do not need a smarter tool. You need an ontology. You need bones.
We start with the Skeleton because nothing stands without it. An ontology has two halves. The things and how they connect, the bones and the tendons, and that is today. The rules that ride on top of them are the blood and the Circulatory System, which are next in this series. But blood needs bones first. So before you buy one more tool, go count your definitions. What is your data actually standing on?
Kathy
THE AI BOSS