I counted last week. 137 Wednesdays.
That is how many Wednesdays have passed since November 2023, when my whole team started building with AI together. Every session runs the same way. Time to talk about what is new, then hours of building, then demos. Nobody watches. Everybody builds. Cameras on. Hands on keyboards. The newest hire and I are in the same tools, hitting the same walls. We took two months off when the fatigue hit. More on that in a minute. Run the math. Over 10,000 team hours on the tools.
That number stopped me. People ask me how we did it. They want the tool list. The tool list does not matter. The tools change every few months anyway. The hours are the answer. AI is a muscle, not a project. You would not train for one marathon and call yourself a runner.
I added it up because I am giving a talk to a room of CEOs about building an AI native org. And when I laid the whole story out end to end, I realized something. I have never given you the whole picture in one place. You have gotten pieces. The pyramid. The hackathons. The support numbers.
So here it is. The total overview. And do not worry, folks. Each piece of this is coming out as its own issue in a series. Today you get the map.
The Spiky Point of View
Departments are a constructed fallacy.
We did not build departments because they are the right way to run a company. We built them because one human brain can only hold so much. When the work got too big for one person, we split it. Then we hired managers to manage the split. Then we hired coordinators to manage the managers. Finance over here. Support over there. Engineering behind a wall. Every box on your org chart exists because of a cognitive limit, not because that is how a business is supposed to be structured. Think about it. No two companies have the same exact org structure!
AI removes the limitation. Learning another domain used to take months. Now it takes an afternoon to get up to speed and a month to become an expert. One person with AI can hold what used to take four specialists. So AI will not just flatten your org. It will narrow it. Fewer boxes. Fewer walls. Fewer handoffs. Some of you breathed a sigh of relief, and some of you just became more tense from that last sentence.
What AI runs at my company today
We did not start with the fancy stuff. We started at the bottom of the pyramid, getting rid of repetitive tasks. The boring wins. That is how you earn your way up to the bigger swings. Three years later, here is where those swings landed.
82% of our support tickets close with no human touching them. A human steps in when the AI flags something it should not handle alone. The rest just gets done.
Our AI renewals team works a real book. Outreach, follow-up, and close. Renewals was the right place to test because there is nowhere to hide. The renewal closes, or it does not. We ran the AI head-to-head against our own people. The AI closed 8% better. Not cheaper. Better. That is the first month of data. Month two is around the corner, and I will share it either way. Renewals go to AI.
Next up is engineering. AI already writes features and fixes bugs across many of our products. We cut 3 million lines of code down to 125 thousand along the way, in just one product. The next step is the whole function. A few humans are in the loop, but AI writes the code. AI watches production, including incident management, not a separate team. Humans do the part that is actually hard. Judgment. Architecture. The calls that matter.
And it does not stop there. Finance, legal, marketing, sales, data. AI runs real work in every one of them. Board prep used to take a full day. Now it takes maybe 30 minutes, don’t tell my board that…
The future org
Now, the part I have not shown you before. Where is all of this heading?
Think about how a typical software company touches a customer today.
Support.
Customer success.
Account management.
New sales.
Product management.
Five groups, all talking to the same customer, handing that customer back and forth and dropping things in between.
Then look at the technical side.
Product management again.
Engineering.
SaaS ops.
DevOps.
Design.
More handoffs. More drops.
And then those two big groups have to talk to each other. That is where the real pain lives. I can tell you exactly where deals stall and where customers churn. It is not inside any one team. It is in the handoffs between them. Every handoff is a place where the customer's words get lost.
In my future org there are three groups.
Customer Management: Support, success, accounts, new sales, and renewals. One team. And every person on it is a product manager. They can log in. They can demo. They can help a customer with the product on the spot. They can tell a customer what is on the roadmap and why. They do not pass the customer to someone else. They are someone else. They make sure the customer is successful. They make sure the customer is getting value. They make sure the customer renews. Successful customer = Renewed Contract.
Product: Product, engineering, ops, design. One build team.
Operations: Finance, legal, HR. It backs the other two and keeps the trains running.
And here is the part people miss. The customer manager does not even need to carry the customer's problems back to the product team. The AI recorder on the call already captured it, summarized it, and wrote the ticket. The customer's voice gets to the product team without a single meeting about the meeting.
Ten teams become three. The handoffs disappear.
Notice that the product function is in both the Customer Management Team and the Product Team. I have a whole newsletter coming on that topic. Stay tuned.
What went wrong
It was not clean. Three things nobody puts on a slide:
I top-graded people out. I had people who were good at their jobs and would not pick up the tools. I gave them training. I gave them all the Weds.ai AI sessions. Some still would not move. So I moved them out. You are going to have to make hard decisions on this. Top grading is one of them. But here's what I learned. Your team will never out-AI you. You set the ceiling. If the CEO is not in the tools, nobody has permission to be. Keeping people who pull the ceiling down is a choice, too. I believe it's called setting the bar.
We burned out, I burned out. AI fatigue is real, and pretending it is not will break your team. Every week brought a new tool, a new model, a new way of working. People got tired, I got tired. We stopped for two months, and I let us stop. It was the right thing to do at the time.
Then it flipped. In the last few months, my team started asking me for more time with the tools, not less. Nobody asks for more meetings. They ask for more build time. That is how I know the DNA of my team craves learning and doing.
That is what 10,000 hours looks like. Never a straight line, and I still am not at 10,000 hours if I do the math.
The wall I hit next
Here is the part I am still living. Everyone and every team got faster, but the silos seemed to get worse.
Our support AI does not talk to our finance AI. Our finance AI does not talk to our engineering AI. Each one is great at its job and blind to everything else. Then I asked a simple question. Which customers cost us the most to serve, why, and is the fix already on the roadmap? Support knows the tickets. Finance knows the cost. Engineering knows the roadmap. Nobody could put the three together. Not the people. Not the AIs. That was my figure-it-out moment. We built prosthetics. Not a body.
So the next build is a business operating system. Six systems, built like a body. The skeleton is your data model, the structure everything attaches to. The circulatory system is context, the shared state that flows everywhere. This is the one almost everyone skips, and then they wonder why their AI tools do not work together. The muscles are the AI tools that do the work. Swap them out as better ones show up. The nervous system coordinates across functions. The brain is human judgment. Still me. Still you. The immune system catches problems before they become a crisis.
I have not finished this. Three systems built. Three to go. I am building it right now. For the record, you cannot outsource the thing that changes everything. No consultant knows your business as you do. I get asked this question at least once a week from people outside my org, and I always say the same thing. No. Absolutely not. Don't do it.
What is coming in the series
The per head economics. The renewal test. Full AI engineering. The future org in detail. The body, system by system. And the mess, because the mess is where the learning is.
One issue at a time. Starting soon.
So here is your homework. Count your boxes. How many teams touch your customer before that customer gets what they asked for? Write the number down. Then ask what would happen if it were one.
Stop learning about AI. Start doing AI.
Kathy
THE AI BOSS