Why I’m building the AI that runs businesses

· 4 min read

For two years I have been writing about reinvention. I sold CommonFloor, felt the strange grief of an exit, and rebuilt myself from scratch in the AI age. People wrote to me about their own version of that. I meant every word.

But reinvention was never the destination. It was the thing I had to do before I could see the real work. Now I can see it, so I am going to tell you what I am actually building.

I think AI is going to run businesses. Not help with a task here and there. Run them.

Here is the distinction that matters. For thirty years, software meant tools that people operate. You bought a CRM, a transaction manager, an accounting package, and then you hired people to sit in front of those tools and do the work. The software was a better filing cabinet. The human was still the engine.

That is ending. What is new is not a smarter filing cabinet. It is software that does the work itself and hands you the decision. The agent reads the contract, chases the signature, flags the deadline, reconciles the numbers, and comes back with: here is what I did, here is what I need you to approve. You stop operating the tool. The tool runs the business, and you run the relationships.

I call what I build an AI-native business operating system. One agentic core, the same agents, memory and workflows underneath every business. Then it fits an industry through a vertical pack, and it fits your specific business through configuration, not a six-month rebuild. It is an operating system because the core is shared and standard. It fits your business because the pack and the setup are yours. Those two things are usually in tension. Getting them to sit together is most of the work, and it is the part I find most interesting.

I am proving it where the operational load is heaviest first: US real estate. A real estate deal is a small mountain of coordination. Contracts, disclosures, compliance, deadlines, the back and forth between agent, client, lender, title and the other side. Most of that is work nobody became an agent to do. Through ReBillion.ai, I am putting AI on that work, so the agent spends the day on the two things that actually win in real estate: the client and the deal. That is the flagship, and it is where the system is proven deepest.

The same core now runs early in two industries that look nothing like real estate on the surface. A cafe business. A practice business. Different customers, different front office, same bet underneath: let AI run the operational core, and let the people run the relationships. These are early, and I am not going to dress them up with numbers they have not earned. The reason they matter is that they are evidence the core generalizes. One system that travels across industries is a very different thing from three clever tools that each work in one place. The whole thesis is that it travels.

Why now, and why me. Now, because the models finally crossed the line from drafting something you finish to doing the thing and reporting back, and because the hard part was never the model anyway. The hard part is the business: the workflows, the edge cases, the trust, the handoffs, the memory of how a real operation actually runs. That is founder work, not demo work. And me, because I have built the unglamorous back-office product before. CommonFloor, underneath the marketplace, was a system for how property gets bought and sold. I have spent my career in the machinery of how a business actually runs. This is the same instinct with a far more powerful engine.

A fair question: is this not just automation with a new name. No. Automation follows rules you wrote in advance and breaks the moment reality does not match the rule. What is different now is that the system can read an ambiguous situation, decide, act, explain itself, and get better at your specific business every week. The honest version of the promise is not that AI runs your whole company while you watch. It is that AI does the work, you review it, and you own the relationships. I will not oversell it. Overselling this is how you lose the trust you need to actually deploy it.

So that is the reinvention. Not a new personality. A new thing to build. I spent two years learning how to start over. I started over into this.

If you run a business and you are tired of your best people spending their hours on operational work a system should be doing, that is exactly the problem I am building for. I will keep writing here as I build it, with the wins and the misses both, because that is the only way I know how to do it.

I am a four-time founder building AI-native companies. I build the operating systems that run businesses, and I write about the founder mindset, AI opportunities, and the mechanics of building with AI at vikasmalpani.com.