The loudest voices in the AI conversation have rarely run an operation. After two decades building companies across retail, real estate, fintech, and software, I have come to think that is exactly why so much of the advice is wrong.
I have sat in a lot of rooms over the past two years where intelligent people explained artificial intelligence to me. Most of them had never had to make payroll. You can tell within a few minutes, because they talk about the technology as if it were the whole problem. An operator never does. An operator knows the technology is the easy part.
I started in telecom retail in 2003. In 2005 I built a retail business into a chain of ten stores moving up to forty thousand pairs of shoes a month. A few years later I was running a commercial real estate project and a microfinance company that grew to fifty offices before I exited both in 2016. Since 2017 I have built international businesses in payments, software, and gaming, and today I run a group across four cities with more than eleven hundred people. I am not listing this to impress anyone. I am listing it because every one of those businesses taught me the same lesson, and it is the lesson the AI conversation keeps missing.
The Model Was Never the Problem
People want to blame the model. The model is fine. The model has been fine for a year. What goes wrong is everything around it, and everything around it is an operations problem, not a technology problem.
I watch companies buy an AI tool, bolt it onto a stack that does not talk to itself, and then wonder why the answers are generic. The answers are generic because that is all the data they gave it. I watch them sign after a clean demo, plug the system into the real mess of their actual conversations, and lose half the quality overnight. Nobody lied. The demo was honest. The mess was hidden. And I watch them treat the agent like software you install, when it is closer to a colleague you train. If you do not give it feedback, it stays mediocre. The companies that get this right put a real person in charge of teaching the system every day, the way you would manage a new hire. The companies that fail bought it and walked away.
None of that is exotic. It is just unglamorous. The companies willing to do unglamorous work for a quarter or two are the ones that look like geniuses by the end of the year.
What Operators Know That Consultants Do Not
There is a habit of mind you only get from running things. You stop being impressed by launches and start caring about the second order. A retailer placing your product is a hypothesis. A retailer placing a second order is a verdict. A demo that dazzles is a hypothesis. A system that is still improving in month six is a verdict. Operators live in the second column.
That is why I am skeptical of most AI advice that arrives without a P&L behind it. The interesting question was never whether the technology is impressive. It plainly is. The question is whether the organization around it is built to keep feeding it, correcting it, and holding someone accountable for its output. That is a management question. It has been a management question the entire time.
The Window, and Who It Belongs To
Here is the part operators are reading correctly while the category argues with itself. The advantage right now is not capital. It is speed. The technology is good enough, the integrations are clean enough, and the cost per conversation is low enough that the only thing missing in most companies is the decision to start.
That favors the mid-market, not the giants. A large enterprise will take three years to deploy this properly because of how it is organized. A company of sixty or two hundred or five hundred people can decide on a Tuesday and ship on a Friday. I say this as someone who runs a company with more than a thousand employees of my own. My group is built in pieces, and the pieces move fast. If you cannot move fast at this moment, size does not save you. It just gives you more to lose.
I built my offices across Hong Kong, Dubai, Limassol, and London for a future I thought was five years away. It turned out to be about eighteen months away. The window where moving early is cheap and moving late is expensive does not stay open long. I have seen a handful of these moments in twenty years. They are always obvious in hindsight and always contested while they are open.
The Operator’s Bet
I chose customer communication as the place to commit because it is the rare thing that compounds in three directions at once. The agent gets smarter, the team gets sharper, and the customer gets better answers, all in the same loop. You do not see that often. When you see it, you build there.
If you are an operator and you are paying attention, this is the moment. Not because the technology is magic, but because it finally is not the hard part. The hard part is the work, and the work is the thing operators were always good at.
Mikhail Antipkin is the founder of Vivo Chat, an AI-powered customer communication platform, and the chief executive of an international technology group with offices in Hong Kong, Dubai, Limassol, and London. His ventures span payment systems, software development, and gaming, with a combined turnover exceeding $100 million. He has been building and scaling businesses since 2003 and currently leads a team of more than 1,100 employees across four continents.
Explore how Vivo Chat is building the future of AI customer communication at vivochat.ai
