Walk into a mid-sized e-commerce company on a Tuesday afternoon, and you can usually spot the support team without asking. They are the ones with twelve browser tabs open, two monitors, a Slack channel that pings every forty seconds, and a posture that suggests they have not stood up since the morning standup. One agent is answering a question on Instagram. Another is on hold with a payment processor while a customer waits in live chat. A third is hunting through Gmail for an order confirmation that may or may not exist.
Every executive in that building knows the situation is unsustainable. Most of them have already bought an AI tool to fix it. A surprising number have quietly shelved that tool six months later, telling nobody outside the office why.
Mikhail Antipkin has spent the past two years talking to those executives, and he has come to a conclusion that runs counter to most of the noise in the AI category. The technology is not the problem. It almost never has been.
"People want to blame the model," Antipkin said. "The model is fine. The model has been fine for a year. What goes wrong is everything around it."
Antipkin is the founder of Vivo Chat, an AI-powered customer communication platform built to unify Facebook, Instagram, WhatsApp, and other channels into a single interface. He is also the chief executive of an international technology group with offices in Hong Kong, Dubai, Limassol, and London, and more than 1,100 employees across the four. He started his first business in 2003 in telecom retail and has spent the past two decades building things and scaling them, including a ten-store shoe chain that moved 40,000 pairs a month at its peak and a microfinance company that grew to fifty offices before he exited in 2016. He is, by any reasonable measure, an operator. The reason that matters is because operators tend to talk about AI differently than consultants do. They talk about it the way someone who has had to make payroll talks about anything: with specifics.
The Failure Modes Nobody Writes About
Antipkin lists the reasons enterprise AI deployments stall the way a mechanic lists the reasons an engine misfires. There are usually three or four of them. They are almost always the same ones.
The first is data fragmentation. "You cannot put an intelligent agent on top of a stack that does not talk to itself," he said. "Companies have a Facebook tool, a WhatsApp tool, a chat tool, a CRM, a help desk, and they are all storing customer history in different places. Then they bolt an AI on the front and wonder why it gives generic answers. It gives generic answers because that is all the data you gave it."
The second is what he calls the demo trap. A vendor shows a polished demo on a clean dataset. The buyer signs. The implementation team plugs the system into the actual mess of the company's real conversations, and the quality drops by half. Nobody is to blame, exactly. The demo was honest. The mess was hidden. The fix takes months, and the patience for it usually runs out at week six.
The third is the part Antipkin sounds most exercised about. "Companies treat the agent like a piece of software you install," he said. "It is not. It is a colleague you train. If you do not give it feedback, it does not improve. If you do not invest in the training, it stays mediocre. The companies that get this right are the ones that put a real person in charge of teaching the system, every day, like a manager. The companies that fail are the ones that bought it and walked away."
None of these failure modes is about the underlying intelligence. They are about how the technology gets dropped into a real organization with real history and real people. Which is, in Antipkin's telling, both the bad news and the good news.
"It is bad news because most of the deployments you are reading about are going to disappoint," he said. "It is good news because the fix is not exotic. It is just unglamorous. The companies willing to do unglamorous work for a quarter or two are going to look like geniuses by the end of the year."
What It Looks Like When It Works
Antipkin is more comfortable describing the failure cases than the success ones, partly because he is wary of sounding like a salesman and partly because the success stories are quieter than the disasters. They tend to look like nothing happening for several weeks, followed by a slow and unmistakable shift in the numbers.
He offered one example, with the company's name kept out of it at his insistence.
"Software company, headquartered in London, sixty employees," he said. "They had Brazilian customers who were churning at twice the rate of their British ones. Not because the product was bad. Because nobody was answering them in their language at a useful hour. The London team was good. They just could not be awake when São Paulo needed them."
The fix was not a hiring round in Brazil. It was a configured agent layer that could answer in fluent Portuguese, in the company's brand voice, at three in the morning local time, with full context on every customer it spoke to. The implementation took about a month. The Brazilian churn rate dropped by half within eight weeks. Nobody in London worked a different schedule. Nobody got fired. The team in London actually expanded six months later, because the Brazil revenue made it possible to hire the senior engineer they had been wanting for a year.
"That is the part of this story that gets missed," Antipkin said. "Everybody wants to talk about cost reduction. The interesting story is the businesses that finally get to compete in markets they were locked out of, and what they do with the money once they win those markets. In every case I have seen, they reinvest in their people. The agent is not replacing the team. It is what makes the team affordable to grow."
Three in the Morning in São Paulo
The São Paulo story is not unusual in Antipkin's portfolio. It is, in his telling, the most reliable pattern he has seen in two years of watching companies deploy unified messaging with an AI layer.
Multilingual support has been the white whale of mid-market expansion for as long as anyone can remember. A British company that wants to sell into Brazil has historically had three options. Hire Portuguese speakers in São Paulo, which is expensive and operationally complicated. Outsource to a contact center in Lisbon, which works until it does not. Or simply ignore the market and tell the marketing team to focus elsewhere. Most companies have picked door number three, not because they wanted to, but because the math never worked.
The math is starting to work.
"I am watching companies that were locked out of half their natural market suddenly walk into it," Antipkin said. "And not just into Brazil. Into Mexico. Into Indonesia. Into Vietnam. These are countries with serious purchasing power and very few mid-market British or American companies serving them properly, because the support economics never made sense. They make sense now. The companies who figure that out first are going to look like they got very lucky in 2026. They did not get lucky. They did the work."
The Quiet Window
Antipkin is careful with predictions. He has watched too many of them fail in the AI category to want to add another one to the pile. But pressed on what the next eighteen months look like, he offered something closer to a window than a forecast.
"Right now, you can still get a real head start," he said. "The technology is good enough. The integrations are clean enough. The cost per conversation is low enough. The only thing missing in most companies is the decision to start. That is going to be true for maybe another year, probably eighteen months. After that, this becomes the baseline. Nobody wins on a baseline. You only win during the window when everyone else is still arguing about whether to do it."
He was emphatic about one thing. The window does not belong to the largest companies. "The Fortune 500 are going to take three years to deploy this properly because of how they are organized," he said. "The companies that move first are going to be the mid-market ones. Sixty employees, two hundred employees, five hundred employees. They are nimble enough to make a decision on Tuesday and ship it on Friday. That is the entire game right now. The advantage is not capital. The advantage is speed."
It is a notable framing, coming from someone who runs a company with more than a thousand employees of his own. Antipkin is aware of the irony and waved it off. "My group is structured in pieces, and the pieces move fast. That is the whole point. If you cannot move fast at this moment, the size does not save you. It just gives you more to lose."
The Operator's Bet
There is a particular kind of confidence that comes from having built things before. It does not sound like the confidence of a founder pitching investors. It sounds quieter, more specific, more interested in the details than in the story. Antipkin has it, and it shows up most clearly when he talks about why he chose customer communication as the place to plant a flag in the AI cycle.
"I have built a lot of businesses," he said. "Retail. Real estate. Microfinance. Payments. Games. Each one taught me something different about what scales and what does not. Customer communication is the one that scales the cleanest, because the value of the product compounds with every conversation. The agent gets smarter. The team gets sharper. The customers get better answers. Three things improving at the same time, in the same loop, is not something you see often. When you see it, you build there."
It is the closest thing to a thesis he offered in a long conversation that mostly avoided thesis statements. Asked whether he was confident the bet would work, he did not hesitate.
"I am more confident in this one than I have been in any business I have built before," he said. "The conditions are right. The technology is ready. The market is asking for it. And the team we have built across four offices is the best I have ever worked with. If you are an operator and you are paying attention, this is the moment. I would not be spending my time on it otherwise."
About the Source
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.
ⓒ 2026 TECHTIMES.com All rights reserved. Do not reproduce without permission.
