AI in the crypto industry: an accelerator, not a replacement for humans — analysis of real-world applications
Artificial intelligence in the crypto sector has not yet become an independent player. Even code writing and exchange monitoring remain under strict human control. This is not a futuristic hypothesis, but a practical conclusion based on the real-world experience of leading industry experts.
My interlocutors — Polina Demchuk, CEO of TradeSanta and a blockchain consulting expert since 2017, and Nikita Kutsenko, a financial manager and crypto analyst — agreed: AI today is an accelerator, not a replacement. Companies have already delegated two key functions to neural networks: content marketing and accelerating development. In marketing, AI gathers news, analyzes TikTok trends, and generates videos. In development, it monitors exchange API changes and writes code. But the final decision always rests with a human.
How AI is integrated into workflows
Polina Demchuk emphasizes: "We use agents to monitor API changes and write code, but this is to speed up development work, and it is still under the control of developers." Nikita Kutsenko goes further: his morning ritual is to open Claude, upload data on sentiment, key news, Bitcoin and Ether prices, and the Fear and Greed Index. "It outputs on one page the direction — long or short — and the reasoning. I don't blindly follow it: I cross-check its picture with mine. When they align, it's a very powerful signal. When they diverge, I dig deeper to figure out which of us is wrong," the analyst shares.
The set of tools is selected through trial and error. In development, it's VS Code with Codex and Claude Code. For video generation, it's Kling and Eleven Labs; for landing pages, it's Lovable. The choice, according to Polina, is based on the quality-to-cost ratio: "We use Kling and Eleven Labs for video generation because, out of everything we've tested, we get the best quality and a reasonable cost per unit of finished material."
Boundaries of trust: risks and strategies
There have been no serious AI errors that would have cost companies dearly. The reason is that a human is always in the loop. Regarding trusting an AI agent with real trades, Polina Demchuk considers it possible as a test, but sets boundaries through the lens of risk: "It's a question of how much I'm willing to lose, what drawdown I'm willing to accept, and this determines both the amount I'm willing to entrust to the agent and the access that agent gets to my data."
Nikita uses several AI tools, distributing tasks on the principle of "each covers its own area." ChatGPT, in conjunction with CoinGlass, processes data on open interest, liquidations, funding rates, and the long-to-short ratio — instead of an hour of manual compilation, a market picture is obtained in thirty seconds. Grok, integrated into X, monitors crypto Twitter in real time: early signals and leaks appear there. Claude handles the strategic direction for the day. "The result: analysis takes not a couple of hours a day, but 15 minutes," Kutsenko summarizes.
My conclusion: AI in crypto is a powerful but still auxiliary tool. It saves time, automates routine tasks, and provides an alternative perspective, but critical thinking and control remain with the human. Until neural networks learn to take responsibility for risks, full trust in them in this volatile environment is an unaffordable luxury.