The AI revolution in commerce should be a change of play, but instead, it has become a fast money taking. Wherever it turns, another chatgpt wrap is marketed as the next big thing for cryptography merchants. Promises? “IA IA”, “next generation trade signals”, “perfect agent trade”. Reality? Overvalued vaporware, too expensive and low performance that does not scratch the surface of what is really necessary.
Saad Naja is a speaker at the AI summit during consensus 2025, Toronto, from May 14 to 16.
The AI must be designed to increase the experience of the merchant, not sideways. Companies like Spectral Labs and Creator. They have an excessive dependence on the great language models (LLM) as chatgpt without offering any unique utility, prioritizing AI’s fashion words over the transparency of the substance and the architecture of AI.
AI agents must increase trade
The combination of AI and trade is a transformative leap, so that humans obtain commercial gains more effectively with a powerful forecast, investing less time, but not completely replacing humans of the commercial equation. Merchants do not need another agent without emotions with an agency without restrictions. They need tools that help them trade better, faster and with greater confidence in environments that simulate real market volatility before operating in real markets.
Too many GPT wrappers are rushed to the market with spongy agents already socks that take advantage of fear, confusion and fomo. With large language models (LLM) barely trained and little transparency, some of these “solutions” of the exchange of reinforce the whole and forget the bad habits.
Trade is not just about hyper speed or automation, it is a reflexive decision making. It is about balance science with intuition, data with emotion. In this first wave of agents design, what is missing is the art of the merchant’s trip: its progression of skills, development of unique strategies and rapid evolution through tutoring and interactive simulations.
Only elegant calculators
True innovation lies in developing a metamodel that combines LLM of predictive trade, real -time API, analysis of feelings and chain data, while filtering through Twitter crypto chaos.
Emotion and feeling move the markets. If your AI Trader agent cannot detect when a bullish or bassist community, or in first glance that signal is not a beginning.
GPT wrappers rejecting market movements driven by emotion offer lower risk and rewards within the optimization of the portfolio. A better agent reads nuances, tone and psycholinguistic, as the qualified merchants do.
And although 20 years of high quality commercial data that cover multiple cycles, markets and instruments are a great start, the real domain comes through compromise and progression loops that adhere. The best agents learn from data, people and thrive with coaching.
Better to lose simulated money
Financial systems intimidate most people. Many never start, or explode quickly. Simulated environments help solve that. The excitement of winning, the pain of losing and the joy of recovering are those that generate resilience and changes to change sterile chat and voice interfaces.
The commercial agents of AI should teach this, the subsequent test and simulate trade return strategies in virtual trade environments, not only of successful operations but also of unforeseen events. Think about it how to learn to drive: real growth comes from time to the road and nearby calls, not just read your state’s manual.
Simulations can show operators how to detect candle patterns, manage risk, adapt to volatility or respond to new tariff holders, without losing their heads in the process. By learning through agents, merchants can refine strategies and possess their positions, win or lose.
Before my bags, I wins my trust
The realistic responses of AI agents are rapidly improving to be indistinguishable from human responses through conversational and contextual depth (close the “strange valley” gap). But for merchants to accept and trust these agents, they must feel real, be interactive, intelligent and identifiable.
Personality agents, those who set as real merchants, whether they are cautious portfolio or optimizer managers cautious portfolio, can become trusted co -drivers. The key to this trust is control. Operators must have the right to reject or approve the calls of the AI agent.
Access to chat on request is another lever, together with the visibility of commercial gains and returns on sweat and tears of real merchants. The best agents will not only execute trades, they will explain why. They will evolve with the merchant. They will get access to manage funds only after trying, as an interns who win a seat at the negotiation counter.
The fun and elegant progression of AAA will keep merchants returning in shared experiences opposed to solo missions. Through tokenization and co-learning models, AI agents could become not only tools, but also property assets, solving the problem of Crypto merchant liquidity along the way.
First market players should be seen with healthy skepticism. If merchants’ agents will have a real impact, they must go beyond sterile chat interfaces and become dynamic, educational and emotionally intelligent.
Until then, GPT wrappers remain what are elegant distractions dressed as innovation, extracting more value from the users than they deliver, as indicated by the correction of the AI tokens market.
The convergence of AI and Crypto should empower merchants. With the correct incentives and a merchant mentality, the AI agents could unlock unprecedented learning and profits. Not replacing the merchant but evolving.
