All in the space of a few weeks, Anthropic unveiled new agents for finance, Circle launched nanopayments, MoonPay launched a payment card for agents, and Gemini launched agent trading, signaling that the agent finance battle is here. While the products are new, the underlying business model remains the same. Every exchange and brokerage earns more when clients trade more, and the data on what it does to client portfolios is unequivocal. Ultimately, agent rails have arrived faster than incentives have changed.
The perverse incentive exchanges hope you miss
The conflict is structural for the industry. Brokerages and exchanges don’t need clients to win, they need them to keep trading. Crypto exchanges and neo-brokers made trading faster, cheaper and frankly more addictive. The commercial reality is that banks profit when you stay, exchange profits when you trade, and AI models profit when you ask. The agent you can trust with your hard-earned capital sits outside all three. An independent agent is paid only when the client’s portfolio wins, threatening the current incentive structure for brokerages and exchanges.
The truth is that zero commission trading is not free. In 2025, US market makers paid more than $4.9 billion for order flow in US stocks and options, up from approximately $3.8 billion in 2021 across the 12 largest US brokerages. The same principle applies to crypto. Derivatives volume as of Q1 2026 reached around $18.6 trillion, 70% of global crypto trading, with spot trading forever dominating. Barter economy rewards speed of trade over disciplined decision making.
At its peak, Robinhood relied on more than 75 percent of its revenue from payment for order flow (PFOF), the hidden backbone of “free” trading where market makers pay brokers to route client orders. Any broker using this incentive model needs clients to trade frequently, even though frequent trading works against long-term returns.
Counseling is no better. Robo-advisors charge 0.25 percent of assets per year, regardless of whether the account is up or down. Human advisors charge about 1 percent, billed against principal even in down years. The extract is designed built into the model: the advisor is paid even when the client loses.
Less exchange friction makes bad trades easier to repeat
The harsh truth is that exchanges need clients to trade more, not win. When retail investors lose, the exchanges still get paid. PiP World research found that 74% to 89% of retail users lose money when trading. Platforms charge at every step, and an AI-enabled exchange could just lead you back to the same losing trade faster.
The April 14 SEC approval of FINRA’s elimination of the Pattern Day Trader rule removed the $25,000 minimum friction. Removing the friction results in more trades, which creates more order flow. More order flow means more money for the broker, regardless of whether the client’s profit and loss (P&L) is up or down.
Enter AI agents, paid to improve clients’ P&Ls
The breaker to this vicious cycle for retailers is the agent built to do what the existing exchange model avoids: trade less, size down, wait and protect customers from their worst impulses. In volatile markets, the best move is often to deny the bad trade and reduce exposure before emotions take over. In the end, keep discipline when the market wants a reaction. Discipline is hard to sell to an exchange because it reduces order flow. An agent who earns by protecting clients’ P&Ls breaks the current incentive model.
The next battleground is who profits from the agents’ order flow
Regulators are squeezing the old “free trade” model. The EU’s PFOF ban comes into force on 30 June 2026 and removes the revenue line behind “free” trades for German and Austrian neo-brokers. Trade Republic, a European savings platform, has already found another way to secure a BaFin license to internalize the order flow.
As TradFi struggles to repair the leak, crypto builders are racing to rebuild onchain rails for AI agents. In markets with tight spreads, fragmented liquidity and millisecond execution, agents trade via nanopayment infrastructure like Circle’s protocol. Gasless trading on perpetual DEX Hyperliquid reduces friction, but maker-taker fees still apply. The real battle ahead is not who removes friction, but who profits when agents start hammering these frictionless rails with high-frequency trading.
Independent programmable agents are better intermediaries
The exchanges and brokers have spent years cashing in on customers trading more, understanding less and absorbing small costs they barely notice. Each agent built by an exchange will inherit the exchange’s incentives. Would an exchange build an agent that sends trades through the rails of a cheaper competitor? Not voluntarily.
Whereas an independent agent has one job: to grow and protect the client’s portfolio, directing deals where they work hardest for the client. Programmable incentives encoded in smart contracts tie the agent’s incentives to portfolio gains. The customer can see where the money is going, control what the agent gets paid, when and why. With independent agents, the client retains more of the value that used to leak to the exchange through order flow, spread of markups and idle cash interest on the exchange.
The agent is rewarded for disciplined trading, not constant trading. It can trade frequently when the signal is strong, reduce exposure when risk increases, and sit out when the market is just noise. The first agent platform to prove this on-chain alignment will give retail investors a fairer counterpart if the economy is finally moving in the same direction as theirs.
