resources
How AI is changing retail trading platforms
22 Sept 2026
Ask a trader in 2026 what has changed on their platform screen since 2020, and AI comes up within a sentence or two. Every major broker now advertises some version of it: chatbots that explain unfamiliar terms, algorithms that scan headlines for sentiment, bots that manage a strategy overnight. But AI trading covers a wide spread of tools, and understanding what each one actually does, rather than just what it is called, is the first step to using any of them properly.
Image from ThinkMarkets: ChelseaAI
Three broad categories, one label
If you are wondering how AI trading works, the honest answer is that it depends on the platform and the approach. Most tools fall into three broad groups: analytical tools, automated bots and conversational assistants, and each carries a different level of risk and involvement, so it is worth knowing which is which before assuming they behave the same way.
Analytical and research tools
These sit closest to how traders already work: reading charts, scanning news, building a view. AI versions of this speed up the process by summarising earnings reports, flagging unusual volume, or surfacing correlations a human might miss across thousands of instruments. They inform a decision. They do not make one.
Automated and algorithmic bots
This is the oldest category by some distance. Algorithmic trading predates the current AI boom by decades, but machine learning has made these systems more adaptive. A bot might rebalance a portfolio, run a backtested strategy continuously, or adjust position sizing based on volatility. Once switched on, it trades without a human confirming each order.
Conversational AI assistants
This is the newest category. For a while, it has mostly meant one thing: an assistant you could ask questions to, check a position through, or get a market summary from, all in plain language. What it has not generally meant is action; placing an order or moving a stop has typically stayed back on the platform itself. Until recently that is…
From advice to action: the shift toward execution
Most conversational trading assistants today are exactly that: assistants. They can explain a concept, summarise a position, or answer a question about margin. What they generally cannot do is act. To place an order, a trader still has to leave the conversation and open the platform itself.
That is beginning to change. A newer wave of tools connects AI assistants directly to a live trading account, so an instruction given in conversation can actually be carried out, not just discussed. ThinkMarkets’ ChelseaAI is a useful example of what that looks like in practice.
Launched on 2 June 2026, ChelseaAI connects a ThinkTrader account, live or demo, to any AI assistant that supports the Model Context Protocol (MCP), an open standard for linking AI systems to external services. Claude is the recommended choice.
Once connected, a trader can place market or pending orders, adjust stop-loss and take-profit levels, partially close a position, or check balance, equity and margin, all from within the AI conversation itself.
The distinction that matters here is what ChelseaAI does not do. It does not generate trading signals, offer recommendations, or place a trade on its own initiative. Every action still has to be explicitly instructed by the trader. The AI executes what it is told, nothing more.
ThinkMarkets has positioned it as a workflow tool rather than an advice engine, and the permission structure reflects that: traders can set read-only or full-access modes, revoke access at any time, and every action is logged for an audit trail.
That combination, real execution but only on explicit instruction, with permissions the trader controls throughout, is a meaningfully different proposition to a bot trading autonomously in the background, or a chatbot that can only describe what a trade might look like.
What this shift means for traders
The practical upside is speed and convenience. Checking exposure or adjusting a stop loss without switching apps removes friction that used to cost time, and for traders who already use an AI assistant daily for other tasks, having a trading account inside that same conversation is a natural extension rather than a new habit to build.
It does not remove the underlying risks of trading itself. CFDs and other leveraged products can still produce losses that exceed the amount deposited, and connecting an AI assistant to an account does not change that.
AI does the mechanics; judgement stays with the trader
AI has not replaced the trader’s judgement, and the more credible tools on the market are not trying to make it do so. What has changed is where that judgement gets executed: increasingly, inside a conversation rather than a dashboard.
Whether that is an analytical tool doing the research, a bot running a defined strategy, or an assistant carrying out instructions directly, the pattern across all three categories is the same. AI is handling more of the mechanical work, while the decisions, and the risk that comes with them, stay with the trader.
