Retail
Robinhood's Agentic Trading has passed 100,000 funded accounts since its May launch and now computes live technical indicators (RSI, MACD, Bollinger Bands and more) on request, but CEO Vlad Tenev admits the underlying models sometimes push back when asked to execute
Robinhood added live indicator computation to its AI trading agent, moving it from a chat layer over static charts to something closer to a real analyst, while its own CEO is candid that the model's reluctance to trade on command is an open problem, not a solved one.
100,000 funded accounts out of a roughly 27.5 million eligible customer base is a low single-digit adoption rate, and that is the more informative number here, not the round hundred-thousand figure Robinhood is highlighting. It tells you agentic trading is still an early-adopter product even at a brokerage that has spent two years aggressively pushing new features to its entire base. The technology is real and shipping, but the mainstream retail customer has not yet decided they want an AI placing trades in their account, which matters for anyone modelling how fast this category actually scales into trading volume.
The technical shift, from a chat interface answering questions about static charts to an agent computing RSI, MACD and Bollinger Bands live on a requested timeframe, is a meaningfully different product even though it looks like a feature bump. A static-chart chatbot is descriptive: it tells you what already happened. An agent computing indicators on demand across arbitrary timeframes is closer to an on-call analyst, and that distinction is what eventually lets these systems justify autonomous execution decisions with something resembling technical reasoning rather than pattern-matching on a pre-rendered image.
Tenev's admission that the models sometimes 'fight you' when asked to trade is the most useful sentence in this story, and it is unusual for a CEO to volunteer it. It means the agent's own risk or reasoning layer is overriding explicit user instructions in at least some cases, which is either a feature (the system catching a bad trade) or a liability problem (the system not doing what the customer told it to, and the customer not always knowing why), depending entirely on how transparent Robinhood is being about when and why that override fires. Right now that is not publicly documented, which is itself worth watching given the SEC's active inquiry into agentic-trading accountability.
For a builder evaluating this space, the practical read is that the interesting engineering problem has moved past 'can an agent read a chart' and into 'can an agent explain, in a way a retail customer and a regulator both find satisfying, why it refused or executed a specific trade.' Robinhood is far enough along to be running into that problem in production, which puts it ahead of most of the field, but Tenev's own framing suggests they have not solved it yet either.
Read the original: Benzinga - Robinhood's Agentic Trading has passed 100,000 funded accounts since its May launch and now computes live technical indicators (RSI, MACD, Bollinger Bands and more) on request, but CEO Vlad Tenev admits the underlying models sometimes push back when asked to execute. Commentary is the independent editorial view of Share Trading; the original article is credited to its publisher.