Devexperts' Jon Light @jonlfa6c9 on how AI is really being used across the brokerage industry right now - from MCP services and DevExa to user profiling, fraud detection, and where trading platforms are heading next. Sam Low of LiquidityFinder sits down with Jon Light, Senior Director of Products at Devexperts, to cut through the AI hype and talk about what's actually in production today, what isn't reliable yet, and why the real value is shifting from the front end to the back end and the API beneath it. We cover LLM stability and why trading can't yet fully rely on it, connecting Claude to live trading data via MCP, whether traders are getting more sophisticated, Devexperts' AI tools (Grenadier and Summerfox), the move from chatbots to agents, and Jon's practical advice to brokers making technology decisions now.
Chapters:
Introduction1:05 The state of AI adoption in brokerage today
2:07 DevExa, natural-language indicators & user profiling
4:45 What MCP is, and connecting Claude to trading data
6:57 Why LLMs aren't reliable enough for live trading yet
9:51 Devexperts market data inside Claude
11:02 Traders building their own front ends with AI
14:41 Are traders getting more sophisticated?
16:09 Expert Advisors vs learning AI models, and crypto bots
17:29 AI tools for brokers: fraud, config & multi-platform control
21:13 Guardrails, and why LLMs won't admit what they can't do
22:27 Grenadier and Summerfox explained
25:18 The future: the front end becomes a commodity
29:23 Agents over chatbots, and models that learn as they go
31:32 Jon's advice to brokers
32:54 Vibe-coding an MCP service in Claude
33:37 Where to find Jon
