In a wide-ranging panel at the Hubbis Thailand Wealth Management Forum 2025, exploring how artificial intelligence is reshaping the wealth management industry, Darell Miller, Managing Director for APAC at Wealth Dynamix, emphasised that the most urgent prerequisite for successful AI deployment is not model sophistication or processing power, but data integrity. He cautioned that while generative tools have made AI more accessible, they have also introduced new forms of operational risk, especially when applied to incomplete or inconsistent client information.
“AI has been around for years,” Darell noted. “What’s changed is that it’s suddenly easier to access, experiment with and implement at speed. But that accessibility creates a false sense of readiness. Without high-quality, well-structured and up-to-date client data, the outcomes will be unreliable at best, and damaging at worst.”
He explained that structured product data tends to be AI-friendly because it is standardised, routinely refreshed and centrally managed. By contrast, client data is fragmented, fluid and often inconsistent. “Clients are messy,” he said. “They forget to mention the property in Dubai, they change addresses, and they update their goals irregularly. If your data capture processes don’t reflect that complexity, AI will simply scale your blind spots.”
From clean architecture to strategic advantage
Darell shared that Wealth Dynamix has recently been engaged by a leading Swiss private bank, not for its onboarding or lifecycle management capabilities, but to deliver a full-scale client data mastering solution. The bank, he explained, had recognised that any AI strategy would only be as strong as the foundational data it relied upon. “They’ve made a strategic decision to get their client data model right first. It’s not just a tech investment, it’s a business enabler. With a clean data architecture, they’ll be able to integrate acquisitions more efficiently and scale with confidence.”
When asked to reflect on the risks of poor data quality in AI contexts, Darell described the potential for silent failure. Unlike traditional systems, which flag errors or return blank cells, AI systems will attempt to derive meaning from whatever input they receive, including gaps, duplications or inaccuracies. “The risk isn’t linear,” he said. “It’s exponential. You get an answer that looks plausible, but you have no visibility on how the model got there, and no way to interrogate the logic if it’s wrong.”
For firms looking to adopt AI meaningfully, Darell’s advice was direct: before investing in tools, address the structure, completeness and governance of client data. “Everything else can wait,” he concluded. “If you get the client data right, the rest becomes possible.”