AI adoption splits along the speed of the underlying data

The clearest signal in this week's AI coverage isn't another model release or another chatbot integration. It's the friction visible when LLMs meet industries built on slow, siloed, fragmented data. Wealth management and commercial real estate are both moving—but the constraint is the same: the shape of the information they're trying to modernize.

WealthManagement reports that AI "coworkers" are entering advisory workflows, but success depends on unified data foundations and cost tracking before any LLM flexibility pays off. The bottleneck is not whether the models work—it's whether the firm has cleaned up decades of client data, reconciled custodial feeds, and built the schema layer that makes a language model useful instead of hallucinatory. Commercial Observer covers a parallel story in CRE, where zoning intelligence has been locked in 25,000+ fragmented jurisdictions, buried in legal code, amended constantly, and historically too slow to query at deal speed. AI is now being deployed to parse and surface that layer—but again, the value is in structuring what was unstructured, not in the model itself.

Meanwhile, TechCrunch ran a how-to on disabling Gemini prompts in Google Docs, a small but telling data point on user sentiment when AI is inserted without a workflow fit. And Sequoia's Luciana Lixandru told the Financial Times that Europe is entering "act two" for tech—an acknowledgment that capital, talent, and ambition are all live, but the infrastructure question remains open.

The through-line: AI is repricing not by model capability, but by data readiness. The industries that structured their information early are already compounding returns. The ones still reconciling siloes are paying the integration tax now, in full.