Every wealth management firm wants to talk about AI. Fewer are asking the harder question first: is the data ready?

At this year’s Pershing Insite conference, Chip Kispert led a panel on technology integration and AI adoption across the wealth management industry. The insights validated something we see consistently across conversations with RIAs, broker-dealers, and TAMPs: the firms that win with AI won’t win because of their AI tools. They’ll win because they built the infrastructure underneath them.

Here’s what the panel surfaced, and what it means for firms making technology decisions right now.

The integration burden scales differently by firm size.

Panelists distinguished clearly between enterprise-scale firms and mid-sized wealth managers. Large institutions have the budget for custom builds and larger technology teams. They face a different problem: decades of acquisitions leave behind what one panelist described as “clutter of complicated, rusty pipe old infrastructure.” The challenge isn’t resources. It’s orchestration across a legacy landscape.

Mid-sized firms face the opposite problem. Limited budgets force reliance on custodians and third-party platforms. The risk isn’t overspending on custom builds. It’s falling behind a technology curve they can’t staff their way around.

The panel’s advice to mid-sized firms was direct: be stewards of technology, not a technology shop.

That distinction matters. A firm’s job is to serve clients, not to maintain integrations. The moment a technology team becomes a build team, the core business suffers.

Data discipline is not optional before AI.

This was the panel’s sharpest point. AI learns from data. Poor data quality produces poor decisions.

One example that surfaced: the same security called by different names across different systems. To a human, it’s obvious. To a model, it’s a different asset entirely.

The panel described a mid-sized broker-dealer that got this right. They normalized their data. They standardized across systems. They approached build-versus-buy with discipline. The result: a technology foundation that could actually support AI adoption.

The requirements the panel identified:

• Normalize and standardize across all systems

• Manage the full data lifecycle

• Establish a single source of truth

• Get organization-wide buy-in

That last point is underappreciated. Data quality is not an IT problem. It’s everybody’s problem. Firms that treat it as owned by one team tend to accumulate the fragmented, inconsistent data that AI cannot use.

AI ROI is real, but it’s not where most firms are looking.

The panel identified note-taking applications as the highest-ROI AI implementation right now. Advisor productivity is measurable: throughput, new accounts, wallet deepening. These tools are also expanding into CRM territory in ways that change how advisors manage relationships.

The harder ROI conversations involve risk. Most firms don’t yet have adequate controls around AI feature changes from vendors. The panel raised the concept of contractual protections requiring vendor notification before new AI capabilities are deployed. That’s not paranoia. That’s infrastructure thinking applied to vendor management.

Regulators are paying attention. Firms that move without controls in place are creating exposure they may not see until it’s too late to fix cheaply.

Buy versus build: the only question that matters.

The panel surfaced two approaches that reflect the firm-size reality.

Smaller and mid-sized firms: lean toward buy. Build only what is genuinely differentiating. Integration work can be a differentiator, but only when it’s strategic and deliberate, not when it becomes an ongoing maintenance burden.

Enterprise firms: honest self-assessment. Know where internal build capability is strong and where it isn’t. Use systems that already work. Supplement with best-in-class partners everywhere else.

What both approaches share: don’t build what someone else can do better. The question isn’t “can we build this?” It’s “should we?”

The integration layer is where firms win or lose.

The panel’s closing themes centered on what good integration actually targets: eliminating unnecessary advisor effort, reducing risk, and ensuring data consistency. Not every connection warrants deep integration. Some should be light. Others need to be structural.

The firms making these decisions well ask a different set of questions. Where is data being duplicated? Where are advisors doing manual work that a connected system could eliminate? Where does inconsistency between systems create audit risk?

These are not technology questions. They are business questions with technology answers.

The infrastructure is the strategy.

What Pershing Insite’s panel described, from multiple perspectives and firm sizes, is a consistent underlying reality: the firms investing in data infrastructure and integration discipline are the ones positioned to use AI effectively. Everyone else is building on unstable ground.

The infrastructure for wealth management is not a headline product. It doesn’t generate conference buzz. But it’s the difference between advisors spending time on clients and advisors spending time on their systems.

The firms that build the infrastructure first will move faster everywhere else. That’s not a prediction. It’s already happening.