Most conversations about AI in wealth management focus on what it will do for advisors, like faster reviews or alerts that flag an issue before the client calls. A simpler question gets much less attention; what is AI using to produce an answer. 

Picture an advisor getting ready for a review meeting. They ask an AI tool to summarize a client's total exposure. The summary comes back in seconds and reads well, but it leaves out the client's private equity commitments and an account held at another institution. The summary looks finished, so the advisor has little reason to suspect anything is missing. 

Why Isn't AI Delivering Yet? 

A 2026 study of financial firms worldwide by the Cambridge Centre for Alternative Finance found that 81% of firms are adopting AI in some form. Yet only 14% see it as transformational to their strategy. More than half (55%) say they find it hard to measure the value AI is delivering at all. 

Firms struggle with AI for many reasons, and one of the most common is that their client data is incomplete or scattered. 

Where Does Client Data Actually Live? 

Ask most advisors where a client's full financial picture lives and the honest answer is usually "in a few places." The core book sits in a portfolio management system, with custodian feeds coming in alongside it. Private fund statements arrive as PDFs, spreadsheets fill whatever gaps are left, and some of the picture lives only in emails or in the advisor's memory. 

An advisor at a dealer tracking held-away assets will recognize this, and so will a family office managing a web of entities and trusts. The details look different from one firm to the next, but most firms are dealing with some version of it. 

We looked at this gap in How Much of Your Client's Wealth Can You Actually See?, and at how to close it in Getting the Full Picture: A Practical Guide for Advisors. AI raises the stakes on both. 

The Cambridge study backs this up. When AI vendors were asked about the biggest data challenges their financial clients face, 72% pointed to data quality and completeness. Another 46% cited legacy systems and siloed data. Notably, data quality was also a top barrier in the same research group's 2020 report. Six years and a wave of new AI tools later, the bottleneck hasn't moved. 

Why Is Bad Data More Dangerous With AI? 

When a spreadsheet is messy, people can usually tell. Totals that don't match and blank cells tend to prompt a second look. 

AI output gives fewer of those clues. A summary built on half the picture reads just as smoothly as one built on all of it, so an incomplete dataset tends to produce an answer that sounds right rather than one that looks broken. 

The answers worth worrying about are the wrong ones nobody thinks to check. 

Even with complete data, AI can still get it wrong. Unreliable output is already high on the industry's list of concerns. In the Cambridge study, "model hallucinations and unreliable outputs" ranked among the top two AI risks for every group surveyed, cited by 70% of financial firms and 70% of regulators. Newer models should reduce some of that risk, though no model can report accurately on assets it has never been shown. 

Who's Responsible When AI Gets It Wrong? 

For advisors, this is also a compliance question. Firms and advisors generally remain responsible for the recommendations and advice provided to clients, regardless of whether AI tools were used in the process. If an AI-generated summary misses part of a client's holdings and a recommendation is built on it, the advisor and the firm are still the ones accountable. 

Regulators appear to see it the same way. The Cambridge study found regulators were more than twice as likely as firms to say responsibility for AI-related harms should stay with the financial institution (38% versus 18%).  

Which Data Gets Missed Most Often? 

Some parts of a client's wealth are far more likely to be missing or out of date than others. Private equity, private credit, real estate, and direct investments often arrive as PDF statements or emails instead of data feeds. Held-away accounts are visible only as far as the client chooses to share them. Holding companies and trusts can hold significant wealth without fitting neatly into a standard account view. 

Quarterly valuations add another problem, since a figure from three months ago can sit on screen next to yesterday's market prices and look just as current. For wealthier clients these are often the assets that matter most, and they're also the ones an AI tool is least likely to see. 

Can AI Help Clean Up the Data? 

Getting to good data doesn't have to mean months of manual cleanup. Some of the most practical uses of AI today happen well before anything reaches a client. AI tools can often help extract holdings and capital calls from fund statements that staff would otherwise type in by hand. They can also spot when the same holding shows up under different names across custodians or flag a valuation that hasn't been updated in two quarters. 

Firms are already moving this way. In the Cambridge study, 69% of firms were at least piloting AI for data and knowledge management, and 79% for process automation. The setups that work best still have a person checking the result, ideally someone who knows the client well enough to notice when something looks off. 

Where Should Firms Start? 

The firms that get the most from AI will likely be the ones that sequence it carefully, using it first to build a complete and trusted picture of each client's wealth before relying on that picture for anything a client will see. Point AI at your data before you point it at your clients. 

Is Good Data Enough? 

Clean data matters a great deal, but firms that fix it will still have other work ahead. In the Cambridge study, firms also named talent (41%) and financial resources (29%) among their biggest barriers. Beyond the data itself, firms need to decide which problems AI is meant to solve and who signs off on its output before a client sees it.  

Even if data is complete and accurate, firms should also consider whether client information is being processed by AI tools in a manner consistent with their privacy obligations, internal policies, client commitments, and vendor agreements. 

They therefore need to know where client data goes when it passes through an AI tool, ask vendors hard questions about how their products handle errors, and make sure their teams are trained to use the tools well. 

Cybersecurity deserves its own attention. Every AI tool connected to client data is one more system that could be targeted, and attackers are starting to use AI themselves. Data privacy and protection was the top AI risk for financial firms in the Cambridge study (74%), and half of firms named AI-driven cyber threats as a major concern. AI vendors were noticeably less worried about those threats (35%), which makes it worth asking any vendor how client data is stored and secured, and what happens if their systems are breached. 

More than half of firms in the study (55%) also worry about "loss of human oversight," where teams gradually lose the ability to do the work themselves once it's automated.  

Where Does Your Firm Stand? 

Before your next AI decision, it's worth asking a few honest questions: 

1. Can I see each client's full financial picture in one place, including private and held-away assets? 

2. How current is that picture, really? 

3. Would I trust an AI-generated summary of it in front of the client? 

4. Will I be able to tell if an AI tool gets something wrong? 

If any of those questions gave you pause, you're in good company – most of the industry would answer them the same way. The foundation can be fixed, and AI itself can help build it. Firms that start there will do better at mitigating AI risk and significantly improve the quality of information available to advisors and the AI tools they add later. 

That foundation is what PortfolioXpressway is built for: bringing a client's full picture, including private and alternative assets, into one place advisors can trust.