Artificial intelligence (AI) has quickly become part of the daily conversation across wealth management. Advisers, family office executives and investment professionals are increasingly using AI tools to search, summarise, draft and analyse. Yet beneath the excitement sits a more basic operational question: before AI can create real value, what kind of infrastructure needs to be in place first?
That was the central issue explored at a recent Hubbis discussion hosted with Eton Solutions, where Bryan Henning, President of Eton Solutions, argued that many family offices are still wrestling with an older and more fundamental challenge. Their data remains fragmented across banks, entities, direct investments, documents and reporting formats, often held together by spreadsheets and manual workarounds. In that context, Henning’s core message was straightforward: AI may be powerful, but it cannot compensate for weak operational foundations.
Key Takeaways
- AI Is Not the Starting Point: Henning argued that family offices must first solve data aggregation, reconciliation and workflow issues before AI can be used effectively.
- Operational Complexity Is Rising: As structures become more multi-jurisdictional and intergenerational, manual processes are becoming harder to sustain.
- Privacy Is Non-Negotiable: Family offices remain highly focused on closed-loop environments and data protection when considering AI deployment.
- Integrated Platforms Matter: Investment reporting, accounting, document storage, payments and tax organisation increasingly need to sit in one environment.
- AI Works Best in Defined Use Cases: Document processing, capital call handling, trust calculations and investment summaries are emerging as practical areas of application.
- Smaller Family Offices Need Flexibility: Earlier-stage family offices may benefit from outsourced or hybrid models before building internal capabilities.
The Operational Problem Comes First
Henning was careful not to begin with AI. Instead, he started with the everyday administrative realities facing family offices.
“A lot of family offices struggle with the first initial issue, which is data and multiple banks,” he said. “How do I aggregate those multiple banks with the owner and the family’s private assets, with also private equity and all that?”
That challenge may sound operational rather than strategic, but in practice it has significant consequences. Family offices increasingly need to monitor public and private assets, track multiple entities, support reporting across different stakeholders and produce timely information for principals. When data is spread across banks, managers, spreadsheets and internal teams, consistency quickly becomes difficult.
Henning argued that the first goal should be to get all relevant data into one place so it is “aggregated, accurate, reconciled, and updated as frequently as possible”. In his view, daily visibility matters, because families and their advisers increasingly want a more immediate understanding of positions, exposures and changes across the whole structure.
He also noted that fragmentation can produce conflicting versions of performance. “I have a CIO, and he’s producing one set of results. I have a COO and a CFO. They produce another set of results, maybe three months later,” he said. “And then I have to explain to the owner why the CIO said the return was ten percent, but why it was actually six and a half after we take accruals, mark to markets and other things into account.”
For Henning, that is not simply an annoyance. It is a sign that family offices need a more integrated operational model.
Rising Complexity Is Reshaping the Family Office
A second major theme in Henning’s remarks was the growing complexity of modern family wealth structures. Families are no longer dealing only with listed securities and bank deposits. They are often managing private equity, direct investments, trusts, multiple jurisdictions, different generations and varied beneficiary arrangements.
As Henning put it, “Everybody is going up the complexity curve.”
That complexity is increasing both on the investment side and on the structuring side. Some families are moving from simple investment portfolios into fund structures or co-investment arrangements. Others are dealing with more sophisticated trust and wealth transfer mechanisms as generational succession becomes more pressing.
Henning’s point was that these developments place much greater strain on operational systems. Structures that may look elegant on paper can become difficult to manage if reporting, accounting, workflows and records are not centralised.
He was also candid about the imbalance he sees in how families prioritise spending. “What do you think’s up there with families?” he asked rhetorically when describing where families are most willing to spend money. “It’s all you guys, right? It’s I want to work with a wealth advisor to set up my trust. I want the best tax advice. I want a fund management structure.”
But the lower priority is often the operational side. “Operationalizing that advice,” he said, is what frequently falls into the bottom-left quadrant of perceived value. In other words, families may pay heavily to create complex structures, while underinvesting in the systems needed to run them efficiently.
Why Privacy and Security Shape the AI Conversation
If operational fragmentation is the first issue, privacy is the second. Henning returned several times to the reality that for family offices, confidential data is among their most valuable assets.
“My data is my most valued possession,” he said, describing the mindset of many clients. “I may be worth a hundred billion, but my data is sacrosanct.”
That is why, in his telling, AI adoption cannot be separated from questions of hosting, access control, cyber security and data segregation. He stressed that sensitive information should remain inside a controlled environment and that any AI capability should be applied within that structure rather than through open systems.
He described Eton Solutions’ approach as one in which “the AI comes into your environment” and “you don’t let that data go out in any circumstance”. That point is especially important in the family office world, where concerns around confidentiality often extend beyond investment positions to family governance, trust arrangements, ownership structures and payments.
Henning also argued that AI outputs need to be traceable. “You need to have a citation if you ask a question,” he said. “What’s the citation? Link it to the document you’re referring to.”
That emphasis on verifiability matters because family offices are not just using technology for broad research or drafting. Increasingly, they want systems that can support reporting, document interpretation and workflow decisions in ways that are auditable and defensible.
AI as a Tool for Practical Workflows
Where Henning became most specific was in describing targeted use cases for AI. His argument was not that AI will replace family office teams, but that it can reduce repetitive work and improve process quality when applied to clearly defined tasks.
One such area is document handling. He contrasted older optical character recognition methods with newer AI-driven approaches that can process both structured and unstructured material. “The intelligent document processing now using AI can do structured and unstructured data,” he said. “The AI needs to know what it’s reading. It’s reading a client name, it’s reading an account number, it’s reading an investment, it’s reading an expense.”
That can be useful for everything from invoices and bank statements to capital call notices and expense claims. Henning said that once the system has enough context, it should be able to interpret and route documents accurately, leaving human teams to review rather than manually input large amounts of information.
He also highlighted agent-based workflows. In cases where clients have dozens of external managers or portals, software agents can log in, identify what is new, retrieve data and trigger the next steps. “We’ve replaced that now with agentic AI,” he said, describing a case in which administrative processes that once required a five-person team were significantly streamlined.
The broader point was that AI becomes valuable when it sits on top of clean, centralised data and defined workflows. It is not a substitute for operational design. It is an accelerator.
AI Should Enhance Judgement, Not Replace It
Henning was also careful not to overstate what AI can do. In his framing, AI is highly useful in helping family offices summarise information, organise records, interpret documents and reduce manual effort. But it still needs human review, especially in areas involving judgement, regulation and interpretation.
That is why he described AI as “the icing” rather than the core substance. “First, we’ve got to focus on the cake,” he said.
That metaphor captured his wider point. Before family offices think about the most sophisticated AI use cases, they need to solve the more basic issues around data quality, workflow design, reporting consistency and security. Only then can AI be embedded in a way that produces reliable results.
A Gradual Adoption Path
Henning also acknowledged that not every family office is ready to institutionalise in the same way or at the same pace. Some have the scale to justify a fully built-out internal platform, while others may need a more gradual approach.
In those cases, hybrid or outsourced models can play an important role. Families can use external support and technology together, then bring functions in-house over time as their governance and staffing become more robust.
That flexibility may prove increasingly important. Many families know they need better systems, but are reluctant to overbuild too early or repeat a poor technology decision. Henning suggested that the right approach is often to start with the core pain points, build around them and expand gradually.
For family offices, then, the technology debate is no longer just about whether AI is interesting. It is about whether the office has the digital foundation to manage growing complexity with discipline and confidence.
As Henning put it, “The boring stuff is what actually matters. It’s your data, where is it? How secure is it? How safe is it? And then how am I going to apply AI to it?”
That may be the most useful lesson of all. In family offices, AI is becoming part of the toolkit. But its long-term value will depend far less on novelty than on the strength of the operating model beneath it.