When a principal asks, “Why did my net worth drop?”, a deceptively simple question, the organisational fractures of a typical family office are immediately exposed. The accounting team surfaces one answer; the investment team surfaces another. Both are defensible. Neither is complete. The disconnect is structural, not incidental.
Family office technology has historically been purpose-built in silos: performance systems are optimised to capture market attribution, while general ledger platforms are designed to track cash flows and transactions. When those systems don’t speak to each other, the CIO and CFO effectively operate from different versions of reality, leaving the principal caught between competing narratives instead of receiving a single, authoritative view of their wealth.
Drawing on decades of experience helping family offices, Ethan Bonar raises a point that often goes overlooked when discussing software: the fundamental design of computer chips. He explains that while a chip can process data at incredible speeds, it requires a staggering amount of energy compared to the human brain.
This efficiency gap is exactly why, despite investing hundreds of thousands of dollars in the latest tech, even the most sophisticated family offices still find themselves manually reconciling performance data with accounting outflows.
What we do know
Right now, most family office setups are basically a tower of different tools stacked on top of each other. A typical family might lean on Addepar to keep an eye on their portfolio. Then turn to QuickBooks to track their spending and cash flow. Finally, to top off their stack, they will use Canoe or another aggregator to pull in all their private assets. Each platform is primed to perform at its peak and even has AI-powered features built into it for enhancements.
The problem with this setup, however, is that it is fragmented. The systems function in isolation and are disconnected. And that is what forces teams to jump between different platforms just to round up the numbers before they can even start to make sense of them.
Looking ahead at what the promise of AI can actually bring, Ethan suggests the answer isn’t about finding one “perfect” piece of software to replace everything. Instead, families need to ask the question, what can help their existing systems talk to each other.
What we don’t know
What we do not know is exactly how fast AI tech will improve. For instance, when ChatGPT was launched in 2022, there were no discussions about people losing their jobs. Which leads us to the second unknown. It’s still unclear how daily family office operations will shift and just how much human involvement will be needed as the future unfolds.
Irrespective of these unknowns, what we do know is that family office software is headed toward more automation. That means an increased use of AI technology. And if we have learned anything in these past few years, it’s that for AI to function optimally, it needs accurate data and context. And above all, data security should be a top priority for a family office.
In that regard, Ethan suggests family offices start to look at Small Language Models (SLMs). These models offer secure, affordable access to their own data, rather than relying on massive, public systems. As he puts it, “the ultimate value of the system lies in giving families sovereignty over their information. If you own your data, you’re ready for whatever comes next, no matter which way the wind blows.”
What is needed?
While AI can do some incredible things, it still doesn’t have that human knack for adapting or understanding deep context. Ethan points out that even though pieces of the puzzle, like pulling data through APIs, function well today, the big challenge is getting all these separate parts to actually work together.
Although the rapid progression of AI is certain, a complete system overhaul is not the solution. For family offices wanting to bridge the AI gap, the focus should shift toward establishing an “umbrella layer” that enables seamless interaction between their existing platforms. Ultimately, this requires an agnostic approach that selects software built for integration and interoperability instead of isolated functionality.
