AI finally lets a founder run their own money. Skip the wealth manager, point a few models at the portfolio, do what a family office does — alone, from a laptop. That's the pitch, and it's a good story. For most founders with $5M–$100M in assets, it answers the wrong question.
The better question is narrower. Given what AI can do right now, which parts of running your capital can you take on yourself, and which still need someone who knows what they're doing?
This Week in 30 Seconds
- AI's real effect is cheaper help. It's not much use for the investment calls themselves, but it's good at everything around them. So at $5M–$100M its main effect is making the people and systems of a family office cheap enough to make sense below $100M.
- A narrow slice works solo. The one thing you can genuinely do yourself is point it at a few assets you already know cold and let it keep watch. The rest still needs judgment it can't give you.
- From where I sit. Where I'd trust AI on your own money, and where I'd stop.
- On the Radar. 430-plus tokenised US stocks went live on-chain, a Geneva family office built deliberately small, and franchises as a family-office cash engine.
Three jobs of a family office, and where AI helps
A family office does three jobs. One is managing the money: allocation, manager selection, deciding what goes where. Another is the structure around it: tax, entities, finance, the plumbing that decides how much you keep. The third is the auxiliary work: reporting, admin, pulling private and public holdings into a single view so you can see what you own. AI helps with all three, but it's far more useful in some than others.
Start with the structure and the admin: tax, reporting, chasing paperwork, getting all your accounts into one place. This is where AI is genuinely good already. A lot of what used to require a full-time hire can now be done or drafted in minutes. From inside the industry, you can watch these tools take over work that used to need a person.
The investment decisions are the opposite story, and that's where the hype falls apart. AI is great at pulling information together and handing you an answer. It's no good at telling you whether that answer is worth anything. Ask it about a private credit fund, and you'll get a confident write-up in seconds. Working out whether to believe it is the hard part, and for that you still need to know what you're looking at, which more or less means being a professional or paying one.
There is one thing you can genuinely do yourself with it, and it's smaller than the pitch. If you already know a corner of the market cold (a sector you've lived in, a few positions you run directly), you can point AI at those and have it keep watch for you. That's a real use. It's a long way from running your whole portfolio off a chatbot.
What AI makes cheaper
The bigger effect of AI at this level is simpler: it makes everyone you'd otherwise hire cheaper. A chief investment officer 3 days a month instead of a $500k full-timer. A tax specialist, a structurer, someone to run the portfolio, all of them part-time, all of them backed by software that used to need a $100M portfolio and a full team to pay for.
I went through the build-it-or-rent-it decision the other week, in this piece on family offices under $100m. What AI changes is the price. Most family offices already outsource their investing (about 80% use outside managers), and the smaller ones increasingly rent a chief investment officer rather than hire one.
The stripped-down version, one or two people coordinating a set of outside specialists on shared software, now runs around $50k to $500k a year and gets pitched to families worth $5m to $50m. None of that was affordable not long ago. And where AI does turn up inside these setups, it's doing the research and the reporting, not making the calls. In Deloitte's latest survey of large family businesses, the 86% already using it lean on it for efficiency and admin, not investment decisions.
The obvious pushback is that this line keeps moving, and it does. The models are getting better fast, at exactly the things you'd expect to need a professional for: screening managers, running scenarios, drafting a tax structure. So yes, the gap will keep closing. But right now the judgment is still the hard part, and what's already changed is the cost of buying it in. Vanguard reckons a good adviser is worth up to about 3% a year, and most of that is behavioural — stopping you selling at the bottom. That's the thing AI is worst at. If you're doing this alone, you have to be that steady hand yourself, which is hard when it's your own money on the line.
Where that leaves a founder at $5M–$100M
None of this means hiring or firing someone. It means being honest about what the tool is good for and what it isn't. AI doesn't get rid of the people who know what they're doing — the tax, the structuring, the investing, or someone sensible holding it all together. It makes them cheaper to get to. And at $5M–$100M, the cost of those people used to be what stood between a founder and a proper setup. That's the thing that's changed.
From where I sit
Founders ask me a version of this constantly: can I just run my own money with AI now? It's usually someone technical, who built their company on software and looks at a wealth manager's fee thinking they could do the same job themselves.
Here's roughly what I tell them. Where you already know enough to catch a bad answer (a corner of the market you understand, the admin, the reporting), go ahead and lean on it. That's where it earns its keep.
The trouble is everything else: the confident summary of an asset class, or investment opportunity, you don't know well, or the view on a fund manager you've never met. That's where it feels most useful and is most dangerous, because you can't tell a good answer from a confident wrong one in a field you don't understand, and it sounds equally sure of itself either way.
So the honest answer is a boring one. Use it where you'd have spotted the mistake anyway. For everything else, the change isn't that you can finally do it all yourself. It's that the person who'd catch your mistakes now costs a fraction of what they used to.
On the Radar
An operator published the exact AI prompt he uses to shortlist acquisition targets.
Ben Kelly put out the actual prompt he runs to go from a universe of companies to a shortlist of acquisition targets in minutes, instead of weeks of manual searching. This is the founder-direct use of AI in the wild: a model pointed at your own deal flow rather than an adviser's. Worth trying against your own criteria, and worth stress-testing what it hands back before you trust a shortlist you didn't build yourself. Read more →
Ondo listed 430-plus tokenised US stocks and ETFs for on-chain trading.
Ondo Global Markets put more than 430 tokenised US stocks and ETFs (Nvidia, Tesla, Apple, SPY, and QQQ among them) on Uniswap, tradeable on Ethereum and BNB Chain. On-chain versions of public equities have moved from pitch decks into live market infrastructure. It's walled off from US persons behind KYC gating, and carries DeFi custody and counterparty questions a private bank wouldn't, worth understanding before it reaches your jurisdiction, not after. Read more →
A data-first method for evaluating private equity managers, aimed at individual investors.
Cape May Wealth Weekly walked through applying systematic, quantitative methods to private-equity manager selection, the kind of analysis usually locked inside institutions. For a founder being pitched to PE, it's a way to interrogate a manager's claimed edge rather than take the deck on trust. It's also the exact analytical muscle AI now puts within reach of a single person, with the patience to use it. Read more →
Inside a Geneva single-family office run deliberately small.
Mr Family Office profiled a single-family office in Geneva built lean on purpose: a stripped-down structure rather than the full institutional build. For a sub-$100M founder weighing how much apparatus to put in place, it's a concrete comparison point, and evidence that small-by-choice is a working model rather than a compromise you settle for. Read more →
Family offices are buying franchise operations for the cash flow.
Another Mr Family Office piece made the case for family offices owning franchise businesses as durable cash generators, useful in a higher-rate world where clean exits are harder to come by. It's a category of operating asset most founders overlook when they shift from building one company to allocating across many. Worth a look if you'd rather underwrite cash flow you can see than bet on a future exit. Read more →
A tech operator reframed a sabbatical as a capital-allocation decision.
Christopher Nelson wrote about how a deliberate 9-month sabbatical in 2019 set up the exit that came years later. He frames the time off as an investment with a return, not dead time. For a founder moving from operator to steward, it's a useful reframe: unstructured time as something you plan and deploy on purpose. Read more →
New on the Site
Last Thursday's piece took apart the comfortable middle of a portfolio: the balanced allocation that looks safe and quietly isn't. If this memo is about where AI belongs in running capital, Barbell Wealth is about how to shape the capital itself: heavy at the safe and risky ends, thin in the middle that only feels secure.
Read it: Barbell Wealth: Why the Safe Middle Is the Most Dangerous Position
This was Capital Signals — weekly briefings on what's reshaping founder strategy on wealth.
Go deeper: Playbooks · Wealth Architecture · Investment Strategy · Business Building · Life Design
Found this useful? Forward it to a founder who's thinking about this stuff. Got a question or disagree with something? Get in touch.
New here? Subscribe for one email a week.
