If you own several companies, AI looks like an obvious portfolio project. One plan, rolled out everywhere, savings in every company. In practice that’s the version that stalls first.
The holding company can’t run an AI project from a distance. It doesn’t know which quotes go quiet in which company, which customer always pays late, or who in the office already does three jobs. The people who run each business know that. So the question for the holding is less “what should we roll out” and more “where do we start, and who owns it there”.
Here is how I’d approach it.
Cross off companies first
The fastest way to find where AI fits is to take out the companies where it doesn’t, at least for now.
- The company you plan to sell within a year or so. The payoff lands with the buyer, and the change lands on a team that is already busy with diligence. - Franchise locations. They usually can’t change their own systems. - Minority stakes. Someone else decides, and you’d be asking them to spend time on your idea. - Very small teams. With two or three people, there often isn’t enough repeat work yet to be worth it.
What’s left is usually a short list.
Rank the rest by three questions
For each company on that list, ask:
1. Is the operator asking for help? An operator who wants it is worth more than any business case. One who is pushed into it will find ten reasons it doesn’t work. 2. Does the same work happen every week, in volume? Inquiries, quotes, invoices, scheduling, documents. That’s where a system earns its keep. 3. Does someone already count a number you’d want to move? Days to invoice, open quotes, hours on a task. If nobody counts anything yet, start by counting, before you build.
The company that scores well on all three goes first. That’s often not the biggest one, or the one with the most obvious problem.
Prove it in one company, then copy what travels
Start in one company with one workflow. Get it running on real work, with one person in that company owning the result. Only then move to the next company.
When you do, be clear about what you’re copying. The way of working travels well: how you choose the first workflow, how drafts get approved, what never gets automated, how you measure. The setup inside each company doesn’t travel as easily. Every company has its own data, its own tone with customers, its own exceptions and its own people. A reminder that works for a landscaping company’s commercial clients will sound wrong to a homeowner who booked a plumber.
So the second company is faster than the first, but it is not a copy and paste.
What the holding should do, and what it shouldn’t
The holding’s job is small and important. Choose where to start. Name the number you want to see move in that company. Ask about it every month. Decide which company is next.
Everything else belongs to the people who run the business. The holding shouldn’t pick the tools, write the rules or decide which customers get which message. When it does, the project becomes the holding’s project, and the team treats it that way.
If you hold for the long term
Most holdcos we talk to, and family offices that own operating companies, have one advantage here: time. You’re not working toward an exit in five years, so you don’t need AI to pay back by next quarter.
That fits the way these systems should be introduced anyway. A step runs on its own only after it has proven itself over weeks, never in a company’s busiest month, and it can always move back. A long-term owner can afford to do it that way. Owners on a short clock are often tempted to skip those steps, and that’s usually where things go wrong in front of customers. If your holdco is fund-backed and plans to sell in a few years, the same steps still apply, with a shorter list and more weight on what pays back quickly.
The other thing that matters more to long-term owners than to most buyers is trust inside the companies. The people in those businesses often stay for decades, just like the owner. How AI is introduced to them, and what it means for their jobs, should be said clearly and early.
Where we come in
With owners of several companies, we usually start with a view across the companies before anything is built: where the repeat work sits in each one, which operator is ready, and which company should go first. Then we build in that company, run it, and take what works to the next.
