AI adoption should work like compound interest

Over my last few posts, I've been circling something without naming it directly. It’s the underlying force driving the changing fractional playbook, efficiency paradox, and new way to think about change management.

I’m talking about AI.

Whatever your preconceived notions are around AI, park them for a second. Instead, let's start with compound interest—a concept finance people understand better than anyone. Small deposits made early and consistently beat a big lump sum made later (I've been trying to teach my son this concept, but he’s only 19 months and we are still struggling to learn our numbers).

AI adoption works exactly the same way. And yet most firms I talk to are treating it like a lump-sum decision. They're waiting for the perfect entry point: when the tools are mature, the roadmap is obvious, and the risk is zero.

That moment never comes. And while you wait for it, other fractional firms are making deposits.

Here’s how to embrace AI while it’s still a work in progress.

You can’t afford to sit on the sidelines

Let me be direct about the stakes, because they’re much bigger than "AI will help you serve clients better.”

How you handle AI could determine whether your firm still exists in three years.

Do I have your attention now?

The fractional space is bifurcating:

1. AI-first firms are compounding their early advantage: faster deliverables, better margins, a sharper story in every sales conversation.
2. Head-in-the-sand firms are getting left behind, and this reluctance to innovate will eventually impact the bottom line or even the future sale price of your firm.

The talent flywheel cuts both ways

Whether you realize it or not, your best people are watching how you handle this moment.

A-players now screen firms for whether they'll get access to tokens and an environment where they can build AI skills on the job. When a talented senior analyst is choosing between two fractional firms, "will I actually learn how to use AI here?" is increasingly a deciding factor, even if they don’t say it out loud.

Innovative firms attract and retain those people. Stagnant firms train people on old processes, and then they leave. Which means you're paying to educate talent for your competitors.

This effect compounds in both directions. Firms investing in modern skills get the people who want to build them, those people make the firm better, and that attracts more A-players. Firms that are unwilling or unable to embrace new ways of working slowly become the training ground everyone else graduates from.

I don’t have to tell you how quickly this snowball starts rolling down the hill. A firm that started making AI deposits a year ago is much more than a year ahead of the competition.

The honest objection

So if AI adoption compounds this fast, why isn't every firm racing to make that first deposit?

Because it’s a lot easier said than done. I totally understand that. I hear real-world objections all the time from prospects:

• "We tried one of these tools last year and it didn't stick."
• "What about client data security? I'm not putting financials into a chatbot."

Both fair. And I'll be honest about my perspective here: I work for a software company, so a little skepticism toward a guy like me telling you to adopt more software is healthy.

But across the firms I talk to, when an AI initiative fails, the problem is almost always the approach, not the AI. The tool didn't stick because it was treated like a lump sum instead of a deposit—rolled out everywhere at once, with no owner, no specific workflow, and no way to measure whether it worked.

Sound familiar? That's the change problem from my last post in a different context.

As for data security: it's a real consideration, but it's a vetting question, not a reason to sit out. Treat it like any other vendor diligence you'd run for a client.

The approach problem, though…that's the one worth digging into.

So, where do you actually start?

Small, repeatable bets build on one another and beat huge AI moonshots.

Here's what the ideal flywheel looks like:

1. Pick one painful, repetitive workflow

This shouldn’t necessarily be your hardest problem, but your most annoying, recurring one. The thing everyone quietly dreads every month. Maybe it’s formatting variance commentary or chasing the same reconciliation. Could be rebuilding the same reporting pack for the ninth time.

You want something frequent enough to matter, painful enough to motivate the team, and repeatable enough that a win actually compounds.

2. Buy before you build

Don't burn client-billable hours reinventing tooling when you don't yet know your real requirements. Buying surfaces those requirements cheaper and quicker. You'll learn more from four weeks of using someone else's product than from four months of building your own.

So many firms hire engineers and make serious Capex investments before there's a clear gameplan for how to measure ROI. Meanwhile, you still have to keep delivering to clients while you're doing all of this innovating…which is a hard enough balancing act without a homegrown software project on the side.

Building can feel like ownership, but much of the time it’s just a distraction.

3. Pressure-test it against your actual use case

AI has a tendency to look good on the surface, then fall apart in production. You won’t know how your tools hold up in primetime until you stress-test them.

So, run it in a few real-world scenarios. Watch where it shines, and where it breaks on your clients' messy exports, weird chart of accounts, and one-off quirks.

And when it struggles (which it absolutely will), treat it as a learning opportunity rather than a failure. You’re learning what your requirements actually are and which workflows need cleanup first.

4. Quantify the time saved

How many hours per client per month did this workflow used to take? What does it take now?

Write that number down. It’s important for two reasons:

1. First, it's the efficiency you sell, not hide. If you read my efficiency paradox post, this is that argument coming full circle: a quantified AI win is exactly what you put in front of clients as proof of a modern operating model.
2. Second, that number is your deposit slip. This is compound interest you can actually see accruing.

5. Expand to the next workflow

Then run the loop again. Ideally, you learn something new each run, and the cycle spins faster and faster over time.

That's the compounding working in your favor.

Make the good stuff travel

One last thing here, and I'll keep it short because I've already written a whole post about it.

Wins only compound if they spread. If your AI super-user has figured out a brilliant variance-commentary workflow and it lives entirely inside their own client work, your firm isn't compounding—one person is.

So incentivize sharing. Make it normal, expected even, for people to show off what they've figured out. Celebrate it in front of the whole team.

And yes, this is exactly the kind of initiative that needs a change leader who owns it. I won't re-explain that here since the last post covers who to pick and how to back them.

Standing still is falling behind

Here's what I want to leave you with.

While you're waiting for the perfect moment to get serious about AI, other firms are making deposits. And you know better than anyone what happens to the gap between someone who starts saving now and someone who starts in three years.

Even if you’re at square one, the decision to start today is the most important step you can take.

You've now got the case and the first move. In the last post of this series, I'll pull everything together—the playbook shifts, the efficiency paradox, change ownership, and AI—into a checklist you can actually run at your firm. The "so, what do I do about it?" for everything we've covered.

Stay tuned!