If you're a business leader who's been thinking about starting an AI pilot, be it testing out a custom AI solution, automating a workflow, or building something tailored to your team's way of working, you've probably done the math on what it costs. And maybe that number gave you pause. What most people don't realize, though, is that a meaningful chunk of that cost may come back to you.

If your company is based in the U.S. and you're commissioning custom AI development, there's a good chance your AI pilot qualifies for federal R&D tax credits under Section 41 of the Internal Revenue Code - and depending on your state, additional state-level credits on top of that. No loopholes or "grey" tricks - it's a tax reward mechanism that was designed exactly for this kind of work: projects that involve real technical uncertainty, iterative problem-solving, and building something that doesn't exist off the shelf yet. Custom AI development for your company checks every one of those boxes.

Here's how it works in practice. The federal R&D credit, calculated using the Alternative Simplified Credit (ASC) method, applies a credit rate of roughly 11% to 14% to your qualifying research expenditures. When you hire a third party (like an AI development partner) to build your custom solution, 65% of what you pay them can count toward your qualifying research base. Moreover, the time your own team spends participating in the pilot - joining working sessions, testing outputs, giving feedback - may also qualify as internal research wages, adding another layer to the credit. In a conservative example with a $10,000–$15,000 pilot fee and a team of five people spending a few hours a week over three to four weeks, combined federal and state credits (for a New Jersey-based client, for instance) could come between $2,300 and $6,200. That's potentially a third of your total pilot cost coming back to you - just from doing the work you were already planning to do.

The part that surprises most leaders when we bring this up is how straightforward the documentation side can be, especially when you plan for it upfront. You don't need anything exotic - just organized records of the activities performed, the technical questions you were trying to resolve, and the time spent. Working sessions, testing logs, iteration notes: the kind of things that naturally happen during a well-run AI pilot anyway. The key is capturing them as you go, rather than trying to reconstruct everything at tax time. What's more is that States like New Jersey and Arizona have their own R&D credit programs that layer on top of the federal credit, and if you're operating in or expanding to those markets, it's worth looping in your tax advisor early to understand what applies to your specific situation.

We're not tax advisors, and we'll always say that clearly - but we do think this is one of the most underutilized financial levers available to small and mid-sized businesses exploring AI right now. If you're considering an AI pilot and want to understand what the credit opportunity might look like for your situation, or if you just want to make sure the work is structured and documented in a way that makes it easy for your accountant to work with, we're happy to talk it through. Or you can start with reading more about it with the link below.