Anna Meadows, CTO of CodeROI. Building deterministic infrastructure for regulated software workflows.
You scrutinize every line of the budget. AI tokens, cloud bills, headcount, the tool nobody remembers signing up for. You will spend an hour shaving a few hundred dollars off an invoice. Meanwhile, one of the biggest checks available to you sits untouched in your own code history, because it wears a tax label and you assumed it was finance’s problem.
It is not finance’s problem. It is yours. And the maddening part is how well your work already fits.
Software: The Textbook Case
The R&D tax credit rewards work that chases something new or improved, leans on real engineering, begins with technical uncertainty and resolves it through experimentation. Read that back. That is not a description of a lab. That is a description of a Tuesday.
You do not know if the architecture will hold until you build it. You do not know why latency spiked until you dig in. You ship, it breaks, you form a theory, you test it, you try again. Bugs are uncertainty. Rewrites are uncertainty. Will this scale past ten thousand users is uncertainty you resolve the only way anyone does: by trying things and measuring. If your engineering carried no uncertainty, you were not doing engineering. You were doing data entry.
So qualifying is not the hard part. Software practically qualifies by definition. Which raises the obvious question. If it fits this cleanly, why is the money still sitting there?
Obvious Not The Same As Proven
Here is the catch, and it is the whole game. “This obviously qualifies,” and “I can prove this qualifies to a tax authority,” are two different sentences. The first one you say to yourself. The second one is the one that pays.
A credit is not a trophy for working hard. It is a reward for being able to show you worked hard, in a form that holds up when someone skeptical asks. Tax authorities do not grade on how confident you sound. They ask what the uncertainty was, who worked it and when and how you resolved it. Most teams cannot answer cleanly, because nobody wrote it down while it was happening. So when claim season arrives, they reconstruct. They pull engineers off real work to remember last March. They write a story.
A story is exactly what gets challenged because anyone can make it up. The money does not die because the work failed to qualify. It dies because the hard evidence was never captured and it’s not defensible.
A Credit, Not A Deduction
If you are half-listening because tax credit sounds like a rounding error, here is why it is not. A deduction shaves the income you are taxed on. A credit cuts the tax you owe, dollar for dollar. Different leagues.
And before you write it off because a pre-profit company owes no income tax, there is a move built for exactly that. A qualified small business can take the credit against payroll taxes instead, up to $500,000 dollars a year. That is real cash landing in a company that has engineers and no profit. The rough bar is under $5 million in gross receipts and no receipts going back more than five years, which is most early software companies on the nose. This is the part that turns an abstract tax concept into runway.
The Trap Congress Set, Then Partially Undid
There is a second piece, and it bit software teams hard. Section 174 controls when you deduct R&D spend, software included. For decades you deducted it the year you spent it. The 2017 law broke that, forcing companies starting in 2022 to spread the cost over five years, fifteen for foreign work. Since engineer salaries are most of R&D spend, teams suddenly could not deduct their own payroll and got tax bills bigger than their cash, on work that was not even profitable, which even resulted in bankruptcies.
The 2025 law walked it back. New Section 174A restored immediate expensing for domestic R&D, software included, for tax years after the end of 2024. Deduct it now, the year you spend it. Foreign work still stretches over fifteen years, so where the work happens matters. Smaller companies can even amend 2022 through 2024 to claw back what the old rule cost them, but that window closes soon.
Net of the history: domestic R&D is deductible up front again, and the credit stacks on top. Two benefits, not one.
Why This Is Your Money
Finance owns the filing. You own the truth it depends on. From the outside, finance cannot tell which work involved genuine uncertainty and which was routine plumbing. That call lives with engineering, which means the size of the claim, and whether it survives a challenge, traces straight back to you.
So the real question is not whether your team did qualifying work. You probably did. It is whether your systems can prove it without a forensic dig through memory eight months later.
You do not need to become a tax person, but you do need to stop treating this as someone else’s lane. Recognize qualifying work when it happens. Keep a clean trail as you go instead of rebuilding it under pressure. Get a tax pro in early, because the recent changes come with deadlines. And talk to your CPA about whether these incentives fit your situation, because none of this is tax or legal advice and the specifics move.
The principle does not move. The work is already done. The money is probably already earned. The only question is whether you can prove it. Go talk to a CPA to see if this can help you.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
