AI may be turning more companies into R&D credit candidates

Historically, many businesses did not think of themselves as software companies.

They purchased an ERP system, subscribed to SaaS products, hired third-party developers when needed, and moved on.

AI-assisted coding is changing that.

Finance, operations, marketing, logistics, professional-services, and other teams can now build internal tools, integrations, automations, and applications that would have required a dedicated software-development team only a few years ago.

That can create a tax question these companies may never have considered before:

Are we now performing qualified research?

Using AI does not automatically create an R&D credit

The fact that an employee uses an AI coding tool does not itself make an activity qualified research.

The underlying development activity still has to satisfy the requirements of the federal research credit.

The more interesting question is what the team is actually trying to solve.

For example, employees may be experimenting with:

• software architecture
• system integrations
• data processing
• performance and scalability
• APIs
• automation workflows
• technically uncertain functionality

When development involves genuine technical uncertainty and an iterative process of evaluating alternatives, there may be qualified research activity.

Internal software has additional rules

Software built primarily for a company's own internal use can face additional requirements under the federal R&D rules.

Certain internal-use software may need to satisfy a higher threshold of innovation in addition to the normal qualified-research requirements, although exceptions and separate rules can apply depending on how the software is used.

That makes the facts important.

An internal tool used only for administrative functions may be treated differently from software that allows customers to interact with the company or software being developed for commercial use.

Why AI changes the population of potential R&D taxpayers

The interesting development is not that AI changes the R&D tax rules.

It doesn't.

AI changes who is capable of doing the development work.

A company that historically paid for third-party software might now have its operations team working with AI tools to develop:

• custom reporting platforms
• workflow automations
• internal applications
• customer-facing portals
• proprietary data tools
• integrations between existing systems

That means companies that never historically had a meaningful software-development function may now have employees performing activities that deserve an R&D review.

Don't start with the tax credit

The right question is not:

“Did we use AI?”

It is:

“What did our people actually build, and what technical problems did they have to solve to build it?”

If meaningful software development is now happening inside the company, the answer may look very different than it did two or three years ago.

That is where the R&D credit analysis should begin.