Key insights
- Artificial intelligence spending alone does not create an R&D tax credit opportunity. Eligibility depends on the underlying development work, including the technical uncertainty encountered, the alternatives evaluated, and the process used to resolve those challenges.
- Potential opportunities may extend beyond employee wages. Depending on the facts, qualifying costs may include certain cloud-computing expenditures and third-party development costs tied to eligible research activities.
- Early assessment and documentation can strengthen a potential claim. Companies should identify qualifying projects, track related costs, and preserve technical records as the work occurs rather than attempting to reconstruct the support during tax return preparation
Evaluate AI development to reveal possible tax savings.
Companies developing and/or investing in AI technology may be able to offset a portion of their development costs through the federal R&D tax credit. Eligibility depends on funding and whether the underlying development activities meet the eligibility requirements for the tax credit.
Learn where to look for potential credit opportunities in your AI projects and what to gather for a supportable claim.
Evaluate the work behind your AI investment
The R&D tax credit extends beyond laboratory research, manufacturing, and new product development. Software development has long been an important category of qualifying activity, and many AI projects involve technical uncertainty, experimentation, and iterative development.
AI spending alone doesn’t create a credit opportunity. The analysis instead focuses on what the company sought to develop or improve, the technical uncertainty it faced, and the process used to evaluate possible solutions.
For example, licensing and deploying an off-the-shelf AI tool generally presents a different tax treatment than building a proprietary model, adding new functionality, integrating systems, or testing competing approaches to solve a technical problem.
How the R&D tax credit works
To qualify for the federal R&D tax credit, an activity generally must:
- Develop or improve a business component
- Rely on principles of science or engineering
- Seek to eliminate technical uncertainty
- Use a process of experimentation under Section 41 of the Internal Revenue Code
For an AI project, the analysis may consider whether the desired functionality was technically feasible, which model architecture or system design could meet performance and scalability requirements, what alternatives the team evaluated, and which development challenges it had to resolve.
Teams that compare approaches, test alternatives, refine models, and address unresolved technical questions may be performing the type of experimentation the credit is intended to encourage.
Common AI development activities that may qualify
Several types of AI development work may warrant a closer look. The strongest opportunities usually involve meaningful technical development rather than routine implementation.
A useful distinction is whether your organization is primarily adopting technology or actively developing and improving it. As a project moves toward technical experimentation, a formal R&D tax credit assessment becomes more relevant.
| Activities That May Warrant An R&D Tax Credit Review | Activities That Generally Don’t Qualify |
|---|---|
| Developing software products with AI functionality | Purchasing and deploying commercial software with little customization |
| Creating proprietary tools or models | Standard operational use of AI tools |
| Testing alternative technical approaches during development | General user prompting activities |
| Integrating AI into existing platforms to resolve technical challenges | Routine implementation or configuration |
| Improving software performance or reliability through experimentation | System maintenance and support after development is complete |
Using AI token costs as potential qualified research expenditures
Many companies view AI spending as another software subscription. When AI coding tools are used directly in software development, however, a portion of that spending may more closely resemble cloud-computing costs incurred for qualified research.
Example:
A developer may use an AI model to generate code, test approaches, troubleshoot errors, or evaluate alternative designs. In these cases, the company is using computing resources hosted by a third party to support experimentation, iteration, and the resolution of technical problems during software development.
Amounts paid for the right to use third-party computers may qualify as research expenses when the computing resources support qualified development activities. AI platforms can operate in a similar manner because each prompt and response uses remote computing capacity that may contribute to developing or improving software.
Businesses using AI coding tools directly in qualifying software development may therefore have an opportunity to include a portion of the related token costs in the R&D tax credit calculation. The treatment potentially depends on how the tools are used, how the costs are structured, and whether the underlying activities otherwise satisfy the credit’s requirements. It should be noted though that the IRS has yet to take a formal position on the eligibility of AI costs for the R&D tax credit.
Internal AI projects require additional analysis
Organizations are also investing heavily in AI-powered systems for finance, operations, human resources, forecasting, reporting, and other internal functions, but these projects require closer review.
Software developed primarily for internal use may be subject to additional requirements beyond the traditional R&D tax credit rules.
See how AI, automation, applications, and connected systems can support your people, simplify efforts, and scale smarter operations. Get the guide.
Depending on the facts, a company may need to show a higher level of innovation, economic risk, and technical advancement beyond what’s currently commercially available.
Internal projects can still support a tax credit, but they have a higher threshold for qualification compared to customer-facing software.
Understanding which AI investment costs may qualify
If a project qualifies, the next step is identifying the costs that may be included in the credit calculation. Depending on the facts, those costs may include:
Employee wages
For many companies, wages represent the largest component of an R&D tax credit claim. This can include employees directly performing qualified development activities, along with certain supervisory and support personnel whose efforts are directly connected to qualifying research.
Software developers, machine learning engineers, data scientists, architects, and technical project leaders may all warrant consideration depending on the facts and circumstances.
Cloud computing costs
AI initiatives frequently require significant computing resources.
Certain amounts paid for the right to use computers in qualified research may be eligible, so determine whether your cloud-computing expenditures are directly connected to qualifying development, testing, and experimentation.
Third-party development costs
Many organizations rely on contractors, consultants, or specialized development firms to supplement internal capabilities. Certain contract research expenditures may qualify when the taxpayer retains substantial rights to the results and bears the economic risk associated with the development effort.
What documentation supports an AI R&D tax credit claim?
Technical work alone rarely supports a credit claim. The company must be able to show how its projects, people, costs, technical uncertainties, and experimentation fit together.
Many organizations can explain the problems their teams solved, but fewer maintain records connecting those challenges to the activities and expenditures supporting the claim.
Effective support typically:
- Identifies the product, process, or software being improved
- Describes the technical uncertainties and alternatives considered
- Preserves testing, development, and project-management records
- Explains employee roles
- Ties wages, contractor costs, and computing expenses to qualifying activities
Because AI projects can move quickly from concept to deployment, companies should capture technical decisions and cost information as the work occurs rather than trying to reconstruct the records later.
How to prepare for R&D tax credit assessments
Companies investing in AI should evaluate potential R&D tax credit opportunities before tax return preparation begins. Earlier involvement makes it easier to identify qualifying projects and build documentation into normal development and financial processes.
Business and tax leaders should understand:
- Which initiatives involve genuine technical uncertainty
- How development decisions are documented
- Whether employee and contractor efforts are tracked
- Which costs relate to qualifying activities
- Whether internal-use software rules apply
They should also consider how federal and state research incentives fit within the company’s broader tax strategy.
Reviewing AI projects early can help align project management, cost tracking, and tax documentation while the relevant information is still accessible.
How CLA can help assess AI R&D tax credit opportunities
AI may be one of today’s fastest-growing areas of business investment, but the tax analysis still rests on established R&D tax credit principles. The amount spent matters less than the work performed: developing or improving technology, confronting technical uncertainty, evaluating alternatives, and documenting the process.
A timely assessment from CLA’s R&D professionals can help you identify potential benefits, clarify documentation needs, and distinguish qualifying experimentation from routine adoption or use.
Contact us
Evaluate AI development to reveal possible tax savings. Complete the form below to connect with CLA.