AI patent strategy starts with a concrete question: what technical problem does your invention solve, and how does it solve it? In the video above, patent lawyer Jeff Schell highlights improvements in memory use, model training, and anomaly detection as areas founders should examine with patent counsel.
A useful invention discussion connects the result your product delivers with the architecture, processing steps, or other technical features that produce it. That connection helps counsel evaluate potential patent claims and helps your team explain what it actually developed.
Start with the technical problem
“Our platform uses AI” leaves important questions unanswered. Describe the limitation your team encountered: a memory bottleneck, an inefficient training process, or an unreliable detection method. Then explain the change you made and why it addresses that limitation.
Bring the implementation into the discussion. Which components interact? What happens to the data at each step? Where does your approach differ from a conventional system? A diagram, worked example, or description of alternative implementations can make the conversation more productive than a feature list alone.
Three AI improvements worth discussing with patent counsel
- Memory efficiency: identify the data representation, processing sequence, or system arrangement that reduces memory demands. Explain the operating conditions and tradeoffs.
- Training efficiency: describe the specific change to the training process, model structure, or computing workflow. Record the baseline used to evaluate the improvement.
- Anomaly detection: explain how the detection system works, what technical setting it operates in, and how the relevant processing differs from existing approaches.
These are prompts for an invention review. Each proposed invention still needs a claim-specific legal and prior-art analysis. Use actual test results when available, label estimates clearly, and record the conditions behind any performance comparison.
How technical improvements fit into AI patent eligibility
The USPTO’s December 2025 eligibility update explains that examiners should consider an invention as a whole, including its described technological advance, when assessing a practical application under the existing eligibility framework. Its subject-matter eligibility resource page collects the agency’s guidance.
Eligibility, novelty, nonobviousness, and disclosure are separate inquiries. Adding the words “technological improvement” to an application does not establish patentability. The claims and supporting explanation must do the substantive work. Nor does a favorable eligibility position guarantee that a patent will issue or withstand a later challenge.
Prepare a useful AI invention record
Before a patent consultation, assemble a working record that your engineers and counsel can review together:
- The technical problem and the limitations of approaches your team considered.
- The steps, components, and interactions that implement your solution.
- Examples, diagrams, and meaningful variations of the implementation.
- Available measurements, comparisons, and known limitations.
- The people who contributed to the invention and relevant disclosure, launch, or filing dates.
Keep this technical material in an appropriate confidential review process. An initial website inquiry can use a short, nonconfidential description.
The detail matters because a patent application must adequately describe and enable the claimed invention. The USPTO’s guidance on specification requirements explains the written-description and enablement requirements. A claim to an attractive outcome needs supporting information about the invention that achieves it.
Common questions about AI patent strategy
Can an AI or software invention qualify for a patent?
It can, depending on the claimed invention and the applicable requirements. A patent attorney can assess the technical implementation, relevant prior art, and disclosure needed for a filing strategy.
Does a technical improvement guarantee an enforceable patent?
No. Eligibility is part of the analysis. Claim scope, prior art, disclosure, ownership, and the circumstances of a dispute also matter. The video is a starting point for a product-specific review.
Discuss your AI invention with Schell IP
Schell IP helps AI companies, software and SaaS businesses, and startup founders connect patent strategy with product development and commercial priorities. Bring a nonconfidential overview of your technology and the decisions ahead.
This article provides general information and does not replace legal advice about a particular invention, company, or transaction.
