Software §101 Eligibility Worksheet
A §101 eligibility scaffold for software inventions, Alice Step 2A/2B and the USPTO's 2024 AI guidance. Free worksheet for patent practitioners.
Written for patent practitioners as a work-aid. Not a substitute for professional judgment.
Attorney work-aid sequence
A structured place to capture the §101 eligibility anchors for a software invention: the abstract-idea characterization to defend against, the specific technical improvement, how it is meaningfully integrated into a practical application, its real-world effect, and any hardware integration, organized around the USPTO's current subject-matter eligibility guidance under Alice.
This is a capture scaffold for an attorney's own analysis. The template itself stays neutral, it does not predict a §101 outcome or recommend claim language. The notes below explain the governing framework and cases for context; none of it is legal advice, and legal judgment stays with the practitioner.
The Template
§101 Eligibility Analysis
Linked Disclosure
Optional, the invention disclosure this worksheet builds on.
Abstract Idea at Risk Step 2A · Prong One
How might the claim be characterized as an abstract idea (mathematical concept, method of organizing human activity, or mental process, the enumerated groupings)? Name the strongest version of the challenge.
Technical Improvement Step 2A · Prong Two
The specific improvement to computer functionality or another technology, what it improves and how, concretely. A specific improvement the specification supports, not a general aspiration.
Evidence of the Improvement Step 2A · Prong Two
Where the specification supports it, and any measured or demonstrable evidence of the improvement.
Meaningful Integration Step 2A · Prong Two
How the claimed elements integrate the concept into a practical application (more than "apply it" on a generic computer).
Real-World Effect Step 2A · Prong Two
The concrete real-world effect or output beyond the abstract idea itself, how the result is actually used (in USPTO Example 47, the eligible claim turned on using the detected anomaly in a specific way).
Technological Components & Interactions Step 2A · Prong Two
The technological elements the improvement runs on and how they interact, not only sensors, displays, controllers, or network elements, but logical structures, memory organization, model training, scheduling, database operations, or resource allocation. Software can improve computer technology without distinctive hardware. Describe the interaction in enabling detail.
Inventive Concept & Evidence Step 2B · Berkheimer
If Prong Two is contested: do the additional elements, beyond the exception, add an inventive concept? Whether an element is well-understood, routine, and conventional is a question of fact, record the evidence (references, admissions, market facts), not just the conclusion.
Result-Oriented vs Mechanism-Specifying
Scan each element: is it claimed by desired result ("configured to detect anomalies") or by a specific mechanism that achieves it? Result-oriented elements are the ones most often held abstract. Flag any, and note the mechanism that could replace them.
Business-Method Framing
Pure technical (no business outcome) / has business outcomes but the contribution is technical / primarily a business-process improvement, and a short explanation.
The framework this scaffolds
Software eligibility runs on the two-step Alice/Mayo framework, as operationalized by the USPTO's 2019 Patent Subject Matter Eligibility Guidance and its 2024 update on AI. The worksheet fields map to the steps:
- Step 2A, Prong One, does the claim recite a judicial exception? The three groupings (mathematical concepts, methods of organizing human activity, mental processes) are the Abstract Idea at Risk field.
- Step 2A, Prong Two, does the claim integrate that exception into a practical application? For software that usually means a specific improvement to computer functionality or another technology: the Technical Improvement, Meaningful Integration, and Real-World Effect fields. Under the 2024 update, Prong Two is the focal point for many AI claims, and the improvement has to be a specific one the specification actually supports.
- Step 2B, if the claim clears the exception hurdle but not Prong Two, do the additional elements add an inventive concept? Berkheimer v. HP (Fed. Cir. 2018) makes "well-understood, routine, and conventional" a question of fact that needs evidence, the Inventive Concept & Evidence field.
The recurring failure mode is result-oriented claiming: courts have repeatedly held claims to data analysis or result-oriented processing ineligible (the Electric Power Group line) while upholding specific improvements to computer functioning (Enfish). The USPTO's 2024 AI examples (Examples 47–49, in the AI eligibility update) make the same point: eligible claims describe a particular method of achieving an outcome; claims that recite the desired outcome do not. The Result-Oriented vs Mechanism-Specifying field is where you flag that.
Machine learning cuts both ways. In Recentive Analytics v. Fox (Fed. Cir. 2025), the first precedential decision on machine-learning (ML) eligibility, applying generic machine learning to a new data domain was held abstract. But the USPTO also designated Ex Parte Desjardins precedential (2025) and added it to MPEP 2106.05(a): a specific improvement to how a model is trained, one that preserves the model's knowledge of earlier tasks while it learns new ones, is an eligible improvement in computer functionality under Step 2A, Prong Two. The dividing line is specificity: a claimed, specification-supported improvement to how the model works is eligible; generic ML on new data is not. This is fast-moving law, confirm current treatment. Our write-up on software patents after Alice works through these cases, and §101 declaration strategy covers Rule 132 evidence when eligibility turns on a factual dispute.
These notes describe the governing framework for context. They are educational, not legal advice, and not a prediction for any specific claim.
Use it in the platform (coming soon)
Our platform will let you fill this in interactively from a linked disclosure and export a finished document. Until it launches, use the template above.
See the Concept Scanner.
Frequently Asked Questions
What is Step 2A, Prong Two of the Alice/Mayo framework?
Under the USPTO's 2019 eligibility guidance, Step 2A first asks whether a claim recites a judicial exception (Prong One), then whether the claim integrates that exception into a practical application (Prong Two). For software, Prong Two usually turns on a specific improvement to computer functionality or another technology. The USPTO's 2024 AI update treats Prong Two as the focal point for many AI claims, where the improvement must be a specific one the specification supports. This worksheet's Technical Improvement, Meaningful Integration, and Real-World Effect fields capture Prong Two facts.
What does Berkheimer add at Step 2B?
Step 2B asks whether the additional elements, beyond the exception, amount to an inventive concept. Berkheimer v. HP (Fed. Cir. 2018) held that whether an element is well-understood, routine, and conventional is a question of fact that must be supported by evidence, not merely asserted. The worksheet's Inventive Concept & Evidence field is where you record that evidence.
Is training a machine-learning model patent-eligible after Ex Parte Desjardins?
It can be. In Ex Parte Desjardins (Patent Trial and Appeal Board, 2025, precedential), the USPTO held that a specific improvement to how a machine-learning model is trained, one that preserved the model's knowledge of earlier tasks while it learned new ones, was a patent-eligible improvement in computer functionality under Step 2A, Prong Two, and added it to MPEP 2106.05(a). The dividing line is specificity: a claimed, specification-supported improvement to how the model works is eligible; applying generic machine learning to a new domain (as in Recentive) is not.
After Recentive, does using AI or machine learning make a claim ineligible?
No, but generic use is vulnerable. In Recentive Analytics v. Fox (Fed. Cir. 2025), the first precedential decision on machine-learning eligibility, applying known machine learning to a new data environment was held abstract. Claims that identify a specific architectural improvement, or a specific technical problem solved in an unconventional way, stand on firmer ground. Treat it as recent, evolving law and confirm current treatment.
What are USPTO Examples 47–49?
They are the three AI-specific hypothetical claim sets in the USPTO's July 2024 subject-matter-eligibility update, anomaly detection with a neural network, AI-based speech separation, and AI-assisted personalized medical treatment, each pairing an eligible claim with an ineligible contrast. Example 47's eligible claim turned on using the detected anomaly in a specific way, which maps to this worksheet's Real-World Effect field. Practitioners cite these examples directly in responses.
Scope. Completing this worksheet does not establish eligibility or predict a §101 outcome. Primarily US law and USPTO practice; other jurisdictions differ.
Disclaimer. This worksheet is a practitioner work-aid for capturing an attorney's own analysis. Obviously Not is not a law firm and does not provide legal advice; this worksheet and any output are for informational and documentation purposes only, are not legal, patent, patentability, eligibility, non-obviousness, claim-scope, validity, or freedom-to-operate advice, and do not create an attorney-client relationship. All legal judgment, including whether and how to file or argue, remains with a licensed patent practitioner exercising independent professional judgment. Nothing here predicts an outcome at the USPTO or any court.