In today’s fast-paced world of innovation, artificial intelligence (AI) is transforming industries—and patent drafting is no exception. In his latest webinar, “How to Draft Patents With AI,” Bastian Best dives into the practical tools and techniques that patent professionals can use to integrate AI into their workflow. From understanding inventions to drafting claims and specifications, this session offers a hands-on guide to leveraging AI responsibly and effectively while addressing critical concerns like data security.

Introduction and Objectives

AI has emerged as a powerful ally for patent professionals. Recognizing this potential, Bastian Best designed a comprehensive webinar to explore the practical applications of AI in patent drafting. The session aimed to bridge the gap between theoretical knowledge and real-world implementation, providing participants with actionable insights and hands-on experience.

  • Equipping participants with basic AI literacy was a cornerstone of the webinar. This objective aimed to demystify AI technologies and their specific applications in patent drafting. By understanding the underlying principles of AI, participants could better leverage these tools in their daily work, enhancing efficiency and accuracy in patent drafting processes.
  • Practical implementation skills formed the second key objective of the session. Bastian focused on demonstrating how to effectively use current AI tools in real-world patent drafting scenarios. This hands-on approach allowed participants to gain confidence in integrating AI into their existing workflows, potentially revolutionizing their drafting processes.
  • The third objective emphasized the critical evaluation of AI tools. Bastian stressed the importance of understanding not only the capabilities but also the limitations and security implications of AI in patent drafting. This balanced perspective aimed to equip participants with the knowledge to make informed decisions about incorporating AI tools into their practice, ensuring responsible and effective use.

Key Learning Goals

  1. Understand the evolving landscape of AI tools for patent drafting.
    The seminar provides a comprehensive overview of current AI tools available for patent drafting, from general-purpose language models to specialized patent-specific software. Participants gain insights into the strengths and limitations of different AI tools, including cloud-based and local solutions. The session explores how these tools can be integrated into various stages of the patent drafting process, from understanding inventions to generating claims and specifications.
  1. Learn how to design and implement an AI-powered toolchain tailored to individual needs.
    Attendees learn practical strategies for creating a customized AI toolchain that aligns with their specific patent drafting requirements and workflow preferences. The seminar demonstrates how to combine different AI tools effectively, such as using document analysis tools for invention disclosure review and language models for claim drafting and specification writing. Participants gain hands-on experience in prompt engineering and iterative refinement techniques to optimize AI-generated content for patent applications.
  1. Address concerns around data security and confidentiality when using AI.
    The session tackles the critical issue of data security when incorporating AI tools into patent drafting processes, particularly for sensitive client information. Participants learn about the differences between cloud-based and local AI solutions in terms of data privacy and control. The seminar provides guidance on evaluating AI tools’ security features and implementing best practices to maintain confidentiality while leveraging AI capabilities in patent drafting.

Overview of the Patent Drafting Workflow

The webinar outlined a typical patent drafting workflow:

  1. Understanding the Invention: Analyzing invention disclosures or scientific papers.
    Bastian emphasized the importance of thoroughly comprehending the invention as the first step in the patent drafting process. This involves carefully studying invention disclosures or scientific papers to grasp the core concepts and technical details. By gaining a deep understanding of the invention, patent drafters can better articulate its novel aspects and potential applications.
  1. Creating Claims: Drafting claims that define the scope of protection.
    Crafting well-structured and comprehensive claims is crucial as they define the legal boundaries of the invention’s protection. Bastian highlighted that claims should be carefully worded to capture the essence of the invention while being broad enough to provide adequate coverage. The process often involves iterative refinement, considering various formulations and potential infringement scenarios.
  1. Writing the Specification: Including sections like background, summary, and detailed description.
    The specification provides the technical context and detailed explanation of the invention, supporting the claims. Bastian explained that this step involves writing sections such as the background, which sets the stage for the invention, the summary, which provides an overview, and the detailed description, which elaborates on the invention’s implementation. Each section plays a vital role in ensuring the patent application is complete and understandable.
  1. Creating Drawings: Generating visual representations of the invention.
    Visual representations are essential for clearly conveying the invention’s structure and functionality. Bastian demonstrated how AI tools can assist in creating flowcharts and structural diagrams that illustrate key aspects of the invention. These drawings complement the written description and help examiners and potential infringers better understand the invention’s technical details.

This approach underscores the importance of focusing on the claims as the foundation of the patent application. By starting with well-crafted claims, drafters can ensure that the specification and drawings align closely with the protected subject matter. Bastian stressed that this method helps maintain consistency throughout the application and strengthens the overall patent protection.

The Role of AI in Patent Drafting

AI Fundamentals

Artificial Intelligence has revolutionized many industries, and patent drafting is no exception. To effectively leverage AI tools in patent drafting, it’s crucial to understand the fundamentals of how these systems operate.

LLMs are statistical word prediction machines trained on vast amounts of text data. These sophisticated models analyze enormous datasets of text from various sources, learning patterns and relationships between words to predict the most probable next word in a sequence. This training allows LLMs to generate human-like text across a wide range of topics and styles.

They generate text by predicting the most likely next word based on context. As the LLM produces each word, it considers the preceding text to determine the most appropriate continuation, creating a coherent flow of ideas. This context-aware generation enables LLMs to maintain consistency and relevance throughout longer pieces of text.

While they excel at generating coherent and linguistically correct text, they do not “understand” content and may hallucinate (generate false but convincing information). Despite their impressive output, LLMs lack true comprehension of the content they produce. This limitation can lead to the generation of plausible-sounding but factually incorrect information, a phenomenon known as hallucination, which requires careful verification and fact-checking by users.

AI Tool Landscape

The AI landscape for patent drafting is diverse and rapidly evolving, offering a range of tools to assist professionals in their work. Understanding the different categories of AI tools can help patent attorneys choose the most appropriate solutions for their specific needs and constraints.

  • Cloud-Based vs. Local Tools:
    Cloud-based tools like ChatGPT and Google Gemini leverage vast computational resources, enabling them to process complex tasks with impressive speed and accuracy. However, these tools require sending data to external servers, which can raise significant concerns about data security and confidentiality, especially when dealing with sensitive patent information.
    Local tools such as GPT for All and Rowan Patents run entirely on the user’s device, providing complete control over data and eliminating the need for internet connectivity. While these tools offer enhanced security, they may have limitations in processing power and capabilities compared to their cloud-based counterparts.
  • General-Purpose vs. Patent-Specific Tools:
    General-purpose AI tools like ChatGPT offer versatility and can be adapted for various tasks, including patent drafting, when provided with appropriate prompts and context. These tools can assist with tasks ranging from prior art research to claim language generation, but may require more guidance to produce patent-specific outputs.
    Patent-specific tools like Rowan Patents are tailored to the unique requirements of patent drafting, offering specialized features such as automated claim generation and drawing creation. These tools often integrate seamlessly with existing patent workflows and may provide more accurate results for patent-related tasks without extensive prompting.

Understanding the Invention with AI

  • Example: Bitcoin White Paper
    The webinar used the Bitcoin white paper as a running example to simulate drafting a patent application for blockchain technology.
  • Document Analysis Tools
    Bastian demonstrated how tools like NotebookLM (a document summarization tool) can assist in understanding complex invention disclosures by:
    • Summarizing key concepts.
    • Answering specific questions about the document.
    • Citing sources within the document for accuracy.
  • Insights
    While NotebookLM excelled at summarization, it was less effective for tasks requiring creative writing or claim drafting. For such tasks, general-purpose LLMs like ChatGPT were preferred.

Drafting Claims with AI

Drafting patent claims is a critical step in the patent application process, as claims define the legal scope of protection for an invention. In this chapter, Bastian Best outlined an AI-assisted claim drafting process that combines human expertise with the efficiency of AI tools. The approach emphasizes starting with a broad claim, refining it iteratively, and leveraging AI to explore alternative formulations and anticipate potential objections.

  1. Start with a Broad Claim Capturing the Core Invention
    The first step in claim drafting is formulating a broad claim that captures the essence of the invention. This ensures that the foundational aspects of the invention are protected while leaving room for refinement in dependent claims. Bastian demonstrated how AI tools can assist in drafting such broad claims by analyzing invention disclosures and generating initial drafts based on key technical features.
    AI tools like ChatGPT can help identify the core inventive concept and suggest concise language for broad claims. However, users must critically evaluate AI-generated claims to ensure they align with legal standards and accurately reflect the invention. Starting with a broad claim also provides a framework for structuring dependent claims that add specificity and detail.
  1. Use AI to Explore Alternative Formulations or Generalizations of Key Terms
    AI can be a valuable brainstorming partner for exploring alternative ways to phrase or generalize key terms in a claim. For example, replacing “cryptographic hash” with “digital fingerprint” can broaden the claim’s scope while maintaining its technical integrity. This flexibility allows inventors to capture variations of their invention that might not have been immediately apparent.
    Bastian demonstrated how AI tools can generate hierarchies of generic and specific terms to identify suitable alternatives. This process helps strike a balance between specificity and breadth, ensuring that claims are neither too narrow nor overly vague. By iterating on these alternatives, drafters can refine claims to maximize their enforceability and adaptability.
  1. Iteratively Refine Claims Based on Feedback from AI
    The iterative refinement process involves using AI not only to draft claims but also to simulate feedback from examiners or competitors. For instance, AI can raise potential clarity objections or suggest areas where claims could be more precise or robust. This proactive approach helps identify weaknesses before submission, reducing the likelihood of rejections or amendments during prosecution.
    Bastian highlighted how AI can simulate examiner objections, such as questioning ambiguous terms or unsubstantiated features in a claim. By addressing these issues early, drafters can strengthen their claims and improve their chances of approval. Additionally, iterative refinement fosters collaboration between human expertise and AI capabilities, resulting in higher-quality patent applications.

By following this structured process, patent professionals can leverage AI to enhance their claim drafting workflow while maintaining control over the final output.

Tools Demonstrated

  • ChatGPT was used to draft initial claims and brainstorm generalizations
    ChatGPT proved to be a versatile tool for generating initial claim drafts, offering coherent and linguistically polished outputs. It effectively captured the core concepts of the invention and provided alternative formulations for key claim elements. However, its outputs required careful review to ensure technical accuracy and alignment with patent drafting standards.
  • GPT for All (a local LLM) was also tested but produced less polished results compared to ChatGPT
    GPT for All, a local AI model, offered a secure environment for drafting claims without sending data to external servers. While it performed adequately in generating basic claim structures, its linguistic quality and depth of understanding were noticeably less refined than ChatGPT’s outputs. This model is suitable for users prioritizing data confidentiality over advanced AI capabilities.

Key Takeaways

  • AI can assist in finding the right level of abstraction for claims by suggesting hierarchies of generic and specific terms
    AI tools excel at brainstorming alternative terminologies, helping drafters identify the appropriate level of abstraction for claim elements. For example, replacing “cryptographic hash” with “digital fingerprint” broadens the scope while maintaining technical integrity. This feature is particularly useful when drafting broad claims that aim to cover variations of the invention.
  • It can also simulate examiner objections, helping refine claims before submission.
    By simulating the perspective of a patent examiner, AI can raise potential clarity or scope objections to drafted claims. This proactive feedback allows drafters to address weaknesses early in the process, reducing the likelihood of rejections during prosecution. Such simulations foster iterative refinement, ultimately leading to stronger and more defensible claims.

Writing the Specification

Drafting the background and summary sections of a patent application is a crucial step in presenting the invention effectively. Bastian Best demonstrated various AI-assisted approaches to streamline this process, ensuring comprehensive and well-structured content. Here’s an overview of the methods he showcased:

Background Section

  • Using Rowan Patents’ built-in local AI model to auto-generate text based on claims
    Rowan Patents offers an integrated local AI model that can automatically generate a background section using the input claims. This approach provides a quick starting point for drafting, though it may require refinement to ensure accuracy and avoid revealing the invention prematurely. The generated content is based solely on the information provided in the claims, which can be both an advantage and a limitation.
  • Leveraging ChatGPT with detailed prompts specifying tone, structure, and content guidelines
    Bastian demonstrated using ChatGPT with carefully crafted prompts to generate a more tailored background section. This method allows for greater control over the content and structure, ensuring the background aligns with specific drafting preferences and legal requirements. By providing detailed instructions, the AI can produce a more refined output that better fits the patent application’s needs.

Summary Section

  • Repeating each claim in narrative form:
    Bastian’s approach involves restating each claim in a narrative format within the summary section. This method ensures that all claimed elements are thoroughly explained in the specification, providing support for the claims and enhancing the overall coherence of the application. It also helps in presenting the invention’s scope in a more readable and comprehensive manner.
  • Brainstorming technical effects using AI to highlight advantages of specific claim elements
    AI tools were utilized to generate ideas about the technical effects and advantages of specific claim elements. This approach helps in articulating the invention’s benefits and can strengthen the case for patentability. By using AI as a brainstorming partner, patent drafters can explore a wider range of potential advantages and select the most compelling ones for inclusion in the summary.

Defining Terms

  • AI tools like Perplexity AI were used to define key terms for inclusion in a glossary within the specification:
    Perplexity AI and similar tools can quickly generate definitions for technical terms used in the patent application. This method ensures consistency and clarity throughout the document, reducing the risk of misinterpretation. By incorporating these AI-generated definitions into a glossary, the patent application becomes more accessible to examiners and potential licensees, enhancing its overall quality and effectiveness.

Creating Drawings with AI

In drafting patent applications, figures play a crucial role in visually representing the invention’s structure and processes. During the webinar, Bastian Best showcased Rowan Patents’ integrated drawing editor as a powerful tool for generating and describing figures, including flowcharts and structural diagrams. These figures were further enhanced with AI-generated descriptions, providing detailed explanations tailored to patent drafting requirements.

  • Flowcharts were auto-generated from method claims
    Rowan Patents simplifies the creation of flowcharts by automatically generating them based on method claims. This automation saves time and ensures consistency between the claims and the figures. Each step in the method claim is represented as a block in the flowchart, complete with reference numerals, providing a clear visual representation of the invention’s process.
    The generated flowchart can be further customized to match specific drafting preferences or jurisdictional requirements. Users can modify labels, add annotations, or adjust formatting to enhance clarity. This ensures that the figure aligns seamlessly with the written description and meets formal patent office standards.
  • Structural diagrams were created manually using predefined terms from claims (e.g., “block,” “digital fingerprint”)
    For structural diagrams, Rowan Patents allows users to manually create figures by dragging and dropping predefined terms from the claims into the drawing canvas. This feature ensures that all key elements of the invention are accurately represented and labeled with consistent terminology. For example, terms like “block” and “digital fingerprint” can be directly linked to their corresponding reference numerals in the claims.
    The manual creation process offers flexibility, enabling users to design diagrams that best illustrate the invention’s architecture. By leveraging predefined terms, users can maintain consistency across the claims, specification, and figures, reducing the risk of discrepancies. This approach also allows for easy updates if claim terminology changes during drafting.
  • These figures were then described using ChatGPT, which produced detailed figure descriptions based on prompts specifying structure and content requirements
    Once the figures were created, Bastian demonstrated how ChatGPT could generate detailed descriptions for inclusion in the patent application. By providing specific prompts that outlined structure and content requirements—such as mentioning each element at least once and explaining their interactions—the AI produced coherent and comprehensive descriptions.

These AI-generated descriptions serve as a strong starting point but require careful review to ensure technical accuracy and alignment with legal standards. Users can refine the text to eliminate ambiguities or add jurisdiction-specific language. This collaborative process between human expertise and AI significantly accelerates drafting while maintaining high-quality output.

Data Security Considerations

A recurring theme throughout the webinar was addressing data security concerns when using AI:

  • Cloud-based tools require careful evaluation of privacy policies to ensure client confidentiality is not compromised.
    When using cloud-based AI tools for patent drafting, it’s crucial to thoroughly review the provider’s privacy policies to understand how client data is handled and protected. These tools often offer powerful capabilities but may pose risks to confidentiality if proper safeguards are not in place. Bastian stressed the importance of carefully assessing the security measures implemented by cloud service providers before entrusting them with sensitive patent-related information.
  • Local tools provide a safer alternative but may require more setup and computational resources.
    Local AI tools offer enhanced security by keeping data entirely on the user’s device, eliminating the need to transmit sensitive information over the internet. However, these tools often demand more technical expertise to set up and may require more powerful hardware to run effectively. Despite these challenges, local tools can provide peace of mind for patent professionals handling highly confidential inventions.

Bastian emphasized that understanding where data is processed and stored is critical when selecting AI tools for professional use.

This point underscores the fundamental importance of data locality in the context of patent drafting and AI tools. Patent attorneys must be acutely aware of the physical and virtual locations where their clients’ data is processed and stored to ensure compliance with confidentiality obligations. Bastian highlighted that this understanding is essential for making informed decisions about which AI tools are appropriate for different types of patent work, balancing the need for powerful AI capabilities with the paramount importance of data security.

Prompt Engineering Techniques

In the webinar, Bastian Best emphasized the importance of crafting effective prompts to maximize the potential of AI tools in patent drafting. He introduced six key elements that can enhance the quality of AI outputs, ensuring they are tailored to specific tasks and contexts. Below is a breakdown of these six elements and their significance:

  1. Role: Define who the AI should emulate (e.g., patent attorney, technical expert)
    Defining a role helps set the perspective and expertise level the AI should adopt when generating content. For example, instructing the AI to act as a “patent attorney” or “technical expert in cryptography” ensures its responses align with the desired professional tone and depth. This approach allows users to simulate input from specific experts, enhancing the relevance and quality of the output.
  1. Task: Specify what you want the AI to do (e.g., draft claims, summarize)
    Clearly defining the task ensures that the AI understands the exact action or outcome expected. Starting prompts with action verbs like “draft,” “summarize,” or “explain” provides clarity and direction for the AI. The more specific the task description, the more likely the AI will produce useful and actionable results.
  1. Context: Provide relevant background information or constraints
    Including context helps the AI understand the bigger picture and tailor its responses accordingly. For instance, providing background about an invention or specifying jurisdictional requirements ensures that outputs are aligned with legal or technical standards. Without sufficient context, the AI may generate outputs that are too generic or misaligned with user expectations.
  1. Format: Prescribe structure or style for responses
    By specifying a desired structure or style, users can guide the AI to produce outputs in a consistent and usable format. For example, prompts can include instructions like “write four paragraphs” or “use numbered lists.” This element is particularly useful for tasks like drafting claims or creating structured sections of a patent application.
  1. Tone: Direct linguistic style (e.g., formal, technical)
    Defining tone ensures that outputs match the intended audience or purpose. For patent drafting, specifying a formal and technical tone can help maintain professionalism and clarity. While tone is often implied by other elements like role and task, explicitly stating it can further refine results.
  1. Examples: Provide benchmarks or model answers for reference
    Including examples helps the AI understand user preferences and mimic desired styles or formats. For instance, providing sample claims or summaries can guide the AI in producing outputs that align with established standards. This approach is particularly useful when working on complex tasks requiring high precision.

By incorporating these six elements into prompts, users can significantly improve their interactions with AI tools, ensuring outputs are accurate, relevant, and tailored to specific needs.

Practical Demonstration Recap

The webinar concluded with a detailed, step-by-step walkthrough of drafting a patent application using various AI tools. This practical demonstration showcased how to integrate AI into every stage of the process, from claims to figures and definitions, highlighting both the strengths and limitations of different tools. Below is a summary of the key steps demonstrated:

  • Claims were drafted iteratively with input from both ChatGPT and GPT for All
    Claims were developed by leveraging ChatGPT for its polished linguistic output and GPT for All for its local processing capabilities. ChatGPT excelled at generating clear, structured claims and provided alternative formulations for claim elements, offering flexibility in drafting. GPT for All, while less refined linguistically, allowed secure brainstorming of claim language without transmitting sensitive data online.
    Both tools were used to refine claims iteratively, with ChatGPT simulating examiner objections to identify potential clarity issues. This process helped ensure that the claims were robust, clear, and aligned with patent drafting standards. The iterative approach also demonstrated how AI can act as a brainstorming partner to explore different levels of abstraction or alternative terminologies.
  • The background section was generated using both Rowan Patents’ local model and ChatGPT
    Rowan Patents’ integrated AI model was used to auto-generate a concise background section based on the claims. While efficient, its output required careful review to avoid inadvertently disclosing inventive features in the background section. This highlighted the importance of human oversight when relying on automated drafting tools.
    ChatGPT was also employed to draft the background section using detailed prompts that specified tone, structure, and content guidelines. This approach allowed for greater customization and alignment with best practices, resulting in a more tailored output. The comparison underscored how combining AI tools can provide flexibility depending on the level of detail and customization needed.
  • Figures were created in Rowan Patents’ drawing editor, with descriptions written by ChatGPT
    Rowan Patents’ drawing editor was used to generate flowcharts automatically from method claims and manually create structural diagrams based on predefined claim terms. This feature streamlined the figure creation process while ensuring consistency between figures and claims through automated reference numbering.
    ChatGPT was then tasked with writing detailed figure descriptions based on prompts specifying content requirements. The AI-generated descriptions provided a strong starting point but required manual refinement to ensure technical accuracy and alignment with patent office standards. This integration of AI for both figure creation and description writing demonstrated how these tools can save time while maintaining quality.
  • Definitions were added using Perplexity AI
    Key terms such as “digital fingerprint” and “transaction data” were defined using Perplexity AI to ensure clarity and consistency throughout the application. The tool provided concise definitions that could be directly incorporated into a glossary within the specification. This step illustrated how AI can assist in standardizing terminology across complex technical documents.
    The use of multiple tools in this final walkthrough highlighted how AI can enhance productivity at every stage of patent drafting when used thoughtfully and critically.

Closing Remarks

Bastian encouraged participants to experiment with different tools and workflows while keeping security considerations top of mind. He emphasized that while AI is not a one-click solution, it can significantly enhance productivity when used thoughtfully.

This webinar provided invaluable insights into leveraging AI for patent drafting while addressing practical challenges like data security and tool selection—making it an essential resource for modern patent professionals seeking to stay ahead in an evolving landscape!

Feedback about the seminar by Jörg Smolinski:

Content Relevance and Clarity: How well did the webinar address your expectations regarding the use of AI in patent drafting? Were the topics covered relevant and presented clearly?

The seminar clearly met my expectations. The presentation was easy to follow and the structure was logical. It was particularly good that the basic principles of AI were discussed first in order to create a common understanding of the subsequent processes.

Practical Application: Did you find the hands-on demonstrations of AI tools (e.g., ChatGPT, Rowan Patents) helpful for understanding how to integrate AI into your patent drafting workflow?

The main focus of the seminar was on the practical application of AI for patent drafting. To this end, the different approaches were demonstrated using different tools. The different quality of the results was clearly categorised for the participants.

Key Takeaways:  What were your most significant takeaways from the webinar? Did it provide actionable insights or techniques that you can apply in your practice?

The biggest takeaway for me was that it really depends on how the prompt is designed. More complex prompts containing the 6 elements can produce surprisingly good results, which can be a good starting point for further processing. The use of different tools to analyse complex texts is another valuable aspect of the seminar.

Data Security Concerns: How effectively did the webinar address your concerns about data security and confidentiality when using AI tools in patent drafting?

The security of personal data when using AI was a key issue for many participants. Although cloud-based models are easy to use, they are not suitable for our work when dealing with confidential data. Bastian presented good local programmes that can be used despite the concerns. This now allows us to use AI internally for patent drafting.

Overall Satisfaction and Suggestions: How satisfied were you with the overall quality of the webinar? Do you have any suggestions for improving future sessions on similar topics?

I enjoyed the seminar because it gave me a deeper understanding of how AI works and how it can be used in the patent field. I think we are at the beginning of a development that will lead to the full integration of AI into our work in a few years’ time. This seminar is a good starting point for all participants. The accompanying material is also very helpful for revisiting the examples presented.