A Rock-Solid Patent for under $1,000? AI Can Get Us There if We Have the Right Focus.
Jedi Knight
Patent & Trademark Attorney
I love working with entrepreneurs. I admire their charisma and abundance of positive energy. That’s why my practice focuses on serving entrepreneurs. That’s also why I hate telling them the reality of getting a patent: It’s going to be expensive. The attorney’s fees alone can often run over ten thousand dollars just to get a patent granted. This pales in comparison with the cost of enforcing a patent, which can run hundreds of thousands and even millions of dollars. The worst part of my job is telling an excited entrepreneur how much it will cost to protect what they are constitutionally entitled to (yeah, a patent is a constitutional right!).
But what if I told you that in the next ten years, a patent could cost only a few hundred bucks? I think we’re a lot closer to that than most patent attorneys are willing to admit. The biggest driver of patent cost is attorney time. Focusing technology development on automating the meat of an attorney’s work, rather than the fluff as most current solutions do, has the potential to dramatically reduce the cost of obtaining a patent.
It’s no secret that hiring a patent attorney is not cheap. But this is rightfully so; patent law is deeply complex, to put it mildly. There’s an adage among patent attorneys: “There is no such thing as a cheap patent.” You can pay the price up front to have a well-drafted, broad patent that accurately covers your invention and potential design-arounds. Or, you can pay the price later when your patent application never gets granted, when your patent is too narrow to enforce, or when your patent gets invalidated when you try to enforce it.
This is why attorneys and savvy business people alike always recommend using a skilled patent attorney to draft a patent application. And because it takes the time and detailed attention of a highly-skilled professional to draft a quality patent application, applying for a patent is expensive.
Despite this, there are a number of aspects of a patent attorney’s job that are simple or repetitive enough to be automated. A number of artificial intelligence (AI) companies have built tools that automate these repetitive or simple tasks, purportedly saving patent attorneys time and, ergo saving inventors money. For example, Specifio®, a startup incubated by the legal innovation accelerator LexisNexis® Legal Tech, touts their software allows a patent attorney to “offload the rote parts of patent preparation.” Indeed, the founders of Specifio® were recently granted U.S. Patent No. 10,417,341 for “Systems and methods for using machine learning and rules-based algorithms to create a patent specification based on human-provided patent claims such that the patent specification is created without human intervention” (say that 10 times fast!). Another AI patent company, Turbopatent®, similarly offers software described on their website as designed to help to “limit the tedious drafting tasks” and help patent practitioners “focus on high-value work” (I think this is a brilliant marketing strategy because every patent attorney I know has bemoaned not getting to focus more time and energy on “lawyer work,” as opposed to other more mundane tasks).
However, in my own *admittedly armchair* assessment, these solutions still fall far short of the real potential AI has to revolutionize the patent industry. A lot of the savings of automation are already reflected in the market price of a patent application because patent attorneys, being inventive themselves, have found their own ways to reduce time on these tasks. Thus, the additional cost of the software is not sufficiently offset by time savings. Indeed, my firm did an assessment of one patent automation company’s products and found that, in comparing the cost of the software to the amount of time saved, there would be ZERO net savings we could pass on to inventors.
Unfortunately, after all the mundane tasks are automated or otherwise streamlined, there still remains the bulk of the arduous and time-consuming tasks associated with preparing a legally and technically sound patent application. This is where we should be focusing technological development.
The reason current efforts to revolutionize the patent drafting and prosecution industry are falling short is because the industry seems to have lost touch with a foundational concept. In a patent attorney’s world, an invention is a set of words in the form of a claim. Those words set the bounds of what the invention is. But this is simply an effect of the massive bureaucracy we deal with on a daily basis (i.e., governments).
In the real, tangible world, an invention is a thing. So if we want to build AI tools that truly revolutionize the patent industry, “[w]e must think things, not words, or at least we must constantly translate our words into the facts for which they stand, if we are to keep to the real and the true.” (Thank you for that famous and prescient wisdom, Justice Holmes). In other words, current AI patent drafting solutions focus too much on the words used to describe an invention. The right solutions will focus on the invention itself.
I envision one such solution looking something like this: An invention disclosure is input into a program. The program performs a search based on the disclosure to identify inventive concepts. Perhaps there is some back-and-forth with the inventor to get missing information and hone in on the core inventive concepts. Simultaneously, the inventor works with an attorney to identify the strategy best suited to the inventor’s goals (although this is something I also believe a computer can do). When the inventive concepts and strategy are identified, the program drafts claims and a specification, and perhaps even drawings, for a patent application. After all this is done, the inventor reviews the application draft for technical accuracy and the attorney reviews for legal (and maybe technical) accuracy. Any changes made to the application by the inventor or attorney are fed back into the program so the program can learn, and the application is filed.
And I believe all that can be done for under $1,000.
But do the tools currently exist to build the right solutions? In my estimation, while there are some technological hurdles that put this problem at or just beyond the cutting edge, the real issue is building the right team. What does that team look like? Well, let’s have lunch and talk about it.
Jedi
