“It’s well known in the industry that patent searching is more of an art than a science,” says Massyl Mallem, CEO of PatentPlusAI.
Mallem would know. His background includes hands-on experience as an intellectual property specialist at a multinational corporation. In that capacity, he often acted as a liaison between inventors, patent lawyers, search firms, and other IP professionals in daily quests to find information that would help guide professionals on whether an invention is patentable, valid, or has the freedom to operate. In fact, Mallem performed many of the searches himself, using internal tools, which he found unsatisfactory. “Have I tried enough strategies?” he would ask himself. “Have I looked through enough patents? Have I considered different search terms? You just don’t know when to call it quits.”
This experience was the spark for PatentPlusAI, a company started in 2020 with the aim of applying fully autonomous research to the arduous process of patent search.
Patent insider
Who: Massyl Mallem
Innovation cred: CEO and Co-Founder of PatentPlusAI, chemical engineer, former IP specialist, former research scientist.
Favorite invention: A method for making catalyst film, from the University of Connecticut, Mallem’s alma mater. UConn’s president Radenka Maric is one of the two inventors. Mallem worked in Maric’s lab, which focused on fuel cells and electrolyzers. “This patent feels especially meaningful for me,” he says. “I like it because it shows how advances in catalyst materials become much more powerful when translated into manufacturable layers for systems like fuel cells and PEM electrolyzers.”
Most promising patent: A perovskite-silicon tandem solar cell, granted in 2022 to the King Fahd University of Petroleum and Minerals. This patent “reflects one of the most credible routes to next-generation solar,” Mallem says. “It also resonates with me personally because I had the opportunity to work hands-on with perovskite materials for solar cells as a researcher at Duke University.”
Patents are notoriously extensive documents, jam-packed with all sorts of opaque technological concepts and legal jargon. With some 180 million patent records globally, humans can only do so much. Autonomous patent search is the perfect use case for AI, according to Mallem.
IFI CLAIMS caught up with Mallem last month to discuss the company’s origin story, the appropriate division of patent search labor between humans and AI, why trade secrets are becoming more prevalent, and why PatentPlusAI chose IFI CLAIMS as the data mainstay for its platform. The interview has been edited for clarity and length.
IFI CLAIMS: How did you find yourself in the patent space?
Mallem: My background is in chemistry and chemical engineering. But I’ve always been fascinated by patent law; in fact, I’ve wanted to be a lawyer since I was a kid. So intellectual property is the one industry I’ve been obsessed with my whole life. I did a lot of work at different labs at the University of Connecticut and Duke University. I worked with renewable energy, electrolyzers, and photovoltaics. But my end goal was always intellectual property as it combined my interests in science and technology on the one hand and the field of law on the other. IP is the perfect industry for me because it combines all my passions.
I also have experience working as an intellectual property specialist for a global company that makes chemical products, fabrics for surgical gowns, fabrics for oil filters and the like. As an intellectual property specialist, I was a bridge between the R&D department and inventors, as well as our attorney partners and patent search firms. I had the opportunity to do a lot of patent searching in this capacity and to experiment with various patent search tools and solutions. We worked with a number of search firms. We also had four subscriptions to four different patent search platforms, so I had the great displeasure of trying a lot of solutions and seeing exactly what works and doesn’t work.
I also worked for the IP/TCS department of the University of Connecticut. So I’ve had IP exposure from the corporate side and also from the university side. It was fascinating to see the differences of how big, global corporations think about intellectual property as opposed to how big research institutions think about it.
IFI CLAIMS: Can we assume the displeasure you felt with the other patent search tools paved the way for PatentPlusAI?
Mallem: That’s the foundation story of our company. We designed a tool that we wanted to use. From where I sat as an IP searcher at a global IP department, I wasn’t happy with the available solutions. What would be the perfect tools for us to use? We were privileged to be both end users and clients of our own product, so testing and iterating and making it better was more manageable from both a technical perspective and from a user interface perspective. I knew what IP specialists wanted because I was one of them.
As an IP specialist, we had the opportunity to use search firms and search engines, which is what every IP department in the world currently has available as the only solutions for patent searching problems. Whether it’s invalidity, freedom to operate (FTO) or even a basic patentability, an IP department can outsource to a search firm anywhere in the world. That firm will do the search for them and come back with the results in two weeks. Or the department can use a search engine in-house. And if they use a search engine, they would have to perform the search themselves. There are a lot of different search engines and tools powered by AI, with many different types of features.
There are advantages and disadvantages to each path. When you outsource, you lose some of that internal expertise, and it can become what we call a “telephone game” between the inventors and the searchers. At least that’s what I experienced when I used search firms in my job. The inventor would tell me something. I would tell it to the lawyer. The lawyer tells it to the searcher and then back down the pipeline the information goes. There were a lot of opportunities for misunderstanding or mistranslation. When you’re talking about scientific literature in a legal context, there is already a lot of ambiguity, in both the legal and technical sense, because the text is jargon-heavy. Add to that four or five different players trying to communicate with each other, and it is very, very inefficient. That inefficiency and opportunity for misunderstanding can really undermine the quality. But there are also plenty of pros to using search firms because many times, they put in more effort internally because it’s their entire job. They don’t have competing priorities or tasks. All they do is search.
On the other hand, using a search engine internally keeps the in-house expertise as part of the search equation. Inventors might also do some of the searching. In any event, you have to put a lot of effort into the search because you’re the one responsible for searching, reading patents, and trying different search iterations. That can take many hours. Anywhere from 5 to 30 or much more, depending on the complexity of the search. Ironically, it’s well known in the industry that patent searching is more an art than a science, even though it’s literally marinated in all types of technical domains, because the searcher never knows when to stop. Have I tried enough strategies? Have I looked through enough patents? Have I considered different search terms? You just don’t know when to call it quits. Those are the cons of using a search engine from within the department. Patents are very long and very dense. And yes, there are now AI chatbots that help you interact with patents, but fundamentally, we thought the industry was missing a key component of the solution: autonomous patent research. And we started our company to offer the industry a third option. We tell our clients, “don’t outsource to search firms; don’t waste your time and energy using search engines. Instead, automate your searching with autonomous research systems.” That’s what PatentPlusAI does. It’s not a search engine that you use. It’s not an outsourcing search firm. It’s basically a system of AI workers that perform the search on your behalf.
Yes, we are faster and less expensive, but those are just side effects. The real value is quality: AI at this scale can outperform even the most skilled human searchers. That is why we are moving humans out of the tedious middle work so they can focus on the highest-value human work: strategy, judgment, and decision-making.
IFI CLAIMS: Tell us how your system works.
Mallem: Our system’s workflow can be explained by what we coined the 10-80-10 percentage rule. We also believe this rule perfectly captures the future of AI/human collaboration in all types of white-collar jobs. The first and the last 10 percent are where humans thrive, where humans do their best work. That’s where we target and isolate the human capabilities. The middle 80 percent is done by artificial intelligence systems. The first 10 percent is basically strategizing and ideating, where the abstract thinking happens. Humans are good at deciding what invention they want to search, what commercialization they want to focus on, what competitor they want to target. A lot of times, this first 10 percent involves conversations with the inventor or the lawyer, with the client and their IP departments.
Then AI does 80 percent of the bulk work, the tedious, boring stuff that humans don’t like doing, and more importantly, aren’t good at doing. We’re automating this work because humans are pretty bad at it. Humans are bad at reading 50-page long, dense technical literature filled with legal and scientific jargon. Humans are bad at attempting different search strategies and trying as many as possible. Humans are bad at processing hundreds if not thousands of documents to find a needle in a haystack. Humans will always give up before the AI does. The AI performs the search, finds the best patents, and analyzes them. After that, the human comes back for the final 10 percent of the search project and makes the IP decisions. So if you’re a lawyer, you’ll present the options to your client. If you’re an in-house inventor or work in the IP department, you’ll decide whether or not you want to file, whether you need to go after a certain competitor, or launch the product in a particular jurisdiction. You can decide that you need another search focused on a different set of features or pursue some exploration of white space.
Basically, humans make decisions based on the results the AI system gives them. And that’s what humans should be doing. That’s what our clients are doing right now and that’s what we expect the rest of the industry to do in the next couple of years because this process isolates and takes full advantage of human capabilities, while letting AI do what it does best.
IFI CLAIMS: So for PatentPlusAI, autonomous search isn’t just about being faster or saving money.
Mallem: I tell our clients that we are less expensive and significantly faster, but that those benefits are just side effects of using artificial intelligence at this scale. The actual value proposition is about quality. Whether it’s an invalidity search or an FTO, if the human is doing the bulk of the work, the limitation is actually the human. The tool is only as good as the person using it. Sure, the humans who are our clients are the experts in their fields. And nobody knows their inventions and products better than they do, but they are limited when it comes to how much time they can spend searching, how many search strategies they can try, how many patents they can read and how many figures and tables in each patent they can review. Especially because they have other responsibilities in their job.
Reading a patent and not missing a single quote is inherently a nonhuman task. Reading thousands of patent references without letting anything slip through the cracks is a nonhuman task, particularly when it comes to invalidity searches. A lot of times the answer is that one needle in a haystack. In patent search, volume matters because one reference can destroy a multi-million-dollar patent. The only way to find that reference is to review as much prior art as possible, as thoroughly as possible. AI systems are built for this. They’re built for scale. We focus on the autonomous route because that’s what yields the best results for clients. Meanwhile, it makes our clients’ jobs better and more enjoyable because they’re just doing the human work of decision-making and strategizing.
IFI CLAIMS: What general trends are you noticing in the patent space?
Mallem: Please don’t hate me, but one trend that is so hard to ignore is the emergence of trade secrets. I think trade secrets are becoming a lot more common. As a software company, we’re never going to pursue a patent because our end product cannot be reverse-engineered. We originally wanted to get a patent because we spent more than six years innovating, researching, and building before we launched our platform. Our systems are very complex, and I would love to share this information publicly and get protection for it, but our counsel told us we needed to go the trade secret route because we wouldn’t be able to tell if a competitor is copying us or not. I think trade secrets are going to be more prevalent, but I don’t know to what extent. AI will certainly accelerate it, though. Patents are hard to enforce. So instead of telling the government how things are done and relying on them to protect it, I think companies will try to guardrail their secrets and avoid sharing things.
In patent search, volume matters because one reference can destroy a multi-million-dollar patent. The only way to find that reference is to review as much prior art as possible, as thoroughly as possible. AI systems are built for this. They’re built for scale.
IFI CLAIMS: Why do you use IFI data in your platform?
Mallem: We’ve been doing patent work for a very long time, and when we started we were going to individual patent offices and grabbing the data from each one of them. We also had to maintain that data. But we quickly saw that it was unsustainable. That’s why we partner with IFI CLAIMS. It is because IFI takes all of those headaches away from us. IFI does all of the heavy lifting. And that allows us to focus on our own innovations. IFI does the tedious, difficult stuff that no human wants to do—gathers all the data, prepares it, cleans it, organizes it, and then most importantly, maintains it. An important aspect of having reliable patent data is that it ought to be kept up-to-date frequently. I love when clients ask who our data provider is because it’s a source of pride for our company that we partner with IFI because we think the company is the best in the market in terms of breadth of data and quality. Also in terms of maintenance and reliability.
