Many legal tech startups streamline manual workflow for lawyers, and some legal tech companies provide tech solutions centered around intellectual property (IP). Among them is a startup called Solve Intelligence, which builds AI software specialized for patent attorneys to write drafts for IP analysis and generation.
Solve Intelligence, a Delaware-based legal tech startup, has raised $3 million in funding from investors, including Y Combinator, Amino Capital, General Advance, SAV, Translink Capital and Nomad Capital.
Chief research officer of Solve Sanj Ahilan and CEO of Solve Chris Parsonson both completed PhDs in artificial intelligence at University College London and experienced the time and cost difficulties of getting a patent while working at tech companies such as Huawei and Dyson. Chris also heard from his patent attorney girlfriend that there’s a lack of software available for patent attorneys beyond Microsoft Word.
Ahilan, Parsonson and Angus Parsonson (chief technology officer of Solve Intelligence) founded Solve Intelligence in June 2023 and launched its product in July.
“Patents are fundamentally different from other areas of legal tech because they intersect both technology and patent law,” Chris Parsonson said, adding that drafting a patent requires combining deep technical addition to legal expertise.
“Unlike other legal professionals, many patent attorneys never go to court or review contracts. Instead, they first gain deep technical expertise in a STEM field through an undergraduate, master’s and often a PhD program,” the CEO continued. “They train for years to pass the bar, after which they can become fully qualified patent attorneys. As such, the algorithms and AI systems needed to write and review patents differ from those needed for equivalent tasks in other areas of legal tech.”
Solve’s product, an in-browser document editor that works like Google Docs but under the hood, is powered by an AI copilot to help attorneys with intellectual property generation. Its AI solution can identify novelty and non-obviousness and help the R&D process. Each approved patent defines what is “novel” and “non-obvious” about a new technology over all other technologies in the public domain, Parsonson said.
“Our AI will be able to automatically pick out the most promising novel and non-obvious steps, rank them by commercial viability, and send these over to the IP firm working with the tech company to begin drafting the patent,” Parsonson said, adding that the AI product helps with patent portfolio infringement litigation after the patent is drafted.
The startup has over 25 IP firms across the U.S., Europe, Asia and South America using its product. Solve claims many of them report 60-90% efficiency improvements. The outlet, which is in the early stage, is already generating recurring revenue between around $100,000 and $1 million by selling its product on a per-seat subscription basis.
The patent analytics market is projected to increase by $5.18 billion in 2032, up from $1.3 billion in 2022.
Some potential user attorneys don’t want to use the AI product due to confidentiality concerns on generative AI tools like ChatGPT. Undisclosed inventions are highly confidential since they haven’t yet been patented.
“Attorneys who turn us down due to confidentiality concerns generally do not yet understand the difference between large language model training and inference. We believe this will be temporary and will become less of a problem over time as people gain [a] greater understanding of AI,” Parsonson said. “Everything is encrypted both in transit and storage and is stored on enterprise AWS servers, so our customers get all of the state-of-the-art security assurances that come with enterprise AWS servers. Hosting AWS is likely much more secure than many alternative custom hosting solutions, which is one of the reasons cloud computing has become so popular over the last decade.”
Solve says it does not train on clients’ data but uses several “proprietary models, algorithms, sequential LLM calls and retrieval augmented generation (RAG) approaches which have been specialized for patent drafting and more generally for IP analysis and generation.” The startup is getting patents for many of these methods and working with some of its IP firm clients to write these patents with the help of its own AI product.
The startup believes every legal professional will use AI daily within the next five years. “The winners will be those who rapidly iterate their product to something genuinely helpful for legal professionals. That’s why we’ve formed a lean, highly technical team of AI PhDs and software dev experts capable of building the features our users want faster than our competitors,” the CEO said.
Some companies develop software for IP, like PatSnap, Cipher and IPRally, which focus on prior art search and technology classification and clustering for R&D teams. Other legal tech companies, including Harvey AI, Robin AI and Casetext, build software for legal professionals to draft legal contracts and find and incorporate relevant case law, but they do not focus on patents. What Solve zeroes in on is generative tasks (on patents) like drafting and responding to office actions, Parsonson pointed out.
“There are many exciting areas our product will be able to help across the intellectual property lifecycle. Patent drafting is just at the beginning,” Parsonson said.
Solve will use the funding to hire staff, meet the demand of new customers and scale R&D for its next generation of patent drafting features, which it will release over the next couple of months, including “algorithms and processes for using AI to generate and analyze technical drawings, review patents to increase their quality, customize the AI to draft in the unique style of the attorney, and much more.”