Solving while we work: The AI startup that raised $3m in just a few months
It is a sign of the times that three talented software engineers can launch a generative AI product in the summer and raise $3 million from investors before the leaves turn brown.
It’s also proof that the race to marry the technology with the work of legal professionals began in the months after ChatGPT’s launch in November 2022, and will likely not let up for years to come.
The start-up founders who convinced tech investors they were ahead of the pack are those behind Solve Intelligence, a software tool released this year that uses AI to write patents.
Chris Parsonson, CEO, is well aware that whether the company he co-founded succeeds will be partly down to how quickly it can make the technology slide seamlessly into the lives of highly skilled wordsmiths.
“Generative AI is a greenfield space—there are a lot of improvements to be had with all kinds of products,” says Chris. “So it's all about how quickly we can build it. And how quickly can we make it useful.”
Chris and Solve’s chief research officer, Sanj Ahilan, completed PhDs in AI at University College London, before spells within the R&D departments at Qualcomm, Huawei, Dyson and InstaDeep (acquired this year by BioNTech) introduced them to the ‘pain points’ of patenting.
When OpenAI launched ChatGPT last year, a plan began to form.
“We were blown away,” says Chris, who began looking to see where the large language model could be applied. Thoughts quickly turned to patents. It helped that Chris’ girlfriend is a patent attorney and could share her real-world experience of doing the job.
By June 2023 the pair were joined by Chris’ brother Angus Parsonson, chief technology officer, and the trio founded Solve.
The result is an in-browser document editor that works just like Google Docs but, under the hood, is powered by an AI copilot to help with the patent application lifecycle.
Monzo founder: ‘I’m excited about where they’re going’
That’s the spiel, but does it work? A positive sign is that since the start-up’s launch, more than 25 firms have signed up in the US, Europe, and Asia, ranging from solo practitioners to a few large firms with hundreds of practitioners. And Chris says one happy user told him a patent that would typically have taken five days to draft took just one.
It’s a plan that clearly impressed investors, including Y Combinator (YC), Amino Capital, General Advance, Translink Capital and Nomad Capital, who stumped up multi-million-dollar seed funding.
YC group partner, Tom Blomfield, the founder of Monzo and GoCardless (both now valued in excess of $1 billion), chose to accept Solve into YC and has been a “fantastic advisor”.
Blomfield tells WIPR that the Solve team’s application stood out because the founders have “super technical backgrounds” with AI PhDs from top universities.
“When we interviewed them, it was clear they had spoken with a number of patent attorneys to gain a deep understanding of their workflow and problems.
“I've been impressed with how quickly they ship features to their customers—they launched their product in the first two weeks of the YC batch [and] they started selling to IP firms immediately, getting great customer testimonials.
“They’re addressing the huge IP generation market, and they’re only just getting started. I'm excited about where they’re going!”
Control through customisation
The neat trick is, whatever doesn’t work within the product, probably will soon, thanks to a feedback process that allows users to drive what features the team needs to add. This means new features are added every week, gradually tailoring the software to provide solutions as they go.
It also means the team’s lack of legal training appears to be no hindrance because the product is being built by the users in real time.
Solve is not the only player in the AI patent drafting space. It is rubbing shoulders with PowerPatent, NL Patent, and Dolcera’s IP Author, among others. But Solve has embedded one important element into its product that it believes could make all the difference: customisation.
The team initially built a semantic search engine to find prior art related to a technology, but then realised that a bigger frustration for attorneys was the time and cost of drafting patents, responding to patent office actions, and so on. Nothing unusual there.
What’s interesting is the AI assistant they created next can be used as and when a patent attorney needs it. A user can instruct the AI to review documents and define how it goes about it, not necessarily making changes directly to the document, but drawing the legal expert’s attention to specific parts.
“They can use AI as much or as little as they want,” explains Chris. “They could just use their exact existing workflow or not use AI at all.
“We built an in-browser document editor that works just like Word, so they don't need to change their workflow at all if they don't want to. Enabling the attorneys to customise the AI for a specific use case is super powerful.”
Trusting GenAI with client data
One sticking point faced by all generative AI products is convincing the legal community that confidential information will remain just that.
Chris is adamant that once explained, users understand that while it may have been a worry in the early months of the technology, it need not be anymore.
“No one gets access to the data, no third party; not even us can actually access the data that attorneys are uploading and outputting from our product,” he explains, adding that Solve is undergoing ‘SOC 2’ certification, one of the highest security standards in the world.
The company wants to spend its new cash on adding a senior software engineer, the next step in an ambitious plan to become a global company “used by every IP firm across all regions” within years.
Or perhaps even by next summer.
Did you enjoy reading this story? Sign up to our free daily newsletters and get stories sent like this straight to your inbox
