NLPatent Is Now Clerq

A letter from Stephanie Curcio, CEO & Co-Founder

Today, NLPatent becomes Clerq. It’s the most exciting day in our company’s history! It’s so much more than just a rebrand story - it’s an evolution that reflects both a change in the technology and the direction professional services are heading.

In case the length of this letter is daunting and you read nothing else, here’s the short version: NLPatent is the intelligence. It’s the model we spent five years building to understand patent language, and it isn’t going anywhere. Clerq is what that intelligence can now do on your behalf. You give it an assignment, it goes away and does the work end-to-end, and it comes back with a finished, cited report you can check quickly and confidently put your name on. A law clerk works under an attorney’s direction. So does Clerq. That’s the whole idea, and it’s why we chose the name.

Now let me take you back to the beginning of the story, because the reason this works has everything to do with where we started.

Act I: Betting on LLMs Before They Were Cool

My journey started as an intellectual property attorney at a big law firm. And it’s funny - when you start out as a patent professional, they throw you into prior art searching. Why? Because it’s foundational. It forces you to figure out what an invention, described in words you may barely understand, is actually directed to, and then to spot when two documents that read nothing alike are describing exactly the same thing. Which happens constantly, by the way! You do that over and over, across technology fields you’ve never seen before, and you develop confidence in the language we lovingly refer to as “patentese.” 

And it matters far beyond the search itself. Prior art research sits underneath almost everything else a patent professional does. Drafting, prosecution, validity, freedom to operate, portfolio strategy, litigation. All of it starts with knowing what already exists, and everything downstream inherits the quality of that answer. Every high-value task you do as a patent professional depends on it.

But because of how law firm economics work, it ends up being a learning exercise more than anything else. That time rarely gets billed, or ends up written off. So you learn it, then you mostly stop. As your time gets more expensive, the searching moves elsewhere - to colleagues earlier in their careers, or out to specialist search firms who do this exceptionally well. This always struck me as a little backwards. As you develop your skills as an attorney and gain more context around the “why” behind the work, you’re suddenly one or two steps removed from the source material where that context would be most valuable.

When James and I started this company in 2021, that’s the problem we set out to solve. The tools that existed at the time - keyword, classification, even early semantic search - none of them got close. But there was a seismic shift happening in the technology - a little something called large language models. This was before ChatGPT; before anyone outside a research lab was talking about them. 

Patent language has always been its own dialect, written deliberately to be broad and obfuscated at the same time. Try searching “channel” across 180 million patent publications - you’re going to get hundreds of thousands of results spanning dozens of technology areas. Without context, meaning gets lost. LLMs could solve this, but general models aren’t trained to understand patent language. So we built our own and called it NLPatent.

We went deep on this one problem, and have stayed there ever since. I’ll be the first to admit it’s deceptively hard, but we built a reputation around solving it really well. Honestly, a lot of it was timing. We were early, and the technology caught up to exactly what we were trying to build.

We expected junior lawyers to be the users, and at first they were, because that’s how the pyramid works. But that slowly started to change, and today our power users are the senior attorneys, which is the part I didn’t see coming. The person with twenty-five years of context on the client, the portfolio, and the strategy is now the one running the search, because it costs them so little time to do. And they get back everything the act of searching gives you, the feel for the field, and the things nobody would have thought to tell them.

Oh, the irony! The work nobody could justify billing attorney time for is back in the hands of the most expensive people in the building, and it finally makes sense. In those hands it stops being a cost to manage and starts being a differentiator, because a partner who has actually read the landscape gives better advice and wins work that their competitors can’t.

And this isn’t only a senior attorney story. Junior lawyers get to that same understanding faster than I ever did. The search professionals we work with spend their time on the really hard problems instead of grinding through the obvious ones. And in-house teams get to scope a search to the question rather than to the budget, which anyone who has ever had to settle for a quick knockout search when deeper analysis was needed will appreciate.

Act II: Our Next Big Bet

So about a year ago, we made another bet. We thought professional work was going to start getting delivered by agents - software you hand an instruction to, that goes away, does the whole thing end-to-end, and comes back with a finished work product. Our clients describe it as “AI fatigue,” an overflow of tools nobody asked for. They don’t want more buttons and features. They don’t want to learn another interface. They want outcomes - reliable, defensible work product they can put their name on. So we started building it. We rode a technology wave in 2021 with LLMs, and it turns out we have the perfect foundation for the next one - agentic AI.

Not all agents are created equal - an agent is only ever as good as what it knows. A general-purpose model will reason beautifully over the wrong set of documents and hand you back something confident and completely useless. Who cares how good the analysis is if you’re analyzing the wrong things?

The utility of any agent lives in the retrieval - working out what hundreds of millions of patent documents and non-patent literature teach, and surfacing the handful that matter. That took us five years and our own model to get right. We used to joke that search was unsexy, but here we are in 2026 and it turns out “unsexy” was the right bet.

And I’ll tell you something I didn’t fully appreciate until I started showing this work to in-house teams. They are being bombarded with automation for the glamorous end of the practice - that’s where all the hype is. Meanwhile, the thing they keep describing as a burning need sits at the top of the funnel. IP teams are drowning in disclosures - volumes nobody can meaningfully get through - and AI is making the problem worse, increasing both the volume and the length of what gets submitted. Triage is what decides where attorney time goes, and it’s still a manual, stitched-together process. Clients and colleagues keep telling me we’re the first people to suggest automating it. This feels like the search bet all over again - the biggest gains keep hiding in the work everyone overlooks.

I’d love to tell you we saw all of this coming five years ago. We didn’t! Nobody was talking about agents when we were teaching a model to read patents. But sometimes when you go really deep on a hard problem, you discover you’ve been laying the foundation for the next one the whole time. Agentic AI needs exactly what we spent five years building - and we get to hand our agents that advantage from day one.

Act III: Enter Clerq

NLPatent describes a search engine - it doesn’t describe what the platform has become. We could have put a new label on the agentic products and carried on quietly, but I wanted to signal to the market that we are fully behind the change in how professional services will be delivered.

The industry has a name for this shift - “Service-as-a-Software.” I didn’t coin it, but I wish I had! Legal tech has always sold tools with varying degrees of automation, but the professional using those tools still had to work closely with them, taking time and expertise. Learning which buttons to press and knobs to turn was non-negotiable. With AI, the tools got better - more accurate and more efficient - but you still had to work with them to get the right output. Clerq is the next evolution of legal tech. It hands you the finished work product - a cited, client-ready report with its full reasoning laid out - where you provide as much or as little intervention as you want to give it.

The truth is, we’ve outgrown our name - and I say that with a lot of love! NLPatent built our reputation, and it’s still the intelligence underneath every product we ship. But the work has grown beyond what a search engine’s name can hold. 

So today, we’re Clerq. A law clerk works under an attorney’s direction. They take the assignment, they go away, they read everything, and they come back with a memo that shows their work. You don’t manage them through the task. You don’t learn a system to brief them. You tell them what you need, you read what they bring back, and you keep the judgment for yourself. That relationship is the oldest and most trusted division of labour in law, and it’s the exact framework for the “Clerq” we’ve built.

Now, you might be wondering, if I just spent all this time advocating for how much value lives in the act of searching, why would anyone hand it to an agent? Because the value comes from contact with the prior art rather than the physical act of searching itself. A Clerq report shows its work - every reference and every line of reasoning, so the attorney reviewing it still builds their map of the field - the same way you learn a topic from a well-written law clerk’s memo. The research and intelligence platform you know and trust isn’t going anywhere either; when you want your hands on the wheel, it’s right there. The agents are for the work that was never getting attorney time in the first place - the mountain of disclosures, the patentability assessments nobody has time to do properly, or deep research tasks. That work deserves rigour too.

One last thing about the name. Why the Q? If you’re familiar with James Bond films, you’ll remember the Q Division. Q is the quirky gadget maker who sends the polished professional out better equipped than they could ever be on their own. There’s a second subtlety hiding in plain sight. Look closely at our Q mark and you’ll find the search icon sitting inside the letter, an Easter egg nod to our roots. Search sits at the centre of the mark, which is the most honest thing we could have done, because everything Clerq is able to do rests on it.

Today we’re launching two patentability workflows - a triage step, for teams looking for a quick go/no-go decision on incoming disclosures, and a full patentability report that provides a deeply reasoned analysis, feature by feature. Invalidity is coming in the next few weeks, with more research workflows close behind. 

None of this is launching cold. Members of our Client Advisory Board - senior IP leaders at top firms and in-house teams - have been running these agents against real matters for months, and their fingerprints are all over the design. One in-house leader at a global technology company is already using the patentability workflow to pressure-test AI-drafted disclosures before they reach the patent committee - a case study on that is coming. Stay tuned!

I’m deeply honoured that you’ve been with us these five formative years while we obsessed over patent research. This profession gave me my career, and caring for it is part of the job as I see it - it shapes how we build this company, and how we show up for the industry beyond it. Patent work is going to change a heck of a lot in the next five years, and caring means caring about how that change gets built. We’re building it today.

Welcome to Clerq!

‍

Sincerely, 

‍

Stephanie Curcio

CEO & Co-Founder, Clerq 

The AI Clerk for Patent Professionals

‍