Atrium Raises $65M

On the litigation side we regularly see significant typos, misquotes, and even misspellings of the judge's name - and those are just the more basic types of errors we see.

By building the technology they're using Atrium will be able to not just drive efficiency but also improvements in the quality of the legal work product they sell.

It's a very exciting company.

To be clear, was this with Atrium or a different law firm?

1. Is the goal to remove the need for lawyers by letting software do the work, or to equip lawyers (and law firms) to perform their own work more efficiently (and potentially move away from the per-hour pricing model)?

2. I've reviewed probably 25 NDAs this year and they're all differently worded variations of the exact same concepts. I really just want software to tell me "hey, this is your standard NDA but this company took a more aggressive stance on this once piece." Is this part of what you're aiming for?

3. Reading contracts completely sucks. Any plans to make a UI that presents them in a more human-friendly way?

4. The back-and-forth on contracts completely sucks. Any plans to improve this?

IANAL, but it feels like so much of my time as a co-founder goes to managing legal matters, so I guess I'm excited for a better way.

Legal bills, especially IP, are one of if not the biggest expenses for biotech startups in the early days (in many months we spend more on IP than R&D), and IP is mission critical. It definitely feels like Russian roulette -- very hard to predict timing and magnitude of expenses. And it is hard to control expenses as well, bc of limited transparency

Communication is another big issue. For small teams that don't have in-house IP lawyers, or VCs with good relationships with IP lawyers, just translating emails from lawyers into actionable information can take up a lot of time, and we don't want to pay $600 for a phone call every time we want to clear something up

My assumption and understanding is that Atrium's goal is to productize/commoditize full-service law, as opposed to the self-service offerings currently on the market courtesy of e.g. LegalZoom. Basically much more like OneMedical is to medicine rather than WebMD, though that comparison might not be entirely apt. Do you see Atrium targeting specific markets that self-service currently manages somewhat effectively, e.g. sole proprietorships and single-digit headcount startups? Or is it moreso geared to the entities which can foot the more traditional rates (even given that Atrium's pricing is intended to be predictable)?

I noticed that you've scaled the business quite quickly since inception a year ago, growing to around 100 employees.

I see a lot of advice saying to keep growth to only ~2x FTEs per year to preserve culture. How do you reconcile that with your hyper-scale financing and hiring trajectory?

Thanks, really appreciate your thoughts.

One of the primary gaps we need to bridge at Atrium is simply educating different disciplines on what everyone else does. EPD (eng, product, design) and operations teams need to learn what it is like being a lawyer; lawyers and paralegals need to learn what product can help with (and what it can't). We've slowly added more and more cross functional education around these things (onboarding, shadowing, etc) -- and probably have a ways to go.

I'd say it's a middle ground between tech culture and legal. Certain things that are ok culturally in tech (for example, moving fast and worrying about IT security later in the game) are not ok for a venture like Atrium. At the same time, we are trying to innovate in an industry that doesn't get a lot of innovation, so we do want to re-think things from first principles as often as we can (in fact, it is one of our company values).

Happy to explain more if people are interested.

When we were selling Twitch, we didn't want an AI lawyer, Uber for lawyers, a marketplace of lawyers, or workflow software for selling your company. We wanted an M&A attorney with years of experience selling multi-billion dollar deals to tell us that everything we were doing was right. This is the philosophy we're taking to Atrium, which is the best clients will go where the best lawyers go, and the best lawyers will go where they are empowered and life is better :).

More broadly, technology has always displaced humans and whether we should ramp taxes on capital is largely dependent on the distribution of ownership of that technology. We are probably swinging back towards robber-baron levels of value creation accruing to capital, and so I predict taxes to increase in the medium term.

Yes and no.

On the one hand those taxes would hinder technological progress.

On the other hand products made by robots should eventually become free (in the distant future). I.e. if everything is made automatically and nobody has to work, how would people "buy" those products?

What we've learned is that there is a more interesting/difficult technical challenge to what we want to build. I think of Atrium's technology as a platform comprised of several layers: a document storage system (stores legal docs -- think PDFs and Word documents), an ML layer that parses those legal docs and turns them into structured data, an API / web platform layer that exposes that data to applications, and then a bundle of applications that sit on top of that data.

That may be a little complex, so I'll give you one example of this in action, which we currently use in production:

When a startup does a Series A financing, the first step is a VC hands the founder a term sheet. The founder then gives this term sheet to their attorney, who will take it and read all the company's previous SAFEs and convertible notes line by line. They then will create a pro forma cap table (basically, an excel model that shows who gets what at the end of the financing). This process can take a lot of human effort (i.e. the lawyer will spend hours reading the notes and creating the model), and it can often be error prone (because someone is doing it manually).

Of course, any programmer who looks at this process is going to realize it can largely be automated. We ingest those documents into our platform before the founder gets the term sheet. Our models then turn those SAFEs and convertible notes into structured data by mapping them to a schema of what we expect to be in those documents (this is verified after the fact by humans). Then another application just renders the pro forma cap table, taking a process that took hours and reducing it to a matter of minutes. This spits out an XLS, so the attorney can take it and use it to advise the founders (or send to the VC's counsel) just like they would a cap table they built themselves (i.e. we aren't prescriptive on how they advise).

My goal is that eventually we understand every kind of corporate legal document, and make that data accessible to a universe of client / lawyer applications that can streamline workflows and the delivery of service.

It's weird how similar startups can be even when they are in wildly different spaces.

About 2 years ago I realized that as an investor (partner at YC) I had stopped learning. I felt like what I was doing was helpful to some startups, but personally I didn't feel growth any more; being an investor felt to me like being retired. That's not to say that YC isn't hugely impactful (it is), but rather that the cadence didn't really fit my personality any longer: I wanted a shorter feedback cycle around trying something and seeing results. When you are an investor the feedback cycle between trying something (investing in a startup) and seeing results (M&A, IPO, success) can be many years.

In a chance encounter I met the author of Essentialism, which basically is a book about saying no to things. It made me really try to distill down what I liked doing on a day-to-day basis, and I came up with these 3 things: 1) learning about how tech can affect and interact with new industries, 2) trying things and seeing numbers go up as a result, 3) mentoring people over extended periods of time. I decided I wanted to do something that maximized my ability to do those three things on a daily basis.

At first I thought about raising a VC fund, but then realized that wouldn't be much different from YC. Then I thought about incubating 4 or 5 startups at once and finding someone else to be CEO of each. I realized while that would be interesting and minimize personal stress, it wouldn't maximize my growth opportunity to see even later stages of company development. Ultimately I decided the most personal growth maximizing thing to do (that would enable me to focus on the three above things) would be to just start a single company and swing for the fences (try to make it as big as possible).

At the same time, I had been fascinated by the legal innovator's paradox: the fact that there is an intrinsic disincentive to innovate when you don't capture the gains of increased productivity, which is true in an hourly billing model. I also thought lots of the things we had done at YC around building software could be applied in the legal market, and it seemed like a huge one, which incidentally I had experienced my entire career as a founder. After months of research, it seemed like a meaty enough opportunity that really solved a problem of my own, and I decided this would be my one "swing for the fences" company.

Lots of people have asked before so will repeat here: I'm 100% on Atrium now full time as CEO and not doing any other things.