The billable hour has shaped how lawyers work for over a century, but it has also made legal work slow, expensive, and inaccessible to the very people paying for it: founders, in-house counsel, and businesses of every size navigating legal transactions.
Spellbook has spent years building an AI-native platform that drafts, reviews, and redlines contracts directly inside a lawyer’s existing workflow, giving legal teams a way to handle far more of their own work without adding headcount. That approach has made Spellbook one of the fastest-growing legal AI platforms for in-house legal teams, who now log more AI usage hours than any other function in corporate America.
Spellbook, launched in 2022, tripled its revenue in 2025 and is on track to reach $100 million in annual recurring revenue (ARR) by year-end, with customers across 80 countries and reportedly more customers than Harvey and Legora combined, including in-house teams at LG, Dropbox, eBay, and Franklin Templeton.
We sat down with Spellbook’s co-founder and CEO Scott Stevenson to discuss why he built Spellbook for in-house teams first, how AI is transforming contract work, and what’s next for the company.
1. What’s the origin story behind Spellbook? How did your first company’s experience with legal fees lead you here?
Before Spellbook, I ran a startup called Mune that built a new kind of electronic music instrument. We raised a small angel round, and by the time we’d incorporated, issued shares, and filed a design patent, half of it was gone to legal fees. We hadn’t launched a single product yet.
What actually bothered me was that the cost had nothing to do with the value of the work. Whether an associate spent thirty minutes or three hours on a document, I paid for the time, not the outcome. I met my co-founder, lawyer Daniel Di Maria, who was equally frustrated at the experience of practicing law: he was spending 10 hours a day in Microsoft Word. Legal agreements power the world’s economy, but they are error-prone, expensive, and slow.
We started Spellbook to create a contract infrastructure that removes this bottleneck and lets billions of agreements move safely and simply.
2. What problem is Spellbook solving that Harvey, Legora, and other legal AI companies aren’t?
Harvey and Legora were built for the world’s 20 million lawyers, mostly inside Big Law firms. We built for a much bigger group: the huge number of people across sales, procurement, HR, and finance who touch a contract at some point, with in-house legal teams as the way in. That’s a different bet on where contracting actually needs to get faster.
We didn’t want to build another chatbot. ChatGPT and Claude already exist, and if all you’ve built is a chat window, you don’t have a product, you have a wrapper. Spellbook works directly inside Word, where lawyers spend their entire day, the same way Cursor lives inside a developer’s existing editor instead of asking them to learn a new one. It is now connecting across business teams in Slack, Salesforce, email, and elsewhere–with deep features like “Compare to Market” that allow a user to compare an agreement to the “typical” agreement, with data to explore.
3. Why does trust matter so much when you’re asking lawyers, one of the most risk-averse professions there is, to adopt AI?
We have one advantage here that people miss: nobody wants a contract to be “creative.” Everyone wants contracts to be boring, standardized, and predictable, and that is exactly what these models are good at.
But trust still has to be earned. So instead of telling a lawyer “this clause isn’t market” and asking them to believe us, we show them the comparison directly, with a citation back to the source. Lawyers have spent their whole careers being trained not to trust anything they can’t verify themselves. Our job is to make the AI’s reasoning verifiable, not to ask for faith.
4. What kind of organizations use Spellbook today?
The majority of our customer base is corporate in-house legal teams and enterprise teams pushing serious contract volume through the platform. That’s our center of gravity. We do work with law firms too, but we’re deliberately focused on the in-house teams who deal with agreements day in and day out.
What we learned is that law firm purchasing decisions often get made by committees more worried about the press release than actual usage. Cutting billable hours in half isn’t exciting news for a firm that bills by the hour. So we sell bottom-up instead, usually starting with one lawyer who just wants their own job to be easier. One of our earliest users at a Fortune 10 company put Spellbook on his personal credit card, months before his own company ever signed a contract with us.
5. What makes legal AI an attractive category for investors right now, and what did Khosla Ventures see in Spellbook specifically?
Legal was one of the last major industries software never really fundamentally changed, yet $100T flows through agreements worldwide. Until 2022, there wasn’t much software could do with unstructured text, so lawyers were still working the way they had for decades: copying and pasting inside Word all day. Compare that to finance, which has had roughly 40 years of software accelerating it since spreadsheets showed up.
Once generative AI arrived, that changed overnight. Before it did, we heard ‘no’ from around 80 investors, because the assumption was that lawyers simply didn’t want to buy software. What investors got wrong was the reason for it. The moment the tools became genuinely useful, lawyers adopted fast.
Khosla backed us because we’d already proven we could grow through that adoption curve ourselves. Keith Rabois joined our board because he saw the same pattern he’d seen before: a boring, overlooked domain about to get faster, cheaper, and more transparent because of software.
6. What is Spellbook’s core product and business model today?
Spellbook started as a Word add-in that reviews contracts, flags issues, checks them against a company’s own standards, and redlines them, functioning a lot like AI tools do for a developer, but for lawyers. That’s still core to what we do, and we’ve built an agent called Spellbook Associate on top of it for more complex, multi-document drafting, the kind of work a financing round requires, where one term sheet spins off ten separate documents.
But we’ve since built well beyond that. We just launched Autonomous Contract Management (ACM), our answer to the fact that contracts don’t end at drafting. ACM pulls contracts in automatically from email, Slack, and Salesforce, triages and redlines them against a company’s standards before a lawyer even opens them, keeps every active negotiation organized in one place, and after signature, stores everything so a team can search their contract history and get answers with citations, while flagging renewals and new risks automatically.
It’s rolling out to select teams right now, and it’s the direction the whole product is heading: from a tool that helps you draft one contract to a system that runs the entire lifecycle. We sell primarily seat-based, land-and-expand: start with a few users, and let the rest of the team pull it in as they see it work.
7. How did Spellbook grow to 4,500+ customers across 80 countries, and what’s driving the path to $100M ARR?
That growth started bottom-up, one lawyer, one team at a time, and increasingly now includes larger enterprise deals as bigger legal departments bring Spellbook in company-wide. Our net revenue retention has stayed above 130 percent, meaning the accounts we land keep expanding on their own.
Every time a big general AI headline hits, our funnel spikes instead of shrinking. Those tools work like a gateway drug. Someone tries Claude for a contract question, likes it, and then goes looking for something actually built for their workflow.
After “Claude for Legal” launched, we nearly doubled the number of lawyer signups we get per week. In-house legal departments have been our fastest-adopting customers, more so than even larger law firms, because they’re not constrained by the billable hour the way firms are, and tripling revenue this year is what puts $100 million in ARR within reach by year-end. We’ve processed more than 10 million contracts to date, and the agreement volume that runs through us keeps climbing. Contract reviews were up 30 percent quarter-over-quarter last quarter alone.
8. What’s next for Spellbook?
We just took the first real step with Autonomous Contract Management, rolling out now to select teams. The idea is to become the actual system of record for a company’s contracts, not another CLM. I’m deliberate about not calling it that; CLM as a category overpromised and underdelivered for a decade before AI ever showed up.
Past that, we’re investing heavily in preference learning, tuning the AI to a specific lawyer’s or company’s judgment, because reviewing a contract is closer to a recommendation problem than an objective problem.
The bigger vision beyond that is what I’d call artificial employees: AI agents that work continuously in the background, surfacing risk and picking up the kind of work a lawyer would otherwise hand to a junior associate, not replacing the lawyer, but never clocking out. Right now, AI behaves like an employee who only works the five minutes you’re watching them. Making it behave like a genuinely good employee, all the time, is the next real unlock.
9. Looking ahead five to ten years, what excites you most about AI’s role in the legal profession?
The access problem is the whole reason I started this. Most people can’t afford a lawyer when they actually need one, and a huge share of small businesses skip legal advice entirely because of the cost. Nobody should lose a fair deal, or get stuck in a bad one, just because they couldn’t afford an hour of legal time.
We’re not there yet. AI may never fully replace a lawyer’s judgement. The real opportunity over the next five to ten years is letting legal teams cover far more ground than they can today, with AI doing the first pass on everything so people can spend their time on the calls that actually require judgment.
10. What advice would you give entrepreneurs building AI companies for risk-averse, highly regulated industries?
Start with a problem you truly believe is important to solve–even if you aren’t sure how to solve it. The importance of the problem will propel your team to keep going. And when the nut is finally cracked, you’ll be there first. We spent years on legal efficiency before large language models made our original idea possible. If we’d quit in year two or three, we’d have missed the entire moment. We believed the outcome was important enough to keep going.
And go where the urgency actually is, not where the market looks biggest on a slide. We assumed Big Law was our market and in-house legal was an afterthought. It turned out to be exactly backwards, because in-house teams don’t have a billable-hour model quietly working against them wanting things done faster. Regulated industries are risk-averse for good reason. The founders who win are the ones who build trust and verifiability into the product from day one.
More Articles: