Shashank Bijapur was running due diligence reports on New Year's Eve 2017 at White & Case — instead of celebrating. He read a headline about Elon Musk building self-driving cars, and couldn't square the fact that cars were driving themselves while he was still copy-pasting contract language by hand. That moment became the seed of SpotDraft.

Industries like pharma and defense kept raising the same objection: where does our data go once it touches an LLM? SpotDraft partnered with Qualcomm to run contract analysis entirely on-device — in airplane mode, with nothing leaving the machine — making it the first CLM to offer that option.

The instinct was the same one that shaped the company from day one: if the data is legally required to stay private, you build around that constraint, not through it.

Around year five, Bijapur realized he had become his own company's bottleneck. He was answering every question himself until his team stopped trying to decide anything without him. His COO broke the loop by introducing OKRs — so people knew exactly what they could decide without asking.

Shashank (00:01) People who taught you in kindergarten are not the same people who are going to teach you in middle school and high school. You need to be able to find the right teachers at the right time.

Roland Siebelink — Host (00:42) Hello, everyone, and welcome to Breakthrough AI Operators, the show where we talk with founders who have made AI structurally native to how their company works and how they serve their customers. Not tool adoption, not productivity wins, but actual operating model breakthroughs. Today's guest, I'm very happy to say, runs an AI-native contract lifecycle management company built for legal teams that need speed without surrendering control.

The company has processed over a million contracts, raised a $54 million Series B, and recently doubled down on strictly on-device AI so that sensitive documents never have to leave the laptop. With that, everyone, meet my guest, Shashank Bijapur, the CEO and co-founder of SpotDraft. Welcome to the show, Shashank!

Shashank (01:31) Thank you, thank you, Roland and thank you for having me over.

Roland Siebelink — Host (01:34) Of course, it's a big honor. This is one of the big success stories in legal tech.

Shashank, let's dive right in. You're not selling AI features, you're selling contract operations with memory, as I understand it. Is that right?

Shashank (01:51) Yeah, that's absolutely right. We are helping lawyers automate a lot of their mundane legal work.

Roland Siebelink — Host (01:57) Excellent. Do lawyers all consider their work mundane?

Shashank (02:02) Well, if you have to review an NDA on a Friday night, week after week, then it does start getting mundane. And when you have the sales guy coming to you every week saying "can we sign this contract" — which has Somalia as the governing law — it does begin to start getting a little mundane.

Roland Siebelink — Host (02:20) Tell me a little bit about the origin story, Shashank. Were you already in legal or in legal tech before? How did this all get started?

Shashank (02:30) I like to call myself a recovering lawyer. After my masters at Harvard Law, I used to be a lawyer at a firm called White & Case in New York. And that's where I met my co-founder and CTO, Madhav, who used to be at Google in New York. And at some point, it was New Year's Eve and I was sitting reviewing contracts and running due diligence reports.

I saw the ball drop and I was thinking, what am I doing right now? And I opened the New York Times and I read about this guy called Elon Musk, who was building a self-driving car. In my head, I'm going: wait, cars are driving themselves and I'm still copying and pasting words on a contract. Something had to change. I went around looking at what different tools people use. And there was some amount of automation that happened in pretty much every single department, but legal was still in the Jurassic era. And that's where the idea of SpotDraft came about.

Roland Siebelink — Host (03:42) Okay, very good. And around which New Year's Eve was that, if I may ask?

Shashank (03:47) It was in 2017. That's when Madhav and I came back to start SpotDraft. But then the next two years, we were still trying to figure out what we were doing, how we were doing it, and if we were doing the right thing before we landed where we are right now.

Roland Siebelink — Host (04:04) And this was way before generative AI became all the buzz, right? Because ChatGPT only came out end of 2022, as I recall. What were you working with before? Was there even enough technology around to be able to entertain that hypothesis?

Shashank (04:21) Our original hypothesis was that lawyers spend a lot of time redlining contracts — if we can automate that, that'd be great. We used a bunch of machine learning and BERT, which was the technology then, to train on contracts to help analyze them and tell you the good, bad, and ugly.

What we realized very quickly is that there isn't a lot of training data for this. If it's called a confidentiality agreement, it's not coming to your inbox to get trained. You're having to train on fake data, on invented data, which is not the real world — so your accuracy is as good as a coin toss.

We said: is there a better way to do this? Where do lawyers live? They live in Word documents, and that's where they edit. So we can't ask them to label stuff — let's give them tools so that the contract review becomes faster. Then somebody asked us, how does it come in for contract review? And we said, well, that's called a workflow. Where does it get stored? Well, that's called a repository. Then we took two steps back and said, what is this animal called? And the answer was: it's called a Contract Lifecycle Management Platform. And that's how we ended up building the CLM so that we could train the data to help contract analysis happen.

AI isn't really an afterthought for us. We were built around the fact that contracts need to be reviewed faster.

Roland Siebelink — Host (05:59) Okay. And that makes you very different, I think, from all the other legal management platforms, contract management platforms that were quite prevalent in the SaaS era — when SaaS companies were so fashionable.

Shashank (06:15) It does make it very different and it also makes it problematic, because you don't look like anything else that's out there. If the state of the union is a cycle, nobody wants something different. They are comparing you against what's already there. In some ways, we had to degrade our existing technology to meet what was there, because what we were trying to build was, for the lack of a better term, ahead of its time.

Roland Siebelink — Host (06:47) Yeah, exactly. Getting the timing right is so important. Sometimes you can just be — I think it's my favorite management philosopher, Vladimir Lenin, that said you can only be one street length ahead of the crowd as a leader. Because if you're two street lengths ahead, they cannot see you turn the corner and then you'll lose them. Tell me, how did this company grow?

Shashank (07:11) We've pretty much doubled in size in every way for the last five, six years in a row — whether it's revenue, number of customers, number of people. It's been double or more. We were founded in 2017; 2018 was the first line of code; December 2019 was when we realized this stuff doesn't work, the accuracy isn't good enough. January 2020 was when we decided to make the pivot to the CLM. Three months after that was COVID.

It was a blessing for us, because that was a time when people realized that contracting could not be done just by tapping somebody on the shoulder and saying "hey, can you take a look at this contract?" They needed to build workflows, processes, audit trails.

Roland Siebelink — Host (09:08) You already mentioned your go-to-market a little bit and how in the beginning it was quite challenging to be seen outside of the category that people knew. What have you learned about your go-to-market?

Shashank (09:08) The first big learning is — taking a step back — as a lawyer, that need to be perfect was always there. We were always in product mode. We never thought GTM. We thought we had built the greatest technology available to mankind and it was just a matter of time before people came searching and found it. It wasn't until almost three years in that we hired somebody in sales and marketing. All of that was stuff that we didn't think about; we just kept building product, iterating on product, and taking feedback from customers. Whereas GTM should have been the first function that we should have attacked.

That said, our GTM is predominantly what we call all-bound. Essentially, we're direct to market, direct to consumers. We do a lot of social selling, a lot of value selling, and a lot of proof selling. That's our predominant way of attracting the market. It's a small set of lawyers, so you need to go educate them and build brand in that space. I'm not building this merely because I want to make money from it — I'm really passionate about how contracts are done and how inefficient they are. And just talking to people about it is a great way to sell.

Roland Siebelink — Host (10:02) How do you actually drag a traditional lawyer by their hair out of their Word document? It's almost the same as dragging an accountant out of their Excel. That must be very hard.

Shashank (10:15) Early days, that was the problem — we would have to tell people why CLM, or why technology at all, for that matter. I think we've moved past that phase and now the question is which technology, and educating them on what kind of technology does the thing that they need.

For example, a Claude or a GPT is probably a perceived threat to every business out there. Everybody's asking, well, I can use Claude or GPT to do that. Now we need to go and educate the customer: yes, there are certain things that Claude can do way better than anybody else in the universe. But there are certain things that you need as basic pipeline and plumbing to make that even effective.

Roland Siebelink — Host (11:05) Can you delve a little bit into that? Because I think a lot of AI founders struggle exactly with that question — how do you differentiate from the generic category leaders, the platforms like Claude and ChatGPT and Gemini, and get people to actually pay for a more verticalized solution?

Shashank (11:23) Large language models is what powers AI — language being the operative word. After Shakespeare, what is the most difficult language that you can read? It's that damn contract.

It is the only type of language in any business that requires you to go to law school, pass the bar, and be eligible to write it. Anybody can write code, anybody can write a PRD, anybody can write a tweet. But can anybody opine on law? No. That does require a specialized set of tools.

And second — it's also the only kind of language that costs $1,400 an hour in the real world to produce. And every dollar that's coming in or out of any business is because somebody signed a piece of paper. You need to get this done right. You need to get this done on time. A horizontal solution will not cut it. If in the real world you require a specialized person to do it, then you require specialized tooling to make that happen. We are the system of record that tells you what you've done, how you've behaved, what is acceptable and not acceptable — and then we take AI to derive knowledge from that.

Roland Siebelink — Host (12:52) How closely related is this to your decision to start processing on local machines and not in the cloud anymore? Can you tell us more about that? Because that's a bit of a different pattern than what I see most other companies do.

Shashank (13:08) Yeah, this was an intentional partnership that we did with Qualcomm. One thing that we kept hearing from people over and over again was the objection: where is my data going? Is it safe? Who's looking at it? Are you training on it? Is it going to an LLM? Or: my industry does not allow me to give data to an LLM. Let's look at pharma, let's look at defense — these are sectors where they may not be comfortable, even with a zero data retention policy, sending something to an LLM.

For those cases, we said: what if there was a way that you could process contracts on the device without it ever leaving? We are the only CLM today — actually, the first CLM ever — to have the ability to take a contract and do that analysis on the device, on airplane mode, with zero data leaving the platform. Which means you can choose which contracts you send to an LLM versus which ones remain on device.

Roland Siebelink — Host (14:11) So it's a dual strategy, right? You still support the cloud version, but for those specific very confidential and industry-protected cases you actually offer this as an alternative.

Shashank (14:25) Correct, that's correct.

Roland Siebelink — Host (14:27) Can you tell us a little bit more about the partnership with Qualcomm? Of course, don't share anything you wouldn't want public, but I'm interested on behalf of so many other founders who are always exploring partnerships and trying to make them work. I'd love to hear that story.

Shashank (14:45) Yeah, this was interesting. It was at an event where we met someone from Qualcomm who asked us what we were doing. We told them, and they said: look, this is one of those segments where there's a lot of demand for on-device AI. We are coming up with a new chip and we're looking to partner with companies that can do this. Usually with larger companies, the thing tends to be that you end up doing a lot of talking but action doesn't happen. But here we were two weeks later with two unreleased chips on a laptop for us to build our models and test them.

Roland Siebelink — Host (15:35) Tell me a little bit more about — you already mentioned doubling the company in size in almost every aspect for the last four or five years in a row. That in itself is a very challenging journey. What have been some of the learnings that you as the founder-CEO have had to go through about how to manage a company differently when it grows so fast?

Shashank (16:05) I think the first thing is: if you're in the early stages of starting a company and I had to give advice to myself about eight years ago, I would say — Shashank, remember, this is not a sprint, it's a marathon. Be prepped for it and be ready for what's coming.

Second is: it's people above everyone else. If you surround yourself and have the ability to attract the right talent, you can do wonders. And it's the right talent at the right time — knowing when to hire the right person and when it's time to tell them, look, we've built the foundation story, you love doing foundation work, why don't you move on and do what you love? And getting the new right leaders at the right time. It's not just identifying the right talent, but also ensuring that talent continues to evolve throughout the journey of the company. It's simple: the people who taught you in kindergarten are not the same people who are going to teach you in middle school and high school, and certainly not if you're doing your PhD in nuclear physics. You need to be able to find the right teachers at the right time.

Roland Siebelink — Host (17:16) If I may double-click on that a little bit — does that mean you have a certain tenure in mind that people would stay typically for, let's say, two or three years in a job and then you're expecting them to move on?

Shashank (17:30) To answer your specific question on what's the ideal tenure — I think if people stop learning and they stop growing and having fun, then they should do something else. And I hold myself to that standard as well. If I stop learning, if I stop having fun doing what I'm doing, I'll go to my board and say my time's up.

Roland, think about it — if the valuation doubles, the number of customers doubles, the number of people doubles, and your revenue triples in one year, your brain is not expanding three times over. You need the right people to help you do that. Otherwise, it's a downward spiral.

You're still thinking like a 10-person company when you've grown to 100. You want to be close to everything that's happening and then you become the bottleneck.

Roland Siebelink — Host (18:31) We've often talked to founders where almost their own brilliance becomes the bottleneck in a sense, because they're so used to answering every question and countering every challenge. And then at some point you start realizing: I've actually been blocking my own team's ability to answer questions for themselves. Is that something that rings a bell for you?

Shashank (18:56) Yeah, absolutely. This was me about three years ago. It creates multiple challenges. First, you become the bottleneck. Second, you hire people but don't give them the power to make decisions. Then you continue that loop because now they're like, hey, no matter what I do, this guy comes in and supersedes everything we say — so let's just go ask him.

Then it is a loop where you keep clogging decisions. That's when we got our COO — he joined us about four years ago and he changed some of the trajectory we were on. The first thing he came and asked was: where are your OKRs? And there I was, Googling OKRs.

Instead of going and bothering people about decisions that need to be taken, we now give them a broad overview of what they need to achieve — the objective, the key results — and we also give them the broad parameters of decisions they can and can't take. If you think this is right for SpotDraft, go ahead and do it.

Roland Siebelink — Host (19:54) Can you recall some of the key breakthroughs that you guys went through that really leveled you up to the next stage?

Shashank (19:58) One framework that's worked for us is: at a certain leadership level, if you come up with a problem, then you need to state the problem, state your proposed solution, and explain why you think it's the right solution. Then people aren't just coming and saying support or customer success or marketing is not able to do this — they're saying: here is how I think it can improve, and this is why I think it's the right decision. If they spend 10 minutes analyzing it, the questions automatically become fewer.

Roland Siebelink — Host (20:57) Taking a more meta stance, can you recall some of the key breakthroughs that really leveled you up to the next level?

Shashank (21:02) There was a point where product was ahead and GTM was lagging — GTM was trying to catch up to everything that we were building. Then there came a point where GTM went way ahead and product was trying to catch up: there was customer request demand, newer things being asked, and product was still catching up.

When you build things too fast, sometimes it breaks too. All of that was being matched up. Now we've reached the point where GTM and product run hand in hand. And I think that is a wonderful wave to ride.

The way you want to build product is not for revenue that's already been sold. You want to build product for revenue that's yet to be derived. You're thinking ahead of the market — thinking through what's happening in three to six months. Though it's become a lot more unpredictable with what's happening on the LLM front. But being in sync means: you hear what the customer is saying, you know what a prospect is saying, you know where the tickets are coming from — and you also know that all of that has already been solved by a product built before.

Roland Siebelink — Host (22:25) Excellent. That sounds like you have your flywheel running. What do you guys do inside the company to keep all those silos in sync? Because every department's growing, people start identifying more with their department than with the company at large at some point. How do you keep everyone working towards the common goal?

Shashank (22:47) Yeah, it's not easy to do that. We struggled for a few quarters, but I think we got the right set of meetings and communication logs going so that information was flowing through — whether company-wide or internally. Then we built a bunch of tooling around how we shared information. Third, we made a lot of the problem-solving self-serve.

Our engineering team has built out a wonderful agent that sits on top of our latest code base. You can ask it: hey, the customer is asking — can we do this? And the answer comes back in pretty much real time, rooted in code rather than in somebody's response. The learning doesn't have to happen in a meeting anymore. It can be asynchronous and on demand. And that's what is fun.

Roland Siebelink — Host (23:46) I love it. I was just going to ask you about agents because, of course, all these breakthrough AI operators are also building agents typically inside the company. Just to double-check — out of your 300 employees, how many are agents and how many are humans?

Shashank (24:00) Everybody is a human running at least a few agents by themselves.

Roland Siebelink — Host (24:05) Are you finding that you need to manage the company in a different way or add some sophistication just because you have more of these agents around? The way you share information, the way you share context, the way you provide guidelines. Are there some learnings we can take from you there?

Shashank (24:26) Yeah, I think one thing that happens is different people start using different tools and then it becomes a disjointed madness — people start putting their own credit card in and using whatever platform they like. And that was our worry because we've got customer data out there. Very early on, the team decided on what the policy was: what they could use, what data they could put in. And we tried to consolidate into one single tool, because then the learning wasn't just an individual's learning — it was the collective learning of the team.

And then setting up the right monitoring to make sure that all the PII and all the data we're not supposed to share does not get shared. We also did a bunch of training around it.

Roland Siebelink — Host (25:14) Okay, very, very good. Well, this all sounds super amazing, Shashank. At the last few minutes of the podcast I always like to delve a little bit into the person behind the company, if that's okay with you. Tell us a little bit more about where you grew up, what was your childhood and teenage years like, and was there something that predicted you would always be this successful lawyer and founder when you grew up?

Shashank (25:40) Well, I don't know if there was. I was always the guy who wanted to understand the why behind everything. The answer was always: well, this isn't the best way to do it. That was one thing.

But my parents both had jobs. And for me, I thought — they work so hard, but they always get paid the same amount. Is there a way that you can make more money if you worked harder? That got me into what I think entrepreneurship is — just that anti-incumbency factor more than anything else.

Both of them were bankers and I was blessed to have a very independent, very well-rounded childhood. I grew up around books, a lot of music, and a lot of culture.

When people start out, having the support of family is the most important thing. They're early investors — the earliest investors. They need to be okay with you missing family functions and being on a laptop when everybody else is at the barbecue, and forgiving you for these selfish desires and routines of yours.

Roland Siebelink — Host (27:02) Your parents, your brothers, sisters, your wife or husband at some point, your children even — they are truly your earliest investors and pull you through the hard times, right? That's a very good note to almost finish the podcast on, Shashank.

Asking again — what would be the one thing you would want to impart on every founder who's maybe just around early product-market fit and looking to grow the company as much as you did?

Shashank (27:38) As much as you focus on product, also please focus on go-to-market. They need to go hand in hand — one can't be ahead of the other. That's one. And second is: hang in there. Have faith in yourself. You've already done the tough job of leaving whatever you were doing to build what you want to build. It takes a little bit of perseverance to just cross the chasm, as they like to call it.

Roland Siebelink — Host (28:07) A little bit of perseverance — well, that's the understatement of the year, I would say. But I much appreciate that. Thank you so much, Shashank. Where should people go who want to learn more about SpotDraft? Is there something specific they should download or ask for?

Shashank (28:23) Yeah, we're on spotdraft.com — S-P-O-T-D-R-A-F-T.com. And I am on LinkedIn. I personally look at and respond to all messages. Please feel free to reach out to me.

Roland Siebelink — Host (28:39) Perfect. And of course, for people who know me but don't know Shashank yet and would like to get in touch, I'm happy to always provide an introduction. Shashank, this has been amazing. Shashank Bijapur, the CEO and founder of SpotDraft. And to the audience — please let's all thank Shashank for coming. And of course, next week we'll have another breakthrough AI operator on the podcast. Maybe not quite as amazing as Shashank, but still we'll do our best. Please stay tuned.

Shashank (28:47) Thank you.

Roland Siebelink — Host (28:48) Please stay tuned.

Shashank (29:09) Thank you, thank you, Roland. Thank you for being such an amazing host and for asking all the easy questions like we decided.