How Madhav Bhagat Built SpotDraft by Betting Against His Own Vision (Twice)
The SpotDraft Co-Founder spent five years building the wrong product while waiting for AI to catch up. Then the technology arrived, and everything changed.
The collapse of an entire business model can be compressed into a single number. In 2022, SpotDraft was spending $40 to review a complex contract using artificial intelligence. By 2024, that same review cost four-tenths of a cent.
A 10,000x collapse in unit economics doesn’t just improve margins. It resurrects dead ideas, validates forgotten visions, and turns yesterday’s impossibilities into tomorrow’s table stakes.
Madhav Bhagat, Co-Founder and CTO of SpotDraft, watched these cost metrics plummet in real-time as GPT-4 and Google’s Gemini models rolled out. The metering infrastructure his team had painstakingly built, designed to track every expensive AI query, suddenly became more expensive than the thing it was metering.
We went from $40 to four-tenths of a cent over the course of two years. So now my metering became more expensive than what I was metering.
This wasn’t a pivot. It was a resurrection. The AI-powered contract review tool they’d abandoned in 2018 as technologically impossible was suddenly not just viable, it was their competitive advantage.
In February 2025, SpotDraft closed a $54 million Series B led by Vertex Growth and Trident Partners, hitting a valuation between $190-200 million. The company now serves over 400 organizations globally, including Panasonic, Airbnb, and PhonePe, with more than 60% of its revenue coming from the United States. For fiscal year 2024, the company reported ₹60 crore (approximately $7.2 million) in revenue, a threefold jump from the previous year.
Check out the video of the conversation here or read on for insights.
The Education in Failure
Madhav’s resume reads like a blueprint for Big Tech success. Carnegie Mellon Computer Science degree. Senior software engineer at Google, where he helped scale Google Classroom from a two-person project to tens of millions of daily active users. The classic trajectory toward a comfortable VP role.
But he didn’t leave because Google was failing. He left because it was succeeding-slowly, bureaucratically, perfectly.
I built this AI solution in Google Forms and it had been live within Google for months. We just could not get it launched because there was all this operational checklist, bureaucracy almost. The feature I had built was released two years after I left, using the exact same code.
What followed wasn’t a straight line to startup success. It was three distinct experiments in failure, each one teaching a different lesson about what not to build.
First came DrinkLink, a nightlife concierge service in New York City where venues would bid on groups trying to get into clubs. The business processed $3 million in gross merchandise value but collected on only about 10% of it. Venues paid based on final bills, creating a collections nightmare. The business required staff on the ground every night, making it operationally intensive and impossible to scale.
We realized that while this is a great lifestyle business, scaling it requires just so much. Running an operational heavy business in America is very, very hard.
Next came CoderBhai, a software development shop he started after moving back to India in 2015. It was cash-flow positive but unscalable, caught between Indian clients demanding rock-bottom prices and those wanting enterprise quality without the budget.
The sequence matters. The failure of an operations-heavy consumer company and the struggle to scale a low-margin services firm created what Madhav calls “scar tissue.” These weren’t detours. They were an education in business model viability that would directly inform SpotDraft’s strategy: build a high-margin, globally scalable, product-led SaaS company.
When Technology Isn’t Ready
SpotDraft was supposed to be an AI company from day one. Co-founder Shashank Bijapur, a Harvard Law School graduate and former Wall Street attorney, had spent years manually reviewing contracts to extract key rights and obligations, work that felt ripe for automation. Madhav had just watched Transformer models change machine learning overnight.
In 2017, they built an early version using BERT, then the state-of-the-art language model. They hit two insurmountable walls almost immediately.
First, the technology was immature. Even with extensive fine-tuning, their models could only achieve 70-80% accuracy, insufficient for high-stakes legal contracts.
Second, they faced the classic AI “cold start” problem. To train their models, they needed thousands of examples of good and bad contracts from each customer. But legal teams don’t keep systematic records of rejected contract versions.
We explicitly in our contract said, you have to give us 100 good examples and 100 bad examples so that we can train our AI on top of it. One customer tried really really hard, but they were only able to find some 50 contracts. And when we started running our engine on it, 25 of these are duplicates. Three people send the same file from their inbox because you never store the bad version.
They’d built a solution to a real problem, but the technology wasn’t ready and the data didn’t exist. It was a product vision ahead of its underlying infrastructure.
The team had built a simple contract editor as a means to collect training data. Customers started using the editor. Then they started asking for more: templates, approval workflows, version control.
By listening to these requests and systematically building them out, the team inadvertently constructed a full-fledged Contract Lifecycle Management platform.
We didn’t even know that we were building a CLM till we had built some parts of it. A lot of that build was just happening based on user feedback.
The moment of true validation came from an unexpected source: a bug report. WhatFix, an early customer, opened a support ticket. Something was broken. A deal was stuck because of it. The SpotDraft team cheered.
Everyone was cheering that someone has opened a ticket, means that they’re actually using it. That was the moment that, okay, clearly we are solving a problem because they’re like, I need this fixed because my deal is stuck because of this. Now we are in the critical part of your business, which means we are adding value.
Product-market fit isn’t when people like your product. It’s when your downtime becomes their emergency.
By 2019, SpotDraft had a working CLM platform. It wasn’t sexy. It wasn’t the AI-first vision they’d started with. But enterprises were paying for it, expanding their usage, and building business-critical processes around it.
The COVID Tailwind
March 2020 brought an involuntary digital transformation mandate. Paper-based contract processes became untenable overnight. Gartner reported a 40% surge in inquiries for contract lifecycle management tools that year.
But there was a second, equally important shift: all enterprise sales moved to Zoom.
We were also getting on Zoom calls to sell, they were also getting on Zoom calls. So no one had that advantage of, oh, I’ll drop by your office. It leveled the playing field for us.
This is the structural advantage that remote work created for global SaaS companies building from India. Geographic proximity had been a genuine competitive edge. Zoom neutralized it. The best product and go-to-market strategy could win, regardless of where the team sat.
The customer base grew to include PhonePe, Airbnb, and Panasonic. The platform was working. The business was scaling. And then, in late 2022, the technology finally caught up to their original vision.
The Technology Catches Up
GPT-3.5 launched in late 2022. GPT-4 followed in early 2023. These models could do what BERT couldn’t: reason about complex legal language without massive fine-tuning. More importantly, they could do it with general-purpose instructions rather than thousands of customer-specific training examples.
Madhav saw the implications immediately. He called an all-hands meeting. The lawyers on the team were skeptical.
I was like, see, if you ask it, you can tell it to think in steps and then it gives you the right answer. This is going to be a big deal. And people were just like, you know, this engineer is doing engineering things.
But Madhav was right. These new models, even with zero fine-tuning, were matching or exceeding the performance of SpotDraft’s old, heavily customized BERT models. The breakthrough solved both the accuracy problem and the data problem. And the collapse in inference costs made the economics suddenly work.
SpotDraft launched VerifAI, an AI-powered review tool that works inside Microsoft Word and can accelerate review times by up to 15x. Smart Data Capture followed, automatically extracting over 1,000 types of metadata from contracts. Then came Sidebar, a conversational legal AI agent tuned specifically for lawyers’ workflows.
The pricing model shifted too. The complex credit-based metering system was scrapped in favor of a simple 20% uplift on per-seat pricing for AI features.
Earlier it was a per seat, some fair usage policy, you click a button, you consume a credit, all of that we can now remove. We can just say, okay, we are going to charge you another 20% on your per seat cost if you want AI. That’s it.
The company that had patiently built a successful CLM business while waiting for AI to mature could now deliver on its original vision-from a position of strength, with hundreds of paying customers and a proven platform.
The Opinionated Strategy
SpotDraft’s competitive positioning is deliberate. Madhav explicitly models the company as “the HubSpot for CLM,” in direct contrast to the “Salesforce” model of legacy incumbents like Icertis and Sirion.
Legacy enterprise CLM platforms are infinitely customizable, which means they require extensive and expensive implementation cycles. One SpotDraft customer had been trying to go live with a legacy platform for two years. SpotDraft contractually committed to getting them live in 90 days. They delivered in 60.
We were a lot more opinionated in that, you know, this is how it works. Instead of it being a customization that you can build on top of the platform, we made it a configuration that you can choose one of few paths.
This is the classic disruption playbook: take a complex, services-heavy enterprise product and rebuild it as an opinionated, best-practices-baked-in platform that can be deployed in weeks instead of years. The tradeoff is real-some customers with highly bespoke needs won’t be satisfied.
It’s a double-edged sword. People who fall under our exact ICP, they love it. But as soon as someone needs some change outside of that, it becomes, what do you mean this is not possible?
The strategic bet is that while incumbents increasingly focus on complex enterprise needs to justify their high costs, they’re leaving the mid-market underserved. If SpotDraft can capture 20,000 mid-market customers-the HubSpot model—they don’t need to chase enterprise whales.
The Contract Lifecycle Management market is projected to grow from $1.6 billion in 2024 to approximately $3.5 billion by 2030-2032, representing a compound annual growth rate of 12.7%. Most businesses still rely on manual processes, email, and shared drives. The category is still being defined.
Even opinionated products get things wrong. SpotDraft baked in an early assumption that every contract would have one legal reviewer and one business owner. This worked for their first ten customers. It broke when they scaled.
If you build everything with that assumption, now undoing that assumption in the code, in the process, in every part of the tool is a pretty heavy lift. We’re still unbaking some of the things we baked in super early on.
The fix required a strategic shift in product discovery. SpotDraft initially over-indexed on Indian customers. When they aggressively expanded in the US market, they discovered fundamental differences. American customers wanted self-service configuration. Indian customers preferred white-glove onboarding.
The AI-First Mandate
In early 2024, Madhav made AI usage mandatory across SpotDraft. Not optional. Not encouraged. Mandatory.
Last quarter, the design and product teams shipped 20 changes without involving engineering. Customer success representatives use Cursor, an AI-native coding assistant, to answer technical questions directly. The HR team built an interactive engagement quiz using Replit, work that previously would have required an engineer for several days.
Earlier they would come to an engineer and ask, how does this work? Can I make this change? How much time will it take? All of those questions are now answered by Cursor.
This isn’t the “one PM to one engineer” ratio that some AI-native companies have claimed. SpotDraft still maintains roughly eight engineers per product manager. But the nature of the work has fundamentally shifted. Non-technical employees can now handle 60-70% of implementation work themselves.
SpotDraft has also built an MCP (Model Context Protocol) server, Anthropic’s standard for allowing AI agents like Claude to use external software as callable tools. By exposing SpotDraft’s core functions-search contracts, review contracts, extract data-they’re preparing for a future where these capabilities might be accessed not through their own interface, but by master AI agents orchestrating tasks across multiple systems.
What it allows is like you could maybe get away from building a lot of very very nuanced features because you can say, okay, this you can just go ask an MCP-enabled agent and it’ll give you the answer.
The pricing model is evolving too. Madhav cites Intercom’s shift to charging $1 per customer support ticket successfully resolved by AI-outcome-based pricing rather than seat-based subscriptions.
I would be willing to pay Cursor on a per successfully merged code basis. Because at the end of the day, that is the value. I know I’m only going to pay for things that actually add value.
The Walk-Away Moments
Madhav is often praised for resilience. But he’s also walked away from three ventures: DrinkLink, CoderBhai, and SpotDraft’s initial freelancer focus. His reframing is instructive: Resilience isn’t about sticking to a problem statement. It’s about sticking to the practice of solving valuable problems, even if that means changing what you’re working on.
The resilience part for me is being able and willing to stick and keep at it versus saying that, you know, this is the only problem I’ll solve. We went from freelancers to contract review to CLM then back to contract review. It’s all about making sure that you’re listening to your customers.
The company even moved the entire team from Gurgaon to Bangalore in search of better engineering talent.
We were having such a hard time hiring in Gurgaon. Everyone good we would talk to, they’re like, you know, I actually want to go to Bangalore. And we’re like, okay, we have to go to Bangalore. Otherwise we will never be able to hire.
The moves that mattered weren’t the ones where they stayed the course. They were the ones where they had the conviction to change direction.
The best founders don’t just respond to the market. They respond to the market while holding a thesis about where it’s going.
Madhav wasn’t wrong in 2017 when he thought AI could change contract review. He was just early. And in venture capital, being early and being wrong look exactly the same-until suddenly they don’t.
Listen now!
Other ways to listen:
Your Feedback matters
As always, I’d love to hear your thoughts! Whether it’s about this episode or ideas you’ve been playing around with, shoot me an email at ad@thepodium.in. Your feedback keeps these conversations going, and I’m always up for chatting about your startup ideas too.
Until next time,
Your Host,
Akshay Datt