THE BIT BETWEEN THE ROUNDSAn Upstarts Club editorial series
Would You Walk Away From a $10 Million Business?
One founder sold a successful company to pursue a side project. Another struggled to sell an unfinished product. Two more spent years searching for product-market fit. Four founders, four different decisions — and no way of knowing at the time which would prove right.
The $10 Million Gamble
Andrew Antos had a successful software business generating around $10 million in annual recurring revenue. He’d spent years getting there, including two and a half years trying to sell AI contract review software to lawyers before finding a more receptive market among finance teams.
Then, at a customer conference, his team demonstrated a small internal tool they’d developed to help identify processes suitable for AI automation.
It wasn’t the main attraction. Klarity had spent the previous year developing more substantial products, but customers kept asking about the little tool instead. Some wanted to know how much it cost and how quickly they could buy it.
The team hadn’t even settled on a pricing model. Antos recalls that the first dozen or so deals were sold on different commercial terms as they worked out what customers would pay.
Within five weeks, the tool had generated approximately $350,000 in annualised recurring revenue.
Antos faced an unusual decision. He had an established business with customers, revenue and employees, but something relatively untested was attracting a level of demand he couldn’t ignore.
He eventually sold the established product business to a partner, transferring the people supporting it, and concentrated on the new opportunity. That business became Within, which Antos said had approached $20 million ARR within roughly 18 months.
It wasn’t his first change of direction. Klarity’s original legal contract review product had reached approximately $800,000 ARR, but growth had disappointed. Antos concluded that the market, rather than simply the sales operation, was the problem. Finance proved more promising.
The striking part of his story isn’t that he changed direction twice. It’s that the second time, he walked away from something that was already working.
Ryan Anderson, the founder of Filevine, took almost the opposite approach.
The Engineer Who Nearly Went to Amazon
Anderson was practising law when he started developing Filevine, a legal case management platform. He knew what he wanted the software to do but needed someone to build it.
An engineer named Jim Blake spent several weeks shadowing Anderson’s law firm, learning how the practice operated. Blake was preparing to join Amazon. Anderson couldn’t compete with the financial security of that opportunity, but persuaded him to become his co-founder.
In a 2026 interview, Anderson recalled that Blake tried to quit somewhere between 20 and 30 times during their first three years together. Anderson would talk him into staying.
The first significant commercial breakthrough came at a legal conference in March 2015, where eight law firms signed up. Unfortunately, the product was far from ready.
It lacked basic functionality, including calendar and contact management. Anderson was still practising law while travelling to customers’ offices to implement software that didn’t yet do everything they’d expected. Some customers were furious.
After those initial sales, new business was scarce. Anderson tried cold calling, advertising and attending more conferences, with little success.
What kept him interested was the behaviour of his existing customers. Despite the product’s shortcomings, they were using it increasingly often.
Eventually, a part-time marketing contact experimented with Instagram advertising. Anderson recalls receiving around 14 leads in a single day.
He was so determined to convert one prospective customer that when a lawyer in Alabama asked whether he’d be visiting the area, Anderson said he’d be there the following week. He had no plans to visit Alabama, but booked a flight.
Filevine went on to become a major legal software business. Yet the important distinction in Anderson’s early account is between the difficulty of acquiring customers and the value those customers were already finding in the product.
Antos had decided that his original market wasn’t attractive enough. Anderson believed his product had a market, even when he was struggling to reach it.
Both were making judgements based on limited evidence. Their subsequent success makes those judgements look straightforward. They weren’t.
Five Years of Experiments
Scott Stevenson and his co-founders spent five years developing Rally, a legal technology platform that struggled to achieve the traction they wanted.
During that period, the team ran hundreds of experiments, testing features, advertising messages and landing pages to discover what lawyers would buy. Stevenson described a process of repeatedly testing ideas, discarding those that failed and moving on.
The company remained deliberately lean. Stevenson believed that expanding before establishing product-market fit would be a mistake, although he acknowledged difficult conversations with stakeholders who wanted the business to grow faster.
Then generative AI created an opportunity that hadn’t previously existed.
The team began applying the technology to legal drafting and contract review. The resulting product became Spellbook.
In a March 2024 interview, Stevenson said Rally had served around 100 law firms, while Spellbook had reached close to 2,000. He reported that the business had grown twelvefold in a year.
Five years of experimentation had preceded a remarkably rapid expansion.
But there’s an important difference between Stevenson’s experience and Antos’s. Antos concluded that he needed to leave his original market. Stevenson remained convinced that legal services could be made substantially more efficient, even as individual product ideas failed.
His commitment was to the problem, rather than any particular version of the software.
It helped that the business had kept its costs under control. Had the founders built a much larger organisation before finding the right product, they might not have been in a position to exploit the arrival of generative AI.
Four Years to Find the Market
Richard Mabey’s experience at Juro offers a less dramatic account of the same challenge.
A former Freshfields lawyer, Mabey established Juro to simplify the creation, negotiation and management of contracts. He understood the inefficiencies of legal work, but translating that knowledge into a product businesses would consistently buy took more than four years.
Juro developed its software closely with early customers, including Deliveroo. Rather than treating initial sales as proof of success, Mabey focused on whether customers continued using and renewing the product.
His approach to sales was equally considered. Mabey has argued that recruiting salespeople doesn’t create demand. Juro invested in marketing and generating customer interest before attempting to scale its sales operation.
It’s a distinction that matters when a company is under pressure to grow. Hiring a sales director is an obvious response to disappointing revenue. Establishing whether sufficient demand exists for the product is a different, and often more difficult, exercise.
Juro’s progress was gradual. There was no sudden technological breakthrough or unexpected side project that transformed the business. Its development came through years of customer feedback, product refinement and learning how to build repeatable demand.
The contrast with Filevine is particularly interesting. Anderson could see customers using his imperfect software despite disappointing sales. Mabey placed considerable weight on retention and renewals. Both were looking beyond the number of new contracts signed.
The Benefit of Hindsight
These four founders made markedly different decisions.
Antos left a market that wasn’t growing fast enough, built a successful business elsewhere, then sold it to pursue another opportunity. Anderson persisted with an unfinished product because existing customers were using it. Stevenson kept experimenting within legal technology until advances in AI changed what his company could offer. Mabey spent years refining Juro around customer needs.
Their decisions look convincing because we know what happened afterwards.
Without that knowledge, the picture is less comfortable. Filevine had eight early customers and almost no new sales. Rally had spent five years experimenting without finding rapid growth. Juro was still searching for product-market fit after four years. Antos was preparing to sell an established business to pursue a product with only a few weeks of commercial evidence.
Any of those decisions could have ended badly.
There’s also an obvious limitation to these accounts: they come from founders whose businesses survived. Other founders have persisted for years, changed markets or pursued promising side projects without achieving comparable results.
That doesn’t diminish these stories. It makes the evidence behind the decisions more interesting.
Anderson watched customer usage. Stevenson tested propositions and monitored whether the economics supported scaling. Mabey examined demand and retention. Antos paid attention when customers showed unexpected willingness to buy something his team hadn’t intended to sell.
None of those signals guaranteed success. But they help explain why the founders made the choices they did.
Company histories rarely give much space to that uncertainty. The years spent experimenting become a short paragraph. The disappointing sales figures disappear. The difficult decisions are recast as obvious turning points.
Yet the original accounts contain details that are far more revealing: an engineer who nearly joined Amazon, a founder booking a flight to Alabama to win a customer, five years of experiments before a breakthrough, and a conference audience more interested in a side project than the products it had come to see.
The funding rounds are part of these companies’ histories. The decisions made between them tell us considerably more about how the businesses were built.
Original interviews
Andrew Antos — Klarity / Within: The Product Market Fit Show, August 2026
Ryan Anderson — Filevine: The Product Market Fit Show, March 2026
Scott Stevenson — Spellbook: The Product Market Fit Show, March 2024
Richard Mabey — Juro: The Everyday Founder, August 2025