SaaS Sessions

There are a handful of non-obvious points (the 1993 Word Track Changes observation, the pivot away from NDA analysis due to accuracy failure, LLM training starting in 2019) but much of the 31 minutes is consumed by origin story, analogies to photography/accounting, and generic founder advice. The ratio of signal to filler is low.

“The greatest innovation that lawyers saw was a patch in Microsoft Word in 1993. And that patch that they introduced was called Track changes.”

“we had five very, very, very large banks who are our customers and we failed at it because the accuracy was very poor”

The framing of legal language as 'the deepest and most expensive end of language' is a decent original angle, and the contrarian take that nobody will sign million-dollar deals on AI advice is mildly interesting, but most of the content leans on recycled startup advice (talk to customers, US is the biggest market, it's a marathon not a sprint).

“I don't think in the next 10 years. And this is Kind of contrarian that we are going to have people who say that I signed a million dollar deal listening to a robot lawyer. That's not going to happen.”

“it is at the deepest end of language, it is at the most expensive end of language and it is a language that is needed in everyday business context”

Shashank is a genuine practitioner - a Harvard Law graduate who co-founded and has run SpotDraft for roughly nine years serving real enterprise customers - which gives him credible first-hand authority. However, the interview format elicits mostly narrative and surface-level reflection rather than the deeper operational expertise he likely possesses.

“we had pressure from all sides saying we need to go enterprise. We need to go enterprise, but we remain truly mid market and continue to focus there”

“we lucked out where we knew that we knew when to stop, when we were failing”

A few concrete data points appear (1993 Track Changes patch, five large bank customers, $1400/hour legal cost, 2019 LLM experiments) but the episode relies heavily on hypothetical scenarios ('let's assume it's an airport') and completely omits revenue, ARR, growth rates, or named customers, which limits its evidential weight for operators.

“The only kind of language that goddamn costs fourteen hundred dollars an hour in the real world to solve for.”

“every new contract that comes in, it'll map out against your existing standards and tell you exactly where, what is missing... Here are the 24 contracts that require changes”

The host asks a reasonable range of questions (GTM ICP, risk mindset, AI impact) and lands one genuinely interesting angle on lawyer risk-aversion vs. founder speed, but there is no meaningful pushback, no challenging of vague claims, and the lightning round is entirely generic. Claims about AI accuracy and market positioning go completely unchallenged.

“how has your view on risk overall changed? Right. Because I think lawyers are typically trained to reduce risk.”

“Nice, nice. I think that that's, that's good.”