Every AI Company Is Now a Media Company

Nishant opens Episode 8 with a small admission about how the guest booking itself happened: “In true founder mode, I reached out via cold DM after seeing your podcast on YouTube.” The guest is Kshitiz Sanghi, co-founder of Voosh, and the podcast Nishant had noticed is Kshitiz’s own side project, a hyper-specific show helping early founders understand how to get into Y Combinator, built out of weekly conversations with YC founders that double as his own window into how Silicon Valley actually operates from a base in India. It sets up Nishant’s own aside, which works as a fitting thesis for the whole episode.

“Every AI company today also needs to be a media company to have an authentic voice.”

What Voosh actually does

Voosh is a food-tech company built for restaurant chains across North America, helping them run their third-party delivery business profitably rather than merely surviving on it. The product automates the unglamorous back-of-house work that eats a restaurant operator’s week: reconciling finances, managing reputation, tracking ad spend, and handling disputes across DoorDash, Uber Eats, and Grubhub at once. Kshitiz is explicit that the company measures itself against outcomes, top line or bottom line, rather than usage metrics.

The path there was not straight. Voosh started in India as something closer to an OYO for restaurants, aggregating more than two hundred outlets in Bangalore, where the operational leakage across those kitchens became obvious fast. The first pivot chased 10-minute food delivery, until the capital intensity of that model — real estate and inventory on both the CapEx and OpEx side — made the economics too punishing. The second pivot is the one that stuck: solving back-of-house restaurant management as software, aimed squarely at the US, where the willingness to actually pay for software runs higher than in India. The first customer came from pure cold outreach and became something more valuable than a contract: a design partner who shaped the product with direct feedback and then referred the next customers in.

Know your ICP better than you know your product

Asked what he’d tell an early-stage founder trying to break into the US market, Kshitiz’s advice narrows rather than broadens. Know your ideal customer profile in specific, almost uncomfortable detail, he says, offering his own numbers as proof: pizza chains with under twenty outlets in Texas converted ten times faster for Voosh than the broader restaurant market did. Pair that precision with equally specific messaging, cold email and LinkedIn outreach that talks about the transformation a customer will experience rather than a list of features. Until product-market fit actually arrives, the mandate is to do things that will not scale, and he is precise about what PMF itself means: not interest, not sign-ups, but customers paying repeatedly for something they genuinely cannot operate without.

On the softer parts of building, mentorship and fundraising, his framing is almost stoic. Mentors become useful only once a founder has already taken full responsibility for their own decisions, not before. Capital follows conviction rather than the other way around, and the test he applies to himself is blunt: are you actually the best person in the world to be solving this specific problem, or just someone who happens to be trying? On Y Combinator specifically, his advice mirrors the same instinct. YC reads more than 30,000 applications a batch and is, in his account, unusually good at spotting anything padded or exaggerated. Voosh’s application was candid about what had worked and what had not, pivots included, and got in on the first attempt.

The moat, when anyone can ship the software

The question every founder in Kshitiz’s position eventually has to answer out loud is what happens to defensibility once building itself stops being the hard part. His answer skips the technical framing entirely. In an era where software creation is increasingly commoditised, he argues, the top sliver of companies wins on genuine customer obsession, a depth of understanding of a client’s actual workflow that a faster shipping cadence alone cannot replicate. It’s the same instinct that shaped Voosh’s own pivots: the company kept changing what it built until the problem it solved was one a restaurant operator would actually pay to never think about again.

His quick-fire answers compress the rest of his worldview efficiently. On the tech-moat-versus-wrapper debate, he sides with the wrapper camp more than most founders are willing to admit publicly, arguing that very few companies today have a genuine technical moat, and that the real work is being exceptional at building something useful and solving a real workflow, moat or not. On where value accrues in the stack, he expects applications to keep generating steady, dependable returns while physical AI and robotics become the far larger story over the next several years. His favourite accelerator, unsurprisingly, is YC itself, for a network and community he considers unmatched. And on the question of when to actually raise venture money, his answer is the same discipline that runs through the whole conversation: only once there is real traction and customers already paying, not before.

Asked for one closing piece of advice for anyone watching, he skips strategy entirely and goes for a trait instead. Stay curious, he says, calling it the single most important quality for solving problems that are actually hard.

From this conversation

How to Crack YC and Build AI Startups | ft. Kshitiz Sanghi | S7 Ep8

S07 · E08 · 41m

▶ Watch the episode