Hyperbound co-founders Sai and Atul share how 25,000 hand-personalized LinkedIn messages led to accidentally building AI sales roleplay, why reinforcement learning discovers behaviors like "negging mid-market prospects," and their prediction that cold outbound will die within five years.
Sai and Atul built Hyperbound while trying to sell a completely different product. After sending 25,000 hand-personalized LinkedIn messages and 10,000 cold emails to land 2,000 user interviews, they created a barebones AI to practice sales calls. A CRO at a New York conference asked how two engineers closed $250K ARR in three months, and when they showed him the practice tool, his reaction was immediate: "Why aren't you selling that?"
Listen to this episode to learn why cold outbound will die within five years, how reinforcement learning discovers non-obvious sales behaviors like "negging mid-market prospects," and why one Fortune 10 company just upsold for $500,000.
Sai and Atul are high school friends turned YC founders who met in choir 14 years ago. Hyperbound is an AI sales roleplay and coaching platform that simulates realistic customer conversations, complete with prospects who hang up, get angry, or curse you out.
Sai brings natural language processing research dating back to high school, including published papers and patents. Atul combines engineering depth with go-to-market instincts honed at Meta Ads. Together, they scaled from zero to $1 million ARR in 11 months, maintained negative burn multiple (profitable), and now serve Fortune 100 logos including a Fortune 10 company.
Their reinforcement learning algorithm observes top performer behaviors, hypothesizes which patterns actually matter, then tests whether bottom performers improve when coached on those specific behaviors.
[00:00] The YC advantage: How "YC founder" in a LinkedIn tagline increased response rates by 50x
[02:45] What Hyperbound actually does: AI that hangs up on you, curses at you, and makes sales practice realistic
[05:30] The accidental product: How 25,000 LinkedIn messages led to building an AI SDR, then pivoting to roleplay
[09:15] The CRO moment in New York: "Why aren't you selling that?" changed everything
[12:00] Reinforcement learning explained: Why negging mid-market prospects actually works for one customer
[18:30] The Adam Robinson bot: How a viral LinkedIn competition books out three weeks of sales calendars
[25:00] Fortune 100 logos came inbound: Why enterprises are scared their competitors will adopt AI first
[32:00] Teaching reps to handle angry customers: Use cases consultancies can't deliver
[38:00] Co-founder dynamics: Why Friday strategy sessions keep everything in sync
[44:00] Hot take: Cold outbound will die completely within five years
Sai and Atul sent 25,000 hand-personalized LinkedIn messages and 10,000 cold emails to conduct 2,000 user interviews for their original AI SDR product. To prepare for sales calls, they built a barebones AI practice tool. At a New York conference, a CRO asked how two engineers closed $250K ARR in three months. "We have this AI. We practice with it before some calls," they explained. His response: "Why aren't you selling that?"
As we discuss in the episode, this pattern appears constantly among successful founders. The hardest problems surface while working on other problems. Many YC companies, including HockeyStack, discovered their real product through a similar pivot. For SaaS founders, the lesson is paying attention to the internal tools you build to solve your own problems: they often have broader market value.
Hyperbound's algorithm works in three phases: observe top performer patterns, hypothesize which behaviors actually matter versus coincidence, then test whether bottom performers improve when coached on specific behaviors. "We found out for one of our customers that negging your prospects for a specific segment actually yields higher conversion ratios," Atul explains. "You should try to disqualify hard. That is the best process for you to move mid-market customers with this job title forward."
This approach answers the question that stumps every sales leader: "What does good look like?" Correlation analysis produces thousands of patterns, most irrelevant. Reinforcement learning filters signals from noise by testing hypotheses against actual outcomes. At Understory, we see similar dynamics in paid media optimization: the behaviors that seem obvious often matter less than counterintuitive patterns discovered through rigorous testing.
One Hyperbound client previously spent nearly $2 million on their annual SKO: flying everyone in, hiring consultancies for in-person certification, running a three-day event. Hyperbound delivers equivalent training throughout the entire year at a small fraction of that cost. "This is really a generational unlock in terms of what you're able to do," Atul notes. "It's the definition of being able to scale with AI."
The economics shift dramatically. SKOs happen once per year, leaving 11 months where sales messaging goes stale and behaviors drift. Week-over-week coaching becomes possible when the marginal cost of training approaches zero. For enterprise sales organizations, this represents a fundamental change in how behavior change happens across large teams.
Hyperbound built an "Adam Robinson bot" that captured Adam's direct, abrasive communication style and let people cold call it. The post went viral, Adam engaged, Mark Kosoglow chimed in during their public beef, and the entire ecosystem amplified reach. "Every time we've done one of these competitions, it's been like a step function change for our inbound," Sai shares. "It books out roughly three weeks of our sales team's calendars. Even though we start disqualifying leads aggressively, it'll still book out."
As Alex notes in the episode, this type of content feels up for grabs: "Viral content is making its way to B2B professional spaces, especially as younger people are becoming leaders at these companies." The key is doing what no one else is willing to do. For SaaS companies building awareness, competitions and interactive experiences generate far more engagement than traditional thought leadership.
Hyperbound's largest logos, including publicly traded Fortune 100 and even a Fortune 10 company that just upsold for $500,000, all came through inbound. "These are folks who typically only listen to Gartner reports to make their purchasing decisions who are now actively looking for tools because there's a massive paradigm shift," Atul explains. "They're scared. If they don't adopt these changes, their competitor is and they're going to outpace them."
This fear-driven buying behavior is new. Enterprises that previously moved slowly on technology adoption are now proactively seeking AI tools. At Understory, we're seeing similar patterns in GTM engineering conversations: companies that ignored outbound automation for years are now urgently asking about Clay workflows and signal-based prospecting because competitors have already adopted these approaches.
One Hyperbound customer needed to teach reps how to lead with empathy when calling back angry prospects who had been mistreated by previous company leadership. "They're going to curse at you and you just got to sit there, take it, lead with empathy, and keep going," Sai explains. "That's not something a training consultancy is going to come in and teach your reps how to do. The only thing that can teach your reps how to do that is to actually get that repetition."
Human role-play partners struggle to sustain emotional intensity across hundreds of practice sessions. AI can maintain consistent difficulty, realistic anger, and unpredictable objections indefinitely. For enablement leaders, this unlocks training scenarios previously impossible to practice at scale.
Atul offers a five-year prediction: "Cold outbound will die completely. The AI gatekeepers are getting better. Email filtering, call screening on iOS 26 where Siri now picks up the phone for you." Android has offered call screening for years; Apple's adoption signals mainstream gatekeeping. The only viable paths forward will be warm introductions or paid advertising.
Sai adds that AI SDRs are also a short-lived bridge: "Why should a person talk to an AI sales rep to buy? Why not just go through an agentic form and click click click you're done?" As we discuss in the episode, sales is fundamentally human, but the efficiency at which we execute will transform completely. Signal-based and warm outbound will survive; pure cold volume will not.
Despite the appeal of remote work, Atul argues early-stage teams need physical proximity: "The best conversations Sai and I ever had about our product iteration was in the hallway or while we were making lunch together. It's the times when you try to turn your brain off where those little pieces of interactions happen that take you so far."
Structured Zoom blocks don't elicit the creativity that serendipitous interaction produces. After 10+ employees, remote becomes viable. Before that, the iteration speed advantage of in-person collaboration compounds daily. For co-founders especially, blocking dedicated strategy time (Sai and Atul block all of Friday) keeps alignment tight as companies scale.
Want more insight on AI-powered sales training and go-to-market strategy? Listen to the full episode on YouTube and subscribe to Understory's podcast for more insights on scaling SaaS growth.
Looking to coordinate your paid media, outbound, and creative without managing multiple specialists? Book a call with Understory to explore how expert allbound execution can accelerate your pipeline.
SmartLead founder Vaibhav Namburi shares how 11 failed businesses, WhatsApp community presence, and platform-as-a-service thinking built an email infrastructure managing 5-7 million mailboxes and 110M+ daily warmup emails.
Catch up on our Understory Unfiltered episode where Nick Velkovski shares how he bootstrapped HeyReach to $7M+ ARR from Macedonia after Y Combinator rejected him twice.
Catch up on our Understory Unfiltered episode where Eli Rubel shares how he scaled MatterMade to $4.2M EBITDA, why VCs are not doing good in the AI era, and how he vibe-coded a SaaS product in seven days.
