Two founders left Cairo for San Francisco, built an AI recruiter that ran a hundred interviews in five days, then walked away from it. Their new bet: a simulated population that tells you what a market will do before you spend a dollar finding out.
Most companies find out what a market wants the expensive way. They run the survey, ship the product, buy the ad, set the price, and then wait for reality to grade the decision. Outrove, a two-person company from Y Combinator's Summer 2025 batch, is built on a stubborn observation: people are unreliable narrators of their own behavior. Ask them what they'd pay, and they guess. Watch what they actually do, and you get a different number. Outrove's product tries to close that gap by building a simulated population of real people and running your decision against it first.
The company is based in San Francisco. Its founders, Saif Elhager and Ahmed ElShireef, are from Cairo. Elhager moved to the Bay Area in January 2025; ElShireef followed within weeks. What they are building now - under the product name Sapien - is not what they set out to build. And that turn is the most interesting thing about them.
Outrove launched in March 2025 as an AI recruiter. The pitch was concrete. The software found candidates across the web, sent personalized outreach, and then did the part everyone dreads: it ran the interview. A digital interviewer with synthesized voice and rendered video could join a Google Meet call, ask follow-up questions, react to answers, and grade the candidate on the way out. It charged around $750 per successful hire.
The idea did not come from a whiteboard. Both founders had built earlier companies that leaned on offshore teams across Egypt, India, and Pakistan, and both had felt the same friction - screening was slow, results were inconsistent, and feedback loops dragged. They automated their own hiring first. Then they noticed the tooling might be worth more than the hires.
The proof point arrived fast. Within a month of launch, an advertising agency hired Outrove to source offshore Amazon specialists and designers. The system ran 100 interviews in five days - a stretch of work that had previously taken weeks. By most measures, this was a product that worked.
"We wanted to build something significant, not just make money."Saif Elhager, Co-Founder & CEO
Here is the part that separates Outrove from the pack: the founders had a functioning product with paying interest, and they still changed direction. Somewhere in the process of building an AI that could conduct a convincing interview, a larger question surfaced. If you can simulate one human well enough to hold a real conversation, how far can you push the simulation? Not one candidate - a whole population. Not a job interview - a decision.
That question became Sapien. Instead of interviewing people you might hire, the platform builds simulated populations of real people so a company can test a decision against them before committing. What price should we set? Which campaign should we run? What will happen if we launch this product, or pass this policy, or move this market? Outrove's answer is to ask a synthetic version of the audience first.
The distinction sits right in the product's taglines: "Test the decision before you commit to it," and, more bluntly, "Sapien tells you what people will do, not what they'd say." It is a market-research company wearing an AI-simulation coat.
A synthetic population is only as good as the ground it stands on, and the obvious risk is that you build a hall of mirrors - agents that agree with each other and drift away from reality. Outrove's design tries to guard against that with a four-part loop.
The personas are meant to be evidence-based - assembled from interviews, records, and observed behavior rather than invented from a prompt. Segmentation shows how different slices of an audience react differently to the same move. And the piece worth noting, because most AI companies quietly skip it: every prediction is logged against what reality returns. The model is asked to be wrong in public, on the record, so it can get less wrong over time.
"Every prediction is logged against what reality returns."Sapien, company motto
Does it work? The company points to a case study in which its model landed 4.2 times closer to actual shelf reality than traditional human conjoint research when valuing an aluminum-free product claim. Treat a single case study as a signal, not a settled verdict - but the direction is the point.
Outrove is aimed at anyone who makes a costly call and would like to be less surprised by it - companies weighing a launch or a price, investors sizing a market, public-sector teams gauging a policy. The competition splits into two camps. There is the old guard it wants to replace: surveys, focus groups, and conjoint studies, which are slow, expensive, and famously good at recording what people say. And there is a new wave of AI-simulation startups - names like Artificial Societies, Aaru, and Societies.io - all chasing the same idea that you can model an audience instead of polling it.
| Approach | Speed | Signal |
|---|---|---|
| Focus group | Weeks | What 8 people say while watched |
| Survey / conjoint | Days to weeks | Stated preference, guessed |
| Outrove / Sapien | Minutes | Modeled behavior, logged vs. reality |
The founders' earlier recruiting product also had rivals - Mercor, Micro1, HeyMilo, Ribbon - which is worth remembering, because it tells you Outrove is willing to be in a crowded room if the room is big enough. The differentiator it leans on now is accountability: not just a plausible answer, but a paper trail comparing the answer to what actually happened.
Neither founder is new to building. Saif Elhager, the CEO, previously built AiHello, an e-commerce advertising automation company he scaled to nearly 40 people and seven-figure revenue - reaching Amazon Advanced Partner status before he was 20. Ahmed ElShireef, the CTO, led an AI speech-therapy application with a team of around 20 and has worked on a UN-recognized robot-doctor project. One brings the sales-and-fundraising instinct, the other the engineering depth.
Built AiHello to seven figures before 20. Sales and fundraising instinct; the outward-facing half of the pair.
Ran an AI speech-therapy app; deep engineering background. The build half of the partnership.
There is a line from ElShireef about landing in Silicon Valley that captures the company's temperature better than any pitch deck: you go from being the top fraction of a percent back to being a normal kid again, so you restart the competition and work harder and faster. Outrove reads like a company built by two people who assume they are behind and act accordingly - which may be exactly why they were willing to abandon a working product for a bigger one.
"You go from being the 0.1% to being a normal kid again. So you have to restart the competition - work harder and faster."Ahmed ElShireef, Co-Founder & CTO
Outrove is a B2B software company. In its recruiting form, it charged per successful hire. As Sapien, it sells access to the simulation platform, with demos booked directly and a small founding team behind it. Public sources report it raised over $1 million from Y Combinator and angel investors, with its Summer 2025 batch stamp attached; some database listings show smaller figures, and the company has not published exact terms, so treat the numbers as approximate. Revenue is undisclosed.
What is clear is the shape of the bet. Market research is an enormous, slow, and much-complained-about industry, and Outrove is trying to compress its core loop - ask the audience, get an answer, check it against reality - from weeks into minutes. If the accuracy holds up outside the case studies, the pitch writes itself. If it doesn't, the logging-against-reality discipline is at least a healthier way to find out than most.