The problem isn't new. Product managers at fast-growing companies swim in oceans of user data—analytics dashboards light up with colorful graphs, support tickets pile up in Zendesk, Slack channels overflow with customer complaints—and yet the simple question of why users actually leave remains maddeningly elusive.
Matthew Wong and Mojmír Horváth believe they've found a way through. Their two-person outfit, Poth Labs, emerged from Y Combinator's Summer 2026 cohort with what they're billing as "the customer brain for your company"—a platform that dispatches AI agents to autonomously hunt for answers buried in the noise.
It's an audacious pitch for a team of two. And they're entering a market that's suddenly gotten very crowded.
The Core Idea: Stop Asking, Start Investigating
Poth's approach differs from traditional analytics in one crucial respect. Instead of forcing teams to manually parse dashboards or dispatch surveys with pre-written questions, the platform pulls data from across a company's ecosystem—Fireflies recordings, Slack threads, CRM notes, support tickets, product analytics, even documentation stored in Confluence—and then goes hunting.
A product lead can type a plain-language query: Why did users churn last quarter? Why are people abandoning checkout? The system surfaces patterns, identifies behavioral segments, and recommends actions. When gaps appear in the data, it doesn't shrug. It generates targeted follow-up surveys or interview questions on the fly, adjusting in real time based on how users respond. The questioning adapts rather than following a rigid script.
The company describes this as hypothesis-driven investigation. According to Poth's methodology documentation, three technical pillars support the system: adaptive surveys powered by language models, a knowledge graph mapping relationships between data points, and autonomous agents that form theories about causation and validate them statistically.
Whether that's enough to stand out is another question entirely.
A Market Already Moving
Poth launched in late June 2026 into a landscape where established players were already sprinting. Amplitude announced Agentic AI Analytics features earlier that year, with executives touting specialized agents tied to business outcomes on earnings calls in May. Mixpanel rolled out its own AI agent as the centerpiece of its product intelligence stack in May and June. PostHog has been shipping AI-powered analytics tools throughout 2025 and into 2026.
The research platform side isn't sitting still either. Qualtrics introduced Experience Agents in March 2025 and published content on agentic market research by May 2026. UserTesting updated its AI capabilities in April 2026. Newer ventures like Measure Protocol's Predict (launched June 4, 2026) and Pogo (announced June 9, 2026, with $32 million raised to date) suggest the field is getting more competitive, not less.
Poth's founders seem to be betting that integration is the differentiator—qualitative feedback and quantitative data treated as one continuous loop rather than separate workflows. It's a cleaner story than many competitors tell, but defensibility in software has always been hard to predict.
The Team: Lean and Ambitious
Wong, the CEO, spent time at Palantir before co-founding Poth. Horváth, handling CTO duties, comes from Tietoevry. Both studied at top institutions—Stanford for Wong, Phillips Academy and HTL-Spengergasse for Horváth. According to Y Combinator's company listing, the team size remains just the two of them as of mid-July 2026. They list their base as both New York and San Francisco, though the company's privacy policy names a Market Street address in San Francisco as the GDPR data controller location.
Funding, so far, appears limited to Y Combinator's standard $125,000 investment. Dealroom lists that investment as arriving in June 2026. There's no pricing page live on the site, no customer logos displayed, no published technical papers detailing the statistical validation methods the founders reference. Interested companies are directed to book a demo. A YouTube video embedded on the Y Combinator Launch post offers a walkthrough, but granular details remain scarce.
That's not unusual for a company fresh out of the accelerator—Demo Day, scheduled for late August 2026, will likely reveal more about traction and go-to-market plans. Still, for a startup targeting enterprise product teams, the lack of public case studies or methodological transparency feels conspicuous.
Trust and the Agentic AI Moment
The timing is both opportune and tricky. Gartner's May 2026 Hype Cycle for Agentic AI and Forrester's June commentary on the state of agentic systems point to genuine enterprise interest, but not without reservations. Forrester's 2026 Security Survey found that 49 percent of security decision-makers cited agentic AI as a concern—a notable figure given that these systems often require access to sensitive customer data.
Building trust around autonomous agents that poke through internal systems and conduct adaptive interviews will be critical. Especially when those agents are being pitched to teams already burned by overpromised analytics tools.
Poth's premise resonates because the pain is real. Product managers are drowning in data. Static dashboards do fail to explain causation. And the promise of an AI that investigates rather than just reports is compelling in theory.
Whether two founders can deliver on that promise at scale, against competitors with multi-million-dollar war chests and established customer bases, is the question they'll spend the next year answering. The market is watching—perhaps more skeptically than the founders might hope, but watching nonetheless.
For now, Poth Labs remains a bet on a very specific future: one where product intelligence isn't just smarter, but genuinely autonomous. We'll see after Demo Day whether investors share that vision.
