The pitch sounds almost too good to automate: What if investment due diligence—those frantic weeks of coordinating expert interviews, parsing data rooms, and synthesizing findings into coherent memos—could be handed off to AI agents that do the legwork themselves?
LATO, a recent Y Combinator graduate, has built its business on exactly that proposition. The startup, still just two people, is selling what it calls an "agent-native research and simulation platform" to private equity and venture capital firms. The system deploys AI agents to conduct voice interviews, digest proprietary fund data, and construct market simulations—all aimed at delivering investment analysis faster and cheaper than traditional diligence workflows allow.
So far, a handful of European investors have signed on. Whether the model can scale beyond early adopters, or whether firms ultimately prefer human experts to synthetic interviews, remains very much in play.
Voice Agents and Digital Twins
At the heart of LATO's product is a somewhat audacious claim: that AI can reliably conduct structured voice interviews at scale. The platform claims to access a network of over four million verified participants, though the company hasn't disclosed which data providers or expert networks feed that pool, and this figure remains unverified by independent sources. Marketing materials showcase studies with more than a hundred completed interviews—tracking not just what respondents say, but how they say it, measuring tone, sentiment, conviction.
It's an appealing vision for time-strapped analysts. Instead of spending days scheduling calls with former employees, customers, or industry observers, an investment team could theoretically task an AI agent with running dozens of parallel conversations, then surface patterns and outliers for human review.
But LATO's ambitions extend beyond automating phone calls. The second layer ingests a fund's internal knowledge—past deal memos, CRM notes, recorded transcripts, data room files—and cross-references that with public filings and market intelligence. The idea, according to the founders, is to build what they describe as a "company brain" that accumulates institutional memory across deal cycles, getting smarter the longer a firm uses it.
The third piece is where things get speculative. LATO converts interview data into simulations populated by "digital twins" of the people it's interviewed. In theory, this allows investors to run what-if scenarios—say, testing how customers might react to a 15% price increase—without going back into the field. It's the kind of feature that sounds transformative on a pitch deck. Whether it holds up under real-world scrutiny is another matter.
The platform itself lives in the tools investors already use. An Excel add-in (which, for what it's worth, carries a five-star rating on Microsoft's AppSource marketplace, though that's based on just six ratings) lets analysts pull research directly into financial models. LATO also hooks into Gmail and Slack, embedding its workflows where deal teams operate day to day.
The Team Behind It
Tymek Staniszewski knows the pain points firsthand. He spent six years at Verdane, a growth-stage private equity firm, cycling through the kind of diligence sprints LATO is now trying to short-circuit. His co-founder, Tien Chu, brings machine learning chops—years spent in enterprise R&D labs and early-stage startups, with an academic foundation in computer science, mathematics, and philosophy from the University of Warsaw.
They've kept the operation deliberately small. Y Combinator's directory lists the headcount at two, though this may change as the company grows. Yet they've managed to sign a few early customers: Blume Equity, FoodLabs, Montis.vc, and Concept Ventures all appear on LATO's website, complete with testimonial quotes from principals and managing directors. (Those relationships haven't been independently verified, and client lists from young startups often reflect pilot projects rather than deep integrations.)
A Crowded Moment for Agentic Finance Tools
LATO is hardly alone in betting that AI agents will reshape financial research. The past year has seen a flurry of similar product launches. Orbit debuted an agent builder for investment research. Coinbase rolled out infrastructure for agent-driven trading workflows. LTX added agentic features to its fixed-income platform, and Waton Financial introduced a multi-agent workbench for institutional investors.
Established players are adapting, too. AlphaSense, which built its business on searchable earnings calls and market intelligence, has been weaving agent workflows and Excel integrations into its offering. Tegus, with its library of expert transcripts, operates at a scale LATO can't yet match. Traditional expert networks like GLG still control much of the high-touch primary research market, while newer entrants such as Solstice and Bastion are carving out niches in AI-native intelligence for private markets.
Academic researchers are paying attention. A peer-reviewed paper in an MDPI FinTech journal recently explored the architecture and systemic implications of AI agents in financial markets, a sign that the conversation has moved beyond vendor hype into more serious inquiry.
LATO's bet is that integrating primary and secondary research into a single agent-driven workflow—voice interviews feeding simulations that flow into Excel models—will prove more valuable than buying discrete tools from specialist providers. Maybe. Or maybe investors will stick with the specialist approach, preferring battle-tested vendors for each piece of the puzzle.
Security and the Fine Print
For a product handling sensitive deal data, security details matter. LATO's documentation outlines standard protections: TLS encryption in transit, AES-256 encryption at rest, and a pledge not to train models on customer data. The company's privacy policy names Anthropic as its AI provider and notes European hosting with contractual clauses covering U.S. data processing. SOC 2 and ISO 27001 certifications are reported by the company as "in progress," though without independent validation or public attestation reports available yet.
Access isn't straightforward. There's no self-serve signup or public pricing page; prospective users must schedule a demo. The terms of service reference "usage-based plans with daily credit allowances," but specifics remain opaque for now—a common stance for enterprise startups still figuring out pricing and packaging.
The Bigger Question
For a two-person startup wading into a market crowded with well-capitalized incumbents and a fresh wave of competitors, LATO represents a particular bet about the future. Not just that AI will assist investment research—that's already happening—but that the work itself will fundamentally reorganize around autonomous agents conducting tasks humans used to coordinate manually.
The early customer logos suggest at least a few firms are willing to entertain that possibility. Whether it's a sustainable business or a feature that larger platforms will eventually absorb is the question every YC startup faces after Demo Day. LATO's founders, for their part, seem convinced that diligence as we know it is ripe for reinvention. The investors testing their platform will determine if they're right.
