The pre-IPO finance team at any tech company worth its valuation knows the calendar by heart. Somewhere between Series C and the IPO roadshow comes SOX compliance season—that special circle of corporate hell where internal auditors spend months excavating evidence from email threads, hand-drawing process flowcharts, and testing controls one mind-numbing transaction at a time. The typical escape route? Outsourcing the work to consultants billing anywhere from $200 to $800 an hour, which can balloon into seven-figure annual expenses faster than most CFOs care to admit.
Enter Oxus, a San Francisco startup that emerged from Y Combinator's Winter 2026 batch with what sounds like the kind of promise that makes audit veterans raise an eyebrow: What if you could collapse those hours-long audit tasks into minutes?
The pitch is straightforward, almost deceptively so. Oxus has built an AI platform designed to automate the grunt work of internal audit—control testing, documentation, scoping—while keeping human auditors firmly in the loop for final approval. Think of it as a junior associate who never sleeps, doesn't bill overtime, and doesn't hallucinate compliance findings (or at least, that's the idea).
The Mechanics of the Thing
Here's what Oxus actually does, stripped of the venture-speak.
Control testing comes first. Upload your evidence—screenshots, spreadsheets, those scanned documents that someone really should have digitized years ago—and the platform applies test procedures across the entire population. It annotates the relevant sections, generates workpapers, and maintains an audit trail back to source materials. The kind of work that typically occupies junior auditors for weeks.
For walkthrough documentation, you feed the AI your process notes or interview transcripts. Out comes a flowchart and narrative description, no manual diagram wrestling required.
SOX scoping is the third pillar: the platform analyzes financial and qualitative risk factors to identify which areas merit closer attention. Not revolutionary conceptually, but potentially valuable if the AI can spot patterns a human might miss while drowning in spreadsheets.
Fourth is document intelligence—search, annotate, analyze. The platform can pull in web searches when needed, but crucially, Oxus says outputs are grounded in specific text excerpts rather than the kind of confident-sounding hallucinations that have plagued other generative AI tools.
Every workflow follows the same blueprint: upload, analyze, generate workpaper, review. Nothing becomes final without human sign-off. For a product targeting an industry where regulatory scrutiny can make or break companies, that's less a feature than a survival requirement.
Oxus also offers customer-hosted deployment and promises data isolation. "Never reused, shared, or retained for model training," the company's website states—a reassurance that matters more in finance than in, say, marketing automation.
The Builders
The founding team is compact: three MIT graduates with overlapping credentials that read almost identically on paper, which is either efficient or slightly unusual depending on your perspective.
CEO Janet Liu holds degrees in Computer Science, Economics, Data Science, and Finance from MIT. Before Oxus, she researched digital payments at the Federal Reserve and worked on investment strategy at D.E. Shaw. COO Melinda Liu (no relation clarified) has the same MIT degree combination and brings fintech AI product experience plus a stint at JPMorgan. CTO Christine Watts studied Artificial Intelligence at MIT and built machine learning systems at startups before consulting stints at BCG and NERA.
They went through YC's Winter 2026 cohort with Diana Hu as their group partner and launched publicly around late 2025 or early January 2026, based on their Launch YC post.
A Market Waking Up, Perhaps Faster Than Expected
The timing might be sharper than it first appears. Internal audit is experiencing something of an AI awakening—though whether that translates into actual adoption at scale remains an open question.
The Institute of Internal Auditors' 2025 North American Pulse survey found that 40 percent of internal auditors already use generative AI, with 65 percent planning to deepen their involvement. A May 2025 Wolters Kluwer survey put current adoption at 39 percent, with another 41 percent planning to adopt within twelve months. If both trends hold, that's potentially 80 percent penetration by next year.
But the incumbents aren't exactly asleep. AuditBoard rolled out advanced AI capabilities in March 2025. Workiva unveiled agentic AI for finance, GRC, and ESG in September. Diligent, Caseware, Hyperproof—all have added AI-driven features to their platforms. For Oxus, the challenge isn't convincing buyers that AI belongs in audit workflows. It's proving that a three-person startup can deliver better accuracy, integration, and speed than established vendors bolting AI onto platforms they've refined over decades.
Direct competition is already forming. Midship, a YC Summer 2024 company focused on end-to-end SOX testing automation, raised $4.15 million in February 2026. Denki, from YC's Fall 2025 batch, calls itself a "full-stack AI financial audit firm" and integrates with platforms like AuditBoard and Workiva rather than replacing them.
The Unanswered Questions
What Oxus hasn't disclosed yet tells its own story. No pricing. No customer logos. No detailed specifications on how it integrates with ERPs or existing GRC systems.
And perhaps most critically for prospective buyers: no public validation that Big Four or mid-tier audit firms will accept AI-generated workpapers as audit-ready documentation. That's not a minor hurdle. It's the entire game. A tool that shaves 70 percent off audit prep time means nothing if external auditors refuse to rely on its output.
The company is actively hunting for introductions to chief audit executives, CFOs, and heads of internal audit or SOX at pre-IPO and public companies. Interested parties can reach the team at [email protected]—a detail that suggests the customer pipeline is still in its early innings.
The Promise and the Proof
For audit leaders exhausted by compliance season's ritualistic drudgery, Oxus offers something genuinely appealing: the same rigor, delivered in a fraction of the time. The same argument, incidentally, that every audit software vendor has made since computers entered the profession.
What separates credible innovation from vaporware usually comes down to three things: execution, integration, and acceptance from the gatekeepers who ultimately sign off. Oxus has the pedigree, the timing, and a clear problem to solve. Whether it can deliver the solution—one that external auditors, regulators, and risk-averse CFOs will actually trust—is a question the company will need to answer in quarters, not pitch decks.
For now, the three MIT grads are betting they can thread that needle. In an industry built on verification and documentation, we'll know soon enough whether the evidence supports their claim.
