There's an old line in the audit world: SOX compliance is the work that never ends.
For companies marching toward an IPO—or those already trading—the annual Sarbanes-Oxley ritual unfolds like clockwork. Weeks disappear into scoping controls, mapping process flows, hunting down evidence, drafting workpapers. Tedious barely covers it. Expensive doesn't quite capture the scale. And by nearly every account, the whole apparatus feels long overdue for reinvention.
Which brings us to Oxus, a startup so new its founding team still numbers just three. Fresh out of Y Combinator's Winter 2026 cohort, the company has staked its claim on a specific, unglamorous corner of corporate life: automating the internal audit grind that bogs down finance teams at pre-IPO and public companies alike. The founders—Janet Liu, Melinda Liu, and Christine Watts—launched their platform in January 2026 with a pitch that sounds almost too clean: turn hours of SOX work into minutes using AI.
The bet itself is telling, even if specific performance metrics and external auditor validation remain limited in the public domain.
The Mechanics of What They're Building
Oxus zeroes in on four friction points in the SOX 404 compliance cycle, the part of the law that requires companies to document and test their internal controls over financial reporting.
First, it accelerates scoping—using AI to flag higher-risk areas based on financial and qualitative signals, the kind of judgment calls that usually involve spreadsheets and committee debates. Second, the platform generates process flowcharts and narrative walkthroughs from meeting notes, a task that can consume entire afternoons. Third, it automates the actual control testing and spits out reviewer-ready workpapers, complete with traceable evidence links. External auditors tend to have strong opinions about documentation trails; Oxus is designing to meet those expectations out of the gate.
Then there's what the company calls Document Intelligence—AI-driven search and annotation across fragmented files and systems, with optional web search layered in for additional context. It's the kind of feature that sounds incremental until you've spent a day hunting for a single control matrix buried in someone's SharePoint archive.
The economics, at least as Oxus frames them, are stark. Companies currently shell out between $200 and $800 an hour to outsource this work, with total annual costs landing anywhere from $300,000 to north of $1 million. The founders' pitch: bring it in-house and speed it up dramatically, though specific cost savings figures haven't been publicly detailed.
The platform promises audit-grade outputs—evidence grounding, human-in-the-loop approvals, standardized formats designed to satisfy both internal teams and the external auditors who ultimately sign off. On security, Oxus offers deployment flexibility: customers can use Oxus-hosted infrastructure or bring their own cloud. The company's privacy policy, dated November 2, 2025, states flatly that customer data won't train foundation models. Oxus says it's pursuing SOC 2 Type II and ISO 27001 certifications, with no public attestations available as of yet.
Three MIT Grads and a Hunch About Audit
All three founders hold MIT degrees spanning computer science, economics, and data science—a combination that perhaps telegraphs the team's view of audit work as an optimization problem waiting to be cracked.
Janet Liu, the CEO, spent time at the Federal Reserve researching digital payments before moving to investment strategy at D.E. Shaw. Melinda Liu, the COO, built AI products in fintech and worked in finance at JPMorgan. Christine Watts, the CTO, focused on machine learning systems at early and growth-stage startups, with stints at BCG and NERA along the way.
Their target buyers: chief audit executives, CFOs, heads of internal audit and SOX, controllers—anyone staring down a first audit or managing recurring compliance cycles. Oxus's Y Combinator profile mentions engagements with "leading payroll providers and public biotech companies," though no customer logos or detailed case studies have surfaced publicly.
A Market Already Getting Crowded
Timing-wise, Oxus is entering just as the incumbents hit the accelerator.
Around the same time Oxus emerged, Diligent—a dominant player in governance and audit software—announced "AuditAI" on March 10 at the Institute of Internal Auditors' annual conference. The pitch: agentic automation with auditor oversight and full traceability. A few months earlier, in November 2025, Workiva unveiled agentic AI across its finance and governance suite, including tools for continuous controls monitoring. Workiva claims more than 6,400 organizations as customers, a scale that's hard to ignore.
AuditBoard has been layering AI-enabled analytics and automation into its platform throughout 2024 and 2025. The company has also published research arguing that internal audit teams must embrace AI to navigate what it calls an era of "hypervolatility"—corporate-speak for the mounting compliance and risk pressures buffeting finance departments.
Then there are the legacy players retrofitting their software. TeamMate+, from Wolters Kluwer, launched an "AI Editor" for audit documentation last year. MindBridge has been weaving LLM-assisted features into its platform and recently partnered with Genpact on AI-led internal audits. Even the Big Four-adjacent firms are mobilizing: RSM announced a $1 billion, three-year investment in AI strategy last June.
Oxus also faces competition from other startups chasing the same idea. Midship raised $4.15 million in seed funding in February to automate SOX testing with AI agents. Auditstage, based in Europe, secured a €750,000 pre-seed round last year with a similar vision. What was once a sleepy corner of enterprise software is suddenly very awake.
What We Don't Yet Know
Oxus's reported funding of approximately $928,000 comes from third-party aggregators and remains unverified by primary documentation. No pricing details are public. No customer testimonials or named references have been shared. The company's website and launch materials lean heavily on feature descriptions and efficiency promises—understandable at this stage, but prospective buyers will almost certainly want harder proof. Specifically: Do external auditors accept the AI-generated workpapers? How much time do customers actually save in real-world deployments?
The certifications Oxus is pursuing matter in this space, no question. But bridging the gap from a three-person team to enterprise-grade reliability is a long walk.
Whether Oxus can carve out sustainable ground against well-funded incumbents and venture-backed peers will hinge on execution speed, customer traction, and something harder to quantify: the willingness of audit committees to trust algorithms with work that's historically been done by humans with spreadsheets and a healthy dose of professional skepticism.
For now, Oxus represents one more wager that the next generation of audit work will be done by machines, not analysts hunched over control matrices. The question—perhaps the only one that matters—is whether controllers and chief audit executives will buy in before the window closes.
