SAN FRANCISCO: Andera, a startup building an AI platform that automates internal audit work, raised $37 million in a Series A round led by Lightspeed Venture Partners, the company announced on June 17, 2026. The round funds a push to land more Fortune 100 customers and expand a system that takes over the spreadsheet-and-email grind that has defined corporate audit since the early 2000s.
The company, founded by Aryo Patel and Tinah Hong, sells software that handles evidence collection, control testing, and workpaper generation for Sarbanes-Oxley (SOX) and other compliance regimes. According to Andera’s website, the platform is built to “save 70% of SOX control testing costs with AI,” and Lightspeed partners Kat Zhang and Bucky Moore wrote in their investment thesis that Andera’s verification engines can reduce SOX testing time by up to 75 percent.
Why Audit Software Suddenly Became Investable
Internal audit has been one of the slowest enterprise functions to modernize. Lightspeed’s Zhang and Moore note that the workflow “hasn’t been reimagined in two decades,” dating back to the passage of Sarbanes-Oxley in 2002. The economics are heavy: average annual audit fees at public companies run about $3 million, and Fortune 100 firms spend more than $20 million each year, according to data cited in the Lightspeed post.
Two forces changed the picture at the same time. Large language models with long context windows and agentic tool use can now read multi-format evidence and produce defensible documentation, work that rules-based automation could never handle. And CFOs have started demanding what Lightspeed describes as “200-300% efficiency gains” from back-office functions, putting compliance headcount on the chopping block. Andera is positioning itself for the cost cut.
The Big Four audit firms face what Lightspeed calls “a genuine innovator’s dilemma”: their billable-hour model makes it costly for them to build the same AI tools that would erode their own revenue. That gap is where startups like Andera plan to sell, both to in-house corporate audit teams and to public accounting firms looking for a competitive edge.
For internal audit leaders, SOX testing has long been the single largest line item in the function’s budget and the work that consumes the most junior staff time during busy season. The Lightspeed thesis argues that automating control testing frees those professionals to focus on enterprise risk identification and strategic advisory work, the parts of the job most auditors say drew them in. The pitch is not replacement, but reallocation of where the most expensive hours get spent.
How will Andera’s AI change internal audit teams at large companies?
Andera replaces the manual portion of control testing, including pulling user access lists, matching evidence to controls, generating workpapers in the same Excel formats auditors already use, and chasing control owners for missing documentation. Fortune 100 audit teams that currently spend thousands of hours per quarter on these tasks would reallocate that time to higher-judgment risk work, per the company’s pitch.
The technical bet, Patel said in the Lightspeed post, is that “the models are already sufficiently smart. The hard problem is getting from billions of tokens to the 30,000 that matter.” Andera’s architecture uses a proprietary agentic retrieval system rather than standard vector search, combined with probabilistic beam search to focus compute on the uncertain portions of a data graph. The platform processes “hundreds of millions of tokens of financial evidence per control,” according to Lightspeed.
The team has paired that engineering bench with audit veterans. Carina Averilla, Andera’s first hire and a former Deloitte auditor, previously led US accounting at Revolut. Jared Lauber, who runs partnerships, spent three decades in audit and risk at firms including EY, Instacart, McKesson, and Williams-Sonoma.
Patel previously worked on Azure infrastructure scaling at Microsoft and on the trading desk at Jane Street. Hong, who has known Patel since middle school in Chicago and overlapped with him at MIT, built machine learning systems at Stripe before launching Andera. Andera is SOC 2 Type II certified and states on its website that no customer data is used to train models or sent to third-party providers, a positioning aimed directly at Fortune 100 information-security review.
What to Watch Next
Andera is competing in a category that includes Workiva, AuditBoard, and DataSnipper, but the Series A signals investor conviction that an AI-native entrant can take meaningful share rather than being absorbed by an incumbent. The Big Four’s response will matter most: whether Deloitte, EY, KPMG, and PwC partner with startups like Andera or push their own internal AI tools will shape how fast audit budgets shift.
For founders watching the back-office automation wave, the next signal will be customer disclosures. Andera says it already works with Fortune 100 customers; named logos and renewal data over the next two quarters will indicate whether the 70 to 75 percent cost-reduction pitch holds up at scale. The company is hiring engineers and customer-success staff, per its careers page, and Lightspeed’s Zhang and Moore said in the funding announcement that the firm spent “several months building a thesis” before writing the check, suggesting more capital is likely to flow into adjacent compliance categories.
GREY Journal has covered related shifts in AI automation of engineering work, the funding flowing to enterprise robotics and automation, and the broader question of what happens when AI eats white-collar functions.
