Corporate auditing is one of those business rituals few talk about and everyone pays for, year after year. Public companies spend thousands of hours gathering evidence, testing controls, and writing up reports, mostly by hand in Excel, in processes that haven’t changed much since 2002. Andera, a startup aiming to hand this work to artificial intelligence, has just raised $37 million, signaling the sector is ready for a shake-up.

From spreadsheet to LLM

Andera’s idea isn’t unique. In recent years, large language models and document automation pipelines have collided with professions based on manual verification. Audit is an ideal candidate: huge volumes of structured and unstructured data, repetitive rules, and the need to spot anomalies and patterns. An LLM trained or fine-tuned on accounting standards and auditing practices can rapidly analyze thousands of transactions, cross-reference evidence, and flag discrepancies that a human reviewer would take hours to uncover.

But the technology isn’t ready for turnkey adoption. Audit work demands absolute precision and full traceability: a model that hallucinates or gives unverifiable answers is unacceptable. Moreover, Europe’s regulatory environment – with GDPR and financial data protection rules – imposes strict limits on where and how data can be processed.

The data sovereignty tangle

For the Big Four and major consulting firms, using AI for auditing runs headlong into the need to retain total control over client information. Balance sheets, ledger entries, and internal documents cannot travel over public clouds without ironclad contractual guarantees. That’s why, beyond the allure of software as a service, many organizations are eyeing on-premise or hybrid deployments, where inference and data processing happen on proprietary infrastructure, inside their own data centers or on dedicated hardware.

This isn’t a minor technical detail: we’re talking about TCO, hardware choices (GPUs with enough VRAM for model quantization), continuous updates to orchestration frameworks, and latency management. Against this backdrop, Andera’s funding signals market momentum, but the path to fully automated auditing is lined with complex architectural decisions.

Beyond the cloud: why audit may push AI on-prem

By its nature, audit requires trust and confidentiality. A model running self-hosted, perhaps in an air-gapped environment, offers reassurance that public cloud struggles to match. It’s not just about compliance – it’s a matter of legal accountability. If an AI assistant misses a fraud indicator, who answers? The cloud provider or the audit firm? The trend toward keeping full control of the inference pipeline is bound to grow, along with interest in solutions that allow LLMs to run close to the data, without a single token leaving the internal network.

In this picture, the Andera funding round carries meaning beyond the dollar figure. It shows the industry is trying to bridge the gap between the power of generative AI and the real-world needs of regulated sectors. For those evaluating on-premise deployment today, there are significant trade-offs among capital costs, in-house skills, and scalability. Yet auditing itself could become one of the first use cases where the return on investment justifies the complexity.

Outlook

The course is set. Automating audit work won’t happen overnight, but serious funding is a clear sign that AI is about to leave the lab and enter the boardroom. For technology providers and audit firms, the game will be won by those who can balance model performance and data sovereignty – a balance that will likely bring on-premise hardware back to the forefront.