Dili raises $21.7M for AI compliance in infrastructure boom

Dili, an AI compliance startup founded by CEO Anand Chaturvedi, announced on Thursday that it had raised $15 million（約24億円） in Series A funding to help infrastructure projects navigate federal compliance requirements. The round was led by Khosla Ventures, with participation from Allianz, Rebel Fund, Brick and Mortar Ventures' Darren Bechtel, and Y Combinator's Garry Tan. Combined with a $6.7 million（約11億円） seed round, the company has now raised $21.7 million（約35億円）. Dili was previously part of Y Combinator's Summer 2023 batch.

The company's core focus is a distinct market niche: U.S. infrastructure projects, particularly those receiving federal funding, which must comply with a complex and overlapping set of regulations. Chaturvedi highlighted Davis-Bacon rules, which allow the Department of Labor to set prevailing wages for certain projects, alongside prevailing wage and apprenticeship (PWA) rules that apply to clean energy projects funded under the Inflation Reduction Act, plus various OSHA and EPA regulations depending on the work's nature. The stakes are substantial: non-compliance can result in millions of dollars of fines. Yet auditing compliance is labor-intensive; traditionally, companies sample data rather than checking all information as it arrives, risking undetected violations.

Dili's solution addresses this gap through a hybrid AI-and-deterministic system. The company uses contemporary AI models only in its data layer to extract and structure information from unstructured documents — company internal records, vendor documents, ERP systems, and payroll data. Once data is structured, a deterministic rule-based system applies the complex-but-static compliance rules, eliminating the fuzziness that contemporary AI models can introduce. The result: a task that once required a full day's work can now be completed in minutes. "Imagine being able to read across the entire context of a company's internal documents, all of their vendors' documents, all of their ERP information, all of their payroll systems information, and then draw out the data that you need specifically for you know reporting or compliance," Chaturvedi explained.

Dili is already deployed at about 700 projects, ranging from manufacturing facilities to data centers. Notably, the company operates under two business models: roughly half its customers use Dili as in-house software, while the other half outsource the entire compliance process to Dili on a contractor basis. Chaturvedi anticipates the industry will shift more toward the software model in the coming years. "Software and AI are going to start eating a lot of those professional services workflows, so I think more and more people will start to bring those in-house," he said, adding that the evolving market will ultimately reflect where customer needs and AI capabilities converge.

The infrastructure boom driven by federal funding — including clean energy projects under the Inflation Reduction Act and data centers — has created a new category of regulatory risk. Projects must navigate overlapping federal, state, and local compliance regimes (Davis-Bacon prevailing wages, IRA-specific rules, OSHA, EPA), each with steep penalties for non-compliance that can reach millions of dollars. Dili's founding and rapid adoption reflects this gap: traditional compliance work is labor-intensive, error-prone, and expensive, yet the stakes are too high for sampling-based audits.

The company's architecture addresses a real tension in AI-for-enterprise applications. Rather than relying on large language models (LLM) to make final compliance decisions — where they can produce hallucinations or inconsistent reasoning — Dili uses AI only in the data extraction layer to convert unstructured documents into structured input, then applies deterministic rule-based systems to enforce the complex but static regulatory requirements. This hybrid approach allows a task that once consumed a full day of human effort to be completed in minutes, reducing both cost and human error while maintaining the reliability demanded by high-penalty environments. CEO Anand Chaturvedi's observation that the market will shift toward in-house software deployment aligns with broader trends in professional services automation, though the current 50/50 split between software and contractor models suggests the transition is still early.