ARCHER: AI Agents Turn Building Codes Into Transparent, Scalable Compliance Checks

Checking whether a building design meets regulatory standards has long been a bottleneck in construction and engineering: auditors must manually validate thousands of rules against detailed architectural models, a tedious and expensive process that doesn't scale. Worse, most automated compliance checkers are proprietary black boxes—users can't see or verify that the rules are being applied correctly.

Researchers have now introduced ARCHER (Agentic Rule and Compliance Harness for Executable Regulations), a system that uses multiple AI agents working in a coordinated sequence to transform written building codes into executable, auditable verification code. The approach treats compliance checking as a code-generation problem, letting it adapt to new regulations and use cases instead of being locked into a single ruleset.

How It Works#

Rather than a single AI pass at the problem, ARCHER deploys a deterministically orchestrated multi-agent workflow. The system evaluates six progressively more sophisticated agent configurations across four different AI models—ranging from frontier APIs like GPT-4 to open-source models that can run entirely on-premises. By combining test-driven development (checking whether generated code actually works) with structured multi-step reasoning, the harness ensures generated verification code is both correct and transparent.

Results#

ARCHER's orchestrated multi-agent approach significantly outperforms naive single-pass prompting. The system achieved the highest accuracy across all backbone models tested, improving mean union accuracy by 82% over simpler baselines. Crucially, the research shows that a self-hosted open-weights model using the ARCHER harness can reach 97.8% of the accuracy of frontier APIs—while costing a quarter as much. This makes data-sovereign compliance checking practical: organizations can run building code verification entirely on their own infrastructure without relying on third-party APIs.

Why It Matters#

Building compliance is a high-stakes, high-volume challenge facing architects, engineers, and real estate developers globally. Manual checking is expensive and prone to error; proprietary tools lack transparency. By generating auditable code directly from regulatory documents, ARCHER creates a pathway toward compliance systems that are simultaneously more accurate, more adaptable, more affordable, and fully transparent to end users—letting stakeholders verify that regulatory intent has been correctly captured in the verification logic.