Maze uses AI agents to investigate and resolve cloud vulnerabilities by analyzing them within the context of the cloud environment, compensating controls, and business requirements. The platform replicates workflows of security engineers to determine exploitability of vulnerabilities. The system connects to cloud environments through read-only roles and integrates with vulnerability scanners via API. AI agents investigate each vulnerability by checking exploitation requirements, presence of weaponized exploits, network configurations, and business context factors such as system confidentiality ratings and IAM permissions. Maze filters out non-exploitable vulnerabilities, which the platform states represent over 90% of findings, allowing security teams to focus on exploitable issues. The investigation process automatically generates audit reports for non-exploitable vulnerabilities. For exploitable vulnerabilities, the platform analyzes technical and business context to prioritize based on likelihood and impact. It generates mitigation and remediation plans with one-click actions routed to appropriate personnel through automated workflows. The platform operates in an agentless manner without requiring software installation on target machines. It uses runtime visibility techniques to gather necessary information. Deployment requires under five minutes according to the vendor. Maze is cloud-hosted on AWS with options for multi-tenant or single-tenant hosting. The platform uses LLM providers that guarantee data is not shared for training purposes, such as AWS Bedrock and Google Vertex. The company has ISO27001 accreditation and is undergoing SOC2 Type 2 observation.