Infrastructure Resource

Infrastructure resource knowledge lives entirely in the heads of the two people who set up the original servers. 'What's running on that box in us-east-1?' gets answered by SSHing in and poking around. There's no inventory of compute instances, no record of storage configurations, and no documentation of why resources were provisioned the way they are. When someone leaves, their infrastructure knowledge leaves with them.

None — AI cannot reason about infrastructure resources because no documentation or inventory exists in any system. Every question requires a human to manually inspect live environments.

Some infrastructure resources are documented but coverage is patchy and inconsistent. A wiki page lists the major production servers but omits staging environments, dev instances, and ancillary storage. Configuration details are scattered across Slack messages, email threads, and personal notes. 'We have a doc somewhere about the database cluster setup, but it's probably outdated by now.'

AI can read the partial documentation that exists but cannot build a complete picture of the infrastructure because coverage is incomplete and the documentation it can find is likely stale or contradictory.

Infrastructure resources are documented using a consistent template covering resource type, configuration, purpose, cost, and owner. The operations team updates documentation when provisioning or decommissioning resources. However, documentation is stored in wiki pages that aren't machine-readable — you can find the docs, but an AI agent can't programmatically query 'show me all resources costing more than $500/month with utilization below 20%.'

AI can search and read infrastructure documentation to answer questions about specific resources. Cannot perform cross-resource analysis or automated audits because the documentation format is human-readable prose rather than structured, queryable records.

Every infrastructure resource is formally defined in a configuration management database or infrastructure-as-code repository. Resource records are current, findable, and consistently structured. An engineer can query 'show me all EC2 instances in the payment processing service with their current configuration and monthly cost' and get an accurate, up-to-date answer within seconds.

AI can query the infrastructure inventory programmatically, correlate resource configurations with cost and utilization patterns, and generate optimization recommendations. Cannot yet autonomously provision or modify resources because changes require human approval workflows.

Infrastructure resources are modeled as a formal ontology with typed entities and validated relationships. Every resource has defined connections to the services it hosts, the network segments it belongs to, and the cost centers it bills to. An AI agent can answer 'what is the full dependency chain if we decommission this Kubernetes node, and what services would be impacted?' by traversing the formal resource graph.

AI can autonomously analyze infrastructure topology, generate capacity plans, and draft provisioning proposals that respect architectural constraints and dependency relationships. Human approval is still required for executing changes that affect production workloads.

Infrastructure resources are self-documenting. Discovery agents continuously scan the environment, automatically registering new resources, updating configurations, and retiring decommissioned assets. The infrastructure ontology evolves in real time as the environment changes. Documentation is never stale because it is generated directly from the live infrastructure state, including cost attribution, utilization baselines, and service dependency mappings.

Can autonomously maintain a complete, real-time model of all infrastructure resources, their relationships, costs, and utilization. AI manages the full lifecycle of infrastructure documentation without human intervention.

Ceiling of the CMC framework for this dimension.