Luca BertonThe first mistake teams make with C5 and AI agents is treating scope as a cloud-provider question. The provider's controls matter, but your agent platform adds its own control surface: identities, prompts, model gateways, vector stores, tools, audit logs, and operational runbooks.
If these components sit outside the C5 control conversation, the resulting assurance story will be incomplete.
Start with the Service Boundary
Define what the AI agent platform actually provides. Is it a chatbot for employees? A regulated customer-support assistant? A tool-using agent that can open tickets, read documents, and call external APIs? The answer changes the scope.
- User entry points: web app, API, Slack, Teams, customer portal, internal developer platform
- Identity layer: SSO, MFA, token issuance, service identities, administrator roles
- Agent runtime: orchestration service, tool router, policy layer, approval workflows
- Model access: model gateway, provider APIs, prompt templates, safety filters
- Data plane: document stores, vector databases, embeddings, caches, logs
- Tool integrations: Gmail, GitHub, Jira, ServiceNow, CRM, internal APIs
The C5-Relevant Components Most Teams Miss
AI agent platforms often have hidden infrastructure that becomes critical during audit preparation:
- Prompt and trace storage: may contain sensitive business data or personal data
- Evaluation datasets: can include production examples copied into test suites
- Embedding pipelines: create derived data that needs lifecycle controls
- Tool credentials: need issuance, storage, rotation, and revocation evidence
- Human approval queues: become part of the control environment for high-risk actions
Kubernetes Recipes
Practical guide for container orchestration and deployment — hands-on patterns you can use today.
View on Amazon →Scope Decision Matrix
Include in Scope When
- Identity systems: they authenticate users or issue tokens used by agents
- Vector stores: they contain regulated, confidential, or customer-specific content
- Model gateways: they enforce policy, routing, logging, or provider controls
- Tool integrations: they can read, write, delete, or transmit business data
- Observability systems: they store prompts, traces, API payloads, or incident evidence
Document Responsibilities Explicitly
A useful C5 scope includes a responsibility matrix. For each control area, name the responsible party: cloud provider, identity provider, model provider, platform team, security team, data owner, or application owner. Ambiguous ownership is where audit findings are born.
Introduction to Chip Design with Open-Source EDA Tools
Design and simulate digital circuits using open-source tools. In collaboration with Starweaver.
Start on Coursera →Bottom Line
C5 scope for AI agents should follow the real control boundary, not the invoice boundary. Include every component that authenticates users, authorizes actions, stores sensitive context, routes model calls, executes tools, or produces audit evidence.
Related Solution
Navigating AI adoption in a regulated environment? Our readiness assessment maps infrastructure, governance, and compliance gaps in 3-4 weeks.
Explore AI Readiness for Regulated Enterprises →