Before a wider rollout
Establish access and ownership boundaries before a pilot becomes normal business infrastructure.
Fixed-scope baseline
AI agent security assessment for permissions, tools, and human controls.
Build a defensible picture of which agents exist, who owns them, what they can reach, how credentials are exposed, where people approve actions, and which logging gaps make review difficult.
Inventory
Permissions
Credentials
Approvals
Best Fit
A practical first step when agent use is growing faster than oversight.
For Portland and Maine organizations that need a fixed-scope answer to “what agents do we have, what can they do, and what should we fix first?”
After informal adoption
Identify shadow AI agent use and bring legitimate workflows into an accountable operating model.
When access is unclear
Translate tools, service accounts, local files, cloud applications, and agent permissions into a reviewable map.
Risks Addressed
Find control gaps before they become operating assumptions.
Unknown ownership
Agents or sub-agents without a business owner, technical owner, approved purpose, or retirement decision.
Broad tool access
Read, write, shell, browser, API, or application access that exceeds the bounded business task.
Shared credentials
Long-lived keys, inherited user sessions, broadly scoped tokens, or unclear credential rotation responsibility.
Missing approval gates
Sensitive sends, changes, transactions, deletions, deployments, or escalations that can proceed without human review.
Thin logging
Insufficient records to connect a request, tool call, command, approval, error, and resulting business action.
No shutdown owner
Unclear authority or procedure for pausing an agent, revoking credentials, and preserving evidence.
AI Agent Security Baseline
- Agent and sub-agent inventory with business and technical owners
- Tool, permission, application, data, and network-destination map
- Credential exposure and service-account review
- Human approval and exception-path review
- Logging, retention, and reconstruction-gap analysis
- Risk-ranked remediation register with suggested owners
- Executive readout with coverage and confidence limitations
Engagement
How the assessment works.
Scope and evidence plan
Confirm named agents, endpoints, owners, supported evidence, sensitive-data restrictions, and the questions leadership needs answered.
Inventory and interviews
Reconcile leadership, IT, and user perspectives with read-only discovery or configuration evidence where supported and approved.
Control mapping
Compare actual access with the approved task, document approval points, and identify where credentials or logs weaken accountability.
Remediation workshop
Prioritize changes by risk, feasibility, owner, dependency, and the degree of human approval required.
Control Principles
Least privilege is a design decision, not a checkbox.
We separate permissions by task, prefer scoped and revocable credentials, distinguish observation from action, and require people to approve consequential steps. Recommendations state where evidence is incomplete rather than assuming coverage.
Common Questions
Assessment FAQ
What do you need from us for an assessment?
A named sponsor, a list of in-scope agents and owners, access to relevant configuration or demonstrations, and staff who can explain the approved use case. Evidence access is agreed before review begins.
Will the assessment change our systems?
The baseline is designed around interviews, configuration review, and read-only discovery where supported. Configuration changes, hooks, enforcement, or active testing require separate approval.
What do we receive?
You receive an agent inventory, ownership map, tool and permission matrix, credential and approval observations, logging-gap analysis, and a prioritized remediation plan with stated limitations.
Turn an unclear agent environment into an owned remediation plan.
Start with a bounded set of agents, evidence sources, and decisions. We will confirm fit and scope before reviewing sensitive material.