Governed by design, not after deployment
AIF controls agents, tools, evidence, cases and deployment boundaries so regulated teams can introduce AI without losing oversight, provenance or reviewer accountability.
Trust model
Trust controls across the operating model
Public trust claims must stay precise: architecture capabilities are product design statements, not certification claims unless independently verified.
Agent control
AIF limits agent behavior through runtime governance instead of treating prompts as policy.
- Capability routing
- Tool policy checks
- Plan and execute modes
- Human-in-the-loop review
Evidence model
Outputs stay attached to the case and can be reviewed as operational artifacts.
- Toolcall replay
- Evidence artifacts
- Report persistence
- Audit-oriented exports
Access boundaries
Need-to-know is modeled across cases, teams, roles, tags and Solution data boundaries.
- RBAC
- Viewer role
- Taxonomy visibility
- Tenant-aware Solution data
Deployment fit
AIF is designed for regulated environments with different sovereignty and connectivity limits.
- SaaS
- Government cloud
- On-prem
- Airgapped and edge patterns
Claim discipline
The public site avoids unsupported certification claims and separates product design intent from verified compliance status.
- Designed for
- Supports
- Needs verification
- No invented logos
Product view
Agents execute through governed capability paths
Tool access, approval requirements and logging happen at runtime. That is the difference between a prompt policy and an operational control plane.
Policies
Capability and role checks before execution
Approvals
Human-in-the-loop where required
Logs
Tool activity preserved for review
CAPABILITY ROUTER
Requests pass through policy before tools execute
Product view
Evidence stays attached to the workflow
Files, tool calls, generated reports, workwall changes and review decisions remain part of the case context.
Evidence
Artifacts and generated reports
Replay
Tool activity and execution history
Review
Human approval and reviewer accountability
CASE TO EVIDENCE FLOW
Every output remains attached to the case
Product view
Deployment controls are part of the trust story
SaaS, government cloud, on-prem, airgapped and edge patterns have different verification requirements. The website states deployment intent and routes specifics into assurance review.
Sovereignty
Data boundary and hosting model matter
Verification
Capability parity checked per deployment
Discipline
No unsupported compliance claims
DEPLOYMENT TOPOLOGY
Same operating model, different sovereignty boundaries
