Three Pillars
ANTS Platform organizes AI operations around three questions:
| Pillar | Question it answers |
|---|---|
| AI Cost | What does our AI cost, who and what drives it, and how do we spend less? |
| AI Resilient | Is our AI fast, available and working correctly, and how quickly do we notice when it is not? |
| AI Governance | Which AI is in use, is it within policy, and can we prove it? |
All three are built on the same foundation: traces from your applications, agents, coding tools and the Budget AI gateway, plus AI usage discovered on employee devices.
AI Cost
| Capability | Where | Docs |
|---|---|---|
| Cost by model, use case, user, environment and Budget AI key | Gauge AI > Cost Dashboard, Cost Performance Metrics, AI Usage | Cost & Latency Dashboards, AI Usage |
| Cost per end user | Gauge AI > Users | Users |
| Model prices used for cost calculation | Project Settings > Models | Models & Pricing |
| Cheaper or faster model recommendations | Gauge AI > AI Optimizer (needs the FinOps recommendations entitlement) | AI Optimizer |
| Budgets, spend limits and model routing for LLM traffic | Budget AI (needs the Budget AI entitlement and feature flag) | Budget AI |
| Per-agent cost | Discover AI > AI Agent Center (Enterprise) | AI Agent Center |
AI Resilient
| Capability | Where | Docs |
|---|---|---|
| End-to-end traces of every request, with errors and levels | Discover AI > Tracing | Traces & Observations |
| Latency by model, use case and user, time to first token | Gauge AI > Latency Dashboard, Latency & Analytics | Cost & Latency Dashboards |
| Agent health and latency | Discover AI > AI Agent Center (Enterprise) | AI Agent Center |
| Notifications and incidents in Slack, Jira, Linear, PagerDuty and ServiceNow | Integrations in Project Settings and Organization Settings | Integrations |
| Quality scores from evaluators and human review | Scores on traces | Scores |
AI Governance
| Capability | Where | Docs |
|---|---|---|
| Discovery of AI tools and agents used across the organization | Discover AI (Enterprise) | Shadow AI |
| Inventory of models and agents | Gauge AI > AI Inventory (Enterprise) | AI Inventory |
| PII, PCI and PHI detection and guardrails | Governance > AI Guardrails, AI Violations (Enterprise, enabled per organization) | AI Guardrails, Violations |
| Risk and drift per agent | Governance > Risk Dashboard (Enterprise) | Risk Dashboard |
| Compliance requests, evidence and framework coverage (EU AI Act, NIST AI RMF, ISO/IEC 42001) | Governance (Enterprise) | Compliance & Auditing |
| Roles, SSO, audit logs and data retention | Organization Settings | Administration |
ℹ
PII detection is not applied automatically to every trace. Guardrails are an Enterprise feature that an administrator turns on and configures.
How the pillars work together
- An agent fails more often after a model change. Tracing shows the failing tool calls (AI Resilient). The Risk Dashboard flags the drift (AI Governance). The cost dashboards show whether the new model also changed spend (AI Cost).
- Spend jumps in one team. Budget AI limits the team's keys (AI Cost). The traces show which agent and users drove the increase (AI Resilient), and guardrails confirm no sensitive data was sent to the new model (AI Governance).