One governed AI assistant per department

Cited answers, drafts, and case triage. Grounded in each department's own guidance.

For governance

One policy.
Every workspace.

Guardrails, routing, and audit as code. Applied the same way everywhere.

Open the dashboard Read the policy

Agent steps, max: 5; Agent steps, max
Models routed: 6; Models routed
Sandbox deadline: 1s; Sandbox deadline

policy.json

{

"version": "1.0.0",

"guardrails": {

"piiEscalation": { "triggers": ["nric", "uen"] },

"evalGate": { "threshold": 80 },

"costCeiling": { "usdPerCall": 0.05 },

"grounding": { "rule": "cite sources" }

"routing": { "fallbackModelId": "..." }

}

Oversight

Audit trail →

09:41callClaude Haiku 4.5 · $0.0025

09:41evalkeyword grader · 5/5

09:42callClaude Opus 4.8 · $0.0712over ceiling

09:43promptassistant-system v4 active

09:44callGPT-4o mini · $0.0003

09:44callClaude Sonnet 4.6 · $0.0121fallback

09:45evalLLM judge · 92%

09:46callGPT-4.1 nano · $0.0001

09:41callClaude Haiku 4.5 · $0.0025

09:41evalkeyword grader · 5/5

09:42callClaude Opus 4.8 · $0.0712over ceiling

09:43promptassistant-system v4 active

09:44callGPT-4o mini · $0.0003

09:44callClaude Sonnet 4.6 · $0.0121fallback

09:45evalLLM judge · 92%

09:46callGPT-4.1 nano · $0.0001

Platform · AI Audit Trail

Every call, run, and change, on the record.

Workspace · AI Instructions

Versioned instructions, one active at a time.

Platform · Workspaces

One workspace per department. One policy for all.

Choose how to get started

Work a case

Ask, draft, and triage in one workspace.

Answers cited to your guidance
Replies drafted for review
Case triage with PII flags
No-code tools

Open the assistant

Govern the platform

Set the rules every workspace runs under.

Policy as code
Deterministic model routing
Eval pass-rate gate
Full audit trail

Open the dashboard