Pilot Access
Govern what your AI agents do — before they do it
A time-boxed pilot of Crelis runtime authorization for regulated financial-services teams in Singapore and APAC whose AI agents take real actions. Other regulated industries are welcome. Shadow mode makes no changes to production systems.
How it works
Start safely in shadow mode
Decision metadata only
Crelis evaluates decision metadata — action types and results — so raw customer data does not need to enter the request.
Shadow mode
Read-only and out-of-band during the pilot. It observes and records; it enforces nothing and changes nothing in your systems.
Our hosted service
Your tenant runs on our hosted service; deployment options beyond it are on the roadmap. Start on synthetic traffic, graduate to a UAT mirror when you’re ready.
What you get
- · A findings report: the consequential actions in your workflow that your current controls would have missed
- · Direct input into the product roadmap
- · A reusable policy set for your workflow, authored with you
- · Named or anonymous reference — entirely your choice
What we ask
- · One bounded workflow your AI agents handle, which can contain several execution tasks (e.g. refunds, data export, account changes)
- · One technical point of contact for a non-prod setup
- · Two to three short sessions across the 4–6 weeks
- · Honest feedback
Timeline
Four to six weeks, end to end
Week 0
Setup
NDA, workflow scoping, your tenant provisioned on our hosted service.
Week 1–2
Shadow
Crelis evaluates traffic; we author your policies together.
Week 3–4
Findings
What was caught, explainability, audit integrity, readout.
Optional
Enforce
Decide whether to move to a binding integration.
One conversation to start.
Bring one agent workflow to a 30-minute working session and we’ll show you the decision, the explanation, and the audit trail — live.
We also work with technology, consulting and implementation partners who bring runtime authorization to their clients — write to sales@crelis.ai.