From zero-credentials demo to live adapters

A new platform team had a mandate to bring order to an estate that had grown faster than the processes around it. They had a budget, a deadline, and, crucially, no appetite for a six-month SaaS rollout. What they needed was a way to evaluate honestly before committing: does this actually hold up with our workloads, our approval culture, our audit expectations?

The problem: evaluating a tool without connecting anything

Most platform tooling demands real credentials in week one, which means a security review, an account, and a decision before you have a feel for the product. The team wanted the opposite: touch the workflows first with no credentials, understand the model, and only then connect a real environment.

The fastest way to evaluate a control plane is to use it with demo data until it gets boring; then connect one account and see if the honesty survives contact.

The approach: demo first, guided live second

The starter path uses the demo workspace (mock infrastructure data, no cloud credentials, no connected accounts). Teams walk the inventory, policy, approval, and audit flows end to end. When they are ready, guided access connects a real account through the SDK adapters, and the same workflows run against live data.

step 1: demo workspace (mock data, zero credentials)      └─ evaluate inventory, policy, approvals, auditstep 2: guided live (one real account, scoped)      └─ same workflows, real data, labeled demo/livestep 3: scale (more accounts, on-prem agents, waves)      └─ policy, approvals, audit as the operating model

Implementation steps

• Walked the demo inventory and mapped it to their real estate's shape.
• Defined a candidate policy in demo: what should pass, deny, and require review.
• Requested guided access and connected one read-scoped account.
• Compared demo behavior with live data and adjusted the policy model.
• Then, and only then, discussed scaling to the full estate with on-prem and waves.

Guardrails from the start

Because the demo carries the same policy, approval, and audit model as production, the guardrails are learned before any real system is connected. The AI assistant in demo proposes next steps; in guided live it remains investigation-only. Nothing about the boundary changes when data becomes real.

Lessons learned

Evaluate in demo until it is boring. The demo exists so teams can learn the model without risk; if a team skips straight to connected accounts, they trade safety for speed and learn the guardrails under fire.

Second, connect the smallest real thing first. One read-scoped account tells you more about honesty than a hundred demo screens. When demo behavior matches live behavior, trust the pattern.

Third, treat demo/live labeling as a feature, not a caveat. Knowing exactly what is mock and what is real is the foundation of every later audit conversation.