Trust

Built for review, designed for governed action

CloudKnife helps teams improve cloud efficiency without giving up control. It starts with visibility and review, and can move toward policy-governed automation where the environment and rules support it.

How we earn trust

Concrete behaviour, not slogans

Review first, explain always, narrow automation to explicit policy. The first pillar is the on-ramp everything else hangs on.

Read-only onboarding first

Connections start read-only. On Azure today we use roles such as Monitor Reader and Cost Reader to analyse usage, cost, metrics, tags, and resource metadata without changing your infrastructure. Other hyperscalers follow as we match the same quality standard for review, and we welcome security-minded early adopters on AWS or GCP.

  • Explainable recommendations

    Each item carries rationale, expected impact, affected resources, and confidence so reviewers can challenge assumptions instead of trusting a black box.

  • Production-aware logic

    Recommendations respect environment context, including different headroom and risk expectations for production and non-production workloads.

  • Review trail and accountability

    History and ownership stay attached to recommendations so teams can see what was surfaced, who should respond, and how review changed over time.

  • Hosted in West Europe

    CloudKnife runs in West Europe, aligned with an EU-focused way of working. We do not list formal certifications we have not completed.

Review workspace

What a governed recommendation looks like

The product keeps evidence, policy, and review state on one row so nothing important hides in a second tool.

Production contextTier 1 · customer-facing

Change window Sat 02:00 to 06:00 UTC. Owner approval required before any execution path, including where governed automation is enabled.

13e55095…98409b7
Review state: Awaiting ownerPriority: HighSafety: Very safeConfidence: 98%
Open in review workspace
Confidence
98%
Expected impact (yearly, illustrative)
€2,444.92

Based on observed usage and stated assumptions. Your numbers and currency follow your tenant.

Rationale
Utilisation signals
Low steady-state CPU with stable tail windows across the decision period.
Behaviour
No burst pattern that would justify the current headroom on this SKU.
Policy alignment
Fits tier rules and the published change window for this environment.
Affected resources
Subscription: prod-coreRG: app-prod-weprod-web-vm-03prod-db-01

Ownership and blast radius stay visible so reviewers know who signs off and what moves together.

Production and policy note

This recommendation uses production headroom rules, the active policy pack, and the service owner list for the subscription. Nothing runs automatically from this view.

Current vs recommended
Current
Standard_D4s_v3
€3,156 / yr
4 vCPU · 16 GB RAM
Recommended
Standard_B2as_v2
€711 / yr
2 vCPU · 8 GB RAM
Review checklist
  • Evidence and assumptions
  • Impact and affected resources
  • Production context and policy pack
  • Owner and accountability
Automation (policy-governed)
Disarmed · review-led

Where your organisation turns execution on, it stays inside explicit policy packs, approval rules, and change windows. Most teams stay review-first while those controls are rolled out. This item will not execute until gates clear.

Approval requiredPolicy pack: EU-prod-defaultAudit trail on decision
Review trail
Last structured check · evidence retained
Accountability
Owner group · platform@example.com
Posture
Read-only data path

Illustrative fields for marketing. Your tenant controls labels, policy packs, and whether any execution path is enabled.

Trust is earned in review before it is earned in automation. We say that plainly because it shapes what we build.

Current state

What CloudKnife does today

We keep this explicit so expectations match reality. The product earns trust in review before it earns trust in automation.

Today the focus is on insights, evidence, and reviewable recommendations. There is no universal unattended apply path across every customer. Where execution or deeper integrations appear, they are framed as governed, policy-based, and opt-in, not as silent change.

  • Insights across usage and spend with enough context to be actionable in review, not only charts.
  • Evidence surfaced next to recommendations so operators can verify claims without exporting raw dumps.
  • Reviewable recommendations as the default path: structured items your team accepts, adjusts, or rejects.
  • A trust-building start: read-only access first, narrow scope, and no promise of silent change across your cloud.
  • Automation described honestly: where it exists or is planned, it is policy-governed and approval-bound, not blind execution.
AI in the product

Guardrails that keep humans in charge

Assisted analysis stays bounded. Outputs are structured for review, not for silent execution.

  • Scoped guidance

    Suggestions stay inside clear operational boundaries so teams know what class of change is being proposed.

  • Human-led decisions

    Recommendations are explainable. Your people approve, defer, or reject. The product does not override change management.

  • Governance alignment

    Risk tolerance, ownership, and policy context stay visible next to each item so approvals map to how you already ship.

  • Traceability

    Rationale and evidence attach to recommendations so reviews and audits stay grounded, not pieced together from memory later.

Security approach

Least privilege, limited scope, you stay in control

Short and direct. We avoid certification language we cannot substantiate.

  • Least privilege

    Access is scoped to what analysis requires. We do not ask for broad admin rights to deliver recommendations.

  • Limited data scope

    We focus on operational and cost signals needed for efficiency work, not a full copy of all your data.

  • Teams stay in control

    Review and approval stay on your side. CloudKnife prepares and explains; it does not override your change management.

CloudKnife is built so teams can review clearly, automate deliberately, and stay accountable.

Questions about access, data handling, or how review maps to automation in your environment? We answer with specifics, not generic safety language.