How KloudSense Determines Findings & Savings
Factual mathematics, 30–90 day empirical telemetry, and certified cloud architect review. We never use AI to blindly guess or hallucinate savings numbers.
The Factual Core: Deterministic Math vs. AI Interpretation
Why our numbers can be defended in front of your CFO and Engineering VP.
Deterministic Calculations (The Numbers)
All cost findings and dollar savings estimates are computed through rigid, deterministic mathematical formulas based directly on AWS public rate cards, provisioned capacity delta, and trailing 30-to-90-day CloudWatch percentiles (p95/p99 CPU, memory, IOPS, and network throughput).
- ✓ Zero mathematical hallucinations or speculative multipliers
- ✓ Empirical volume status: zero IOPS for 45+ days
- ✓ Exact byte-level VPC Flow Log accounting for NAT routes
AI-Assisted Interpretation (The Context)
Large language models are applied strictly for contextual reasoning, natural language summarization, and blueprint synthesis. AI assists with translating raw configuration diffs into clear engineering rationales and structuring safe step-by-step rollback procedures.
- ✓ Explains why an architectural pattern is wasteful
- ✓ Drafts safe CLI rollback commands and safety checklists
- ✓ Every output is gated and reviewed by a human architect
The 10-Step Determination Pipeline
AWS APIs
Data IngestionTargeted, read-only queries dispatched across authorized AWS regions using temporary STS credentials.
Cost & Usage Data
Data IngestionIngestion of AWS Cost and Usage Reports (CUR), Cost Explorer line items, and invoice breakdown by service and tag.
Infrastructure Inventory
Data IngestionComplete catalog of running, stopped, and unattached resources across compute, storage, databases, and networking.
Telemetry & Metrics
Empirical TelemetryHistorical 30-to-90-day CloudWatch telemetry metrics (p95/p99 CPU, memory usage, network I/O, disk throughput, active connections).
Deterministic Analysis
Deterministic EngineRule-based algorithmic heuristics evaluate provisioned capacity versus sustained utilization thresholds.
Savings Calculations
Deterministic EngineFactual dollar pricing models applying current AWS public rate cards, commitment terms, and regional rates to empirical waste.
Risk & Confidence Assessment
Safety FrameworkEvaluation of workload criticality, redundancy, failover requirements, and data stability to assign Low/Medium risk and High/Medium confidence.
AI-Assisted Interpretation
Contextual ReasoningContextual reasoning models synthesize narrative technical explanations, correlate dependencies, and structure safe rollback checklists. AI does NOT calculate savings numbers.
Human Architect Review
Quality & Safety GateA certified senior AWS cloud architect validates every flagged item against engineering reality to eliminate false positives.
Prioritized Action Plan
Final DeliverableFinal synthesis into the central AWS Savings & Findings Report, Executive Financial Summary, and step-by-step engineering blueprints.
Our Risk & Confidence Framework
Cost reduction is only valuable if production systems remain 100% reliable. Every recommendation receives explicit risk and confidence ratings.
Confidence Rating Criteria
Unambiguous empirical evidence across 30+ consecutive days. Examples: unattached EBS volumes with zero IOPS, terminated dev instances, log groups with zero queries, or unmanaged standard S3 objects >90 days old.
Mathematically valid, but requires internal engineering alignment (e.g. database rightsizing requiring maintenance failovers, or 1-year Savings Plans dependent on annual architecture roadmaps).
Risk Assessment Criteria
Changes that carry zero risk of customer disruption. Includes non-production environment scheduling, S3 lifecycle transitions (millisecond latency preserved), and deleting snapshot-backed orphaned volumes.
Changes requiring an approved maintenance window, database reboot, or rolling instance refresh. Always paired with pre-execution safety checklists and rollback procedures.
Certified Cloud Architect Review
Automated tools alert on everything, creating noise. Before your report is generated, a senior AWS-certified solutions architect reviews every single finding to eliminate false positives and ensure recommendations fit real-world production engineering.