All Resources
Doc5 min readApr 27, 2026

Sample Assessment Output

What you get back from the Cloudsaver AI + Cloud Savings Assessment. The examples below represent the types of analysis included in the written report.

Cloud Savings — Rate Optimization

How you're paying for cloud today, where commitment coverage gaps exist, and the recommended instrument mix to close them. Includes current vs. projected coverage, total identified savings, and a breakdown of existing commitments alongside new Cloudsaver-managed instruments.

Rate Optimization

Current Coverage

34%

of eligible spend

With Cloudsaver

91%

projected coverage

Total Savings

$4.3M

annualized

Savings Rate

18.4%

of addressable spend

Coverage Comparison

Current34%
With Cloudsaver91%

Recommended Instrument Mix

Existing 1-Year RIs

$1.2M

28% of coverage

Existing 3-Year RIs

$680K

16% of coverage

Cloudsaver 30-Day

$1.8M

42% of coverage

Cloudsaver 1-Year

$620K

14% of coverage

Cloud Savings — Usage Optimization

The resources themselves — what's over-provisioned, what's idle, what's running on previous-generation hardware. Each finding is categorized, quantified, and broken down to the resource level with projected annual savings.

Usage Optimization

Usage Optimization

287 resources identified across 3 categories

Total Identified

$585K/yr

Rightsizing

142 resources$284K/yr

Over-provisioned compute instances across 8 accounts

EC2 instances (87)$198K/yr
RDS instances (31)$52K/yr
ElastiCache nodes (24)$34K/yr

Idle & Orphaned

89 resources$167K/yr

Resources running with no meaningful utilization

Unattached EBS volumes (34)$18K/yr
Unused Elastic IPs (12)$4K/yr
Idle RDS instances (8)$86K/yr
Dormant Lambda functions (35)$59K/yr

Previous Generation

56 resources$93K/yr

Resources on older instance families with better options available

m4/m5 → m7 migration candidates (38)$67K/yr
gp2 → gp3 volume upgrades (18)$26K/yr

Tag Health

A composite score across three dimensions — coverage, compliance, and clarity — with specific remediation recommendations ranked by severity. This is the foundation that makes cost attribution, showback, and forecasting reliable.

Tag Health
58%

Overall Health

Coverage
72%target: 95%

72% of resources have at least one tag

Compliance
58%target: 90%

58% of tagged resources comply with tag policies

Clarity
44%target: 85%

44% consistency across tag keys and values

Remediations

5 findings
SeverityIssueResourcesRecommended Action
highUntagged resources1,847Apply auto-tagging rules for Owner, CostCenter, Environment
mediumInconsistent keys23Consolidate dept/department/Department into single key
highMissing cost allocation tags412Add CostCenter tag to unattributed resources
lowEmpty tag values189Populate or remove tags with blank values
mediumNon-compliant values67Standardize Environment values to prod/staging/dev

AI — License Optimization

Seat utilization across every AI platform with recommendations organized into three buckets: reduce (reclaim idle seats), elevate (upgrade power users), and expand (fulfill waitlist demand). Includes idle detection and estimated value recovered.

AI License Optimization

Total Pool

3,250

Active

2,088

Idle

362

Claude65%
528 active84 idle
ChatGPT62%
501 active112 idle
Copilot67%
541 active71 idle
Gemini64%
518 active95 idle

Recommendations

Reduce

  • 84 idle Claude seats — no activity in 60+ days. Reclaim and reallocate.
  • 112 idle ChatGPT seats — last active 45+ days ago. $33K/yr recoverable.
  • 9 Gemini Finance seats unused since provisioning. Reclaim entirely.

Elevate

  • 38 ChatGPT Team users hitting rate limits weekly — upgrade to Enterprise tier or shift to API.
  • 22 Claude Standard users averaging 4x the usage of peers — evaluate Premium upgrade for ROI.

Expand

  • 62 Copilot waitlist requests from Engineering — current utilization supports expansion.
  • 41 Claude waitlist requests from Product — idle seats available for immediate reassignment.
  • 28 ChatGPT requests from Customer Success — consider provisioning from reclaimed seats.

AI — Usage Patterns

Aggregate consumption patterns across your AI portfolio — model mix by spend, cost optimization opportunities, and individual flags for concerning behavior. Not a per-user roster, but the patterns that matter.

AI Usage Patterns

Model Mix by Spend

GPT-4o

42% · $38K/mo

Claude Sonnet

28% · $24K/mo

Claude Opus

12% · $18K/mo

GPT-4o Mini

11% · $4K/mo

Gemini Pro

7% · $3K/mo

Aggregate Patterns

PatternDetailImpact
Premium model overuse38% of GPT-4o requests are simple Q&A that could run on GPT-4o Mini at 1/10th the cost.$14K/mo
Burst spending detected3 single-day spikes exceeding $4K in the last 90 days — all from API workloads without rate limits.$12K
Underutilized capacityClaude API committed throughput is 40% underutilized. Consider downsizing or redistributing.$8K/mo
Cross-platform duplicationSame prompts detected across both ChatGPT and Claude APIs — consolidation opportunity.~$6K/mo

Individual Flags

user_a]@corp.comConsumed $12.4K in one week via Claude API — 8x peer average
svc-pipeline-03Retry loop generated 340K requests in 48hrs — no rate limiting
user_b@corp.comFast-mode escalation on 92% of ChatGPT requests — $3.2K premium

AI — Behavioral Insights (Tier 2)

Combined usage metrics and prompt categorization surfacing what your organization is actually doing with AI. Blends token volumes, model selection, and categorized prompt patterns into findings you can act on.

AI Behavioral Insights (Tier 2)

Prompt Categorization

What your organization is using AI for, based on categorized prompt analysis.

Code generation

38%

Content writing

24%

Data analysis

18%

Research / Q&A

12%

Other / unclassified

8%

Key Findings

FindingHeavy content generation users trending up 22% month-over-month

The top 8% of users (by token volume) average 920K tokens/user/month and are primarily focused on marketing content generation. Their usage has increased at a 22% monthly rate over the last quarter — if unchecked, this cohort alone will add $18K/mo by Q3.

FindingPremium models used for simple tasks — $14K/mo recoverable

42% of prompts sent to GPT-4o and Claude Opus are single-turn Q&A or formatting requests. Prompt categorization confirms these produce equivalent results on cheaper models. Shifting this traffic to GPT-4o Mini or Claude Haiku would save $14K/mo with no quality loss.

RiskSensitive data patterns detected in prompts

Categorization flagged prompts containing what appear to be customer names, account numbers, and internal financial data across 34 users. Recommend reviewing AI acceptable use policy and implementing input guardrails before expanding access.

OpportunityCode generation is high volume but low iteration — workflow opportunity

Engineering users generate 3.4x the token volume of other groups, but 78% of code-gen prompts are single-shot with no follow-up. Established AI workflow training could increase output quality and reduce redundant generation — estimated 15-20% token reduction.

FindingShadow AI spend detected outside procurement

API traffic analysis combined with prompt categorization identified 3 unsanctioned tools routing through personal API keys. Combined spend: $4.2K/mo. Usage patterns suggest these tools duplicate capabilities already available through sanctioned platforms.

What the Full Report Includes

The written assessment is delivered as a structured document covering:

  • Executive summary — total identified savings, key risks, prioritized recommendations
  • Cloud savings findings — rate optimization (commitment coverage, 30-day opportunities, over-commitment, 1yr/3yr recommendations) and usage optimization (rightsizing, idle resources, tagging health)
  • AI cost and usage findings — seat utilization vs. contracted across all platforms, spend by team/user/model, anomaly events, license recommendations, cross-platform inventory
  • (Tier 2 only) Behavioral findings — categorized prompt analysis, use-case patterns across teams, governance recommendations grounded in actual usage

Each finding includes the underlying data, the recommended action, the projected savings or risk reduction, and the suggested implementation owner.

Want to see how this applies to your environment?

Get your free savings assessment