← Back to Dashboard
1. Platform Team Patterns2. Shared Services and Chargeback

Shared Services and Chargeback

📚 Enterprise Architecture⏱ 10 min⭐ 115 XP⌨ Hands-on lab

Cost Governance at Scale

Implement chargeback/showback models so each business unit understands AI usage and efficiency trends. Shared AI platforms without cost attribution breed the tragedy of the commons: everyone's workload is someone else's bill.

Attribution Mechanics on AWS

  • Application inference profiles - create per-team/per-app inference profiles wrapping a base model, tag them, and Bedrock usage becomes attributable in Cost Explorer by tag.
  • Cost allocation tags - tag every AI resource (profiles, KBs, guardrails, Lambda/ECS) with team, product, env.
  • Token telemetry - your per-invocation log line ({tenant, model, tokens}) gives finer-grained attribution than billing ever will - join it with pricing to build near-real-time showback.

Showback vs Chargeback

ModelHow it worksWhen
ShowbackVisibility reports per BU, no money movesStart here - drives awareness without budget wars
ChargebackCosts land on BU budgetsMature platforms with trusted attribution + stable baselines

Report Efficiency, Not Just Spend

# the monthly view each BU should see
team: support-ai
spend:            $8,420   (▲ 12% MoM)
cost/successful-outcome: $0.031  (▼ 8% - improving!)
token efficiency: 2,140 tok/request (▼ 15% after caching)
quality:          task completion 91% (stable)

Pair cost visibility with quality outcomes to avoid unhealthy optimization pressure - a team "saving" 40% by silently degrading answer quality is a regression wearing a savings badge. Spend rising while cost-per-outcome falls is often success, not a problem.

Budget guardrails beat budget surprises: per-team token budgets with 80% alerts and a documented burst policy catch runaway loops in hours. The worst cost incident is the one discovered on the monthly invoice.
⌨ HANDS-ON LABAllocate Model Costs by Cost Center Using Cost Allocation Tags
⭐ +150 XP

Query AWS Cost Explorer API for Bedrock token expenditure broken down by project tag.

1Get cost and usage data grouped by CostCenter tag.
lab-sandbox — simulated environment
INFINITY LAB SANDBOX v2.6 — simulated shell
Type the command for the current objective. Helpers: "hint", "solution", "clear".
$
OBJECTIVE 1 / 1 — type "hint" if stuck
🧪 Knowledge Check
Press 1-4 to select1 of 2
Why implement AI chargeback/showback?
To hide costs
To improve accountability and spend transparency
To disable optimization
To remove SLOs