AI Spend Governance
Model, agent and GPU economics
AI Spend Agent attribution across Azure OpenAI, Bedrock and Vertex — tokens, GPU hours, cost per workflow and measured value per dollar of agent spend.
Total AI / model spend
$14.8M
+84% YoY
Token consumption
184.3B
31% served by oversized models
GPU hours consumed
42,100
47% off-hours utilization
Blended AI ROI
3.2x
value per $1 of agent spend
Cost by model
$M annualized
Cost by business unit
$M annualized, showback view
Adoption versus cost
Bubble size = ROI score. Bottom-right is healthy, top-left is waste.
Agent portfolio
Cost, consumption, adoption and governance verdict
| AI agent | Model | Business unit | Annual cost | Tokens | GPU hrs | Adoption | ROI | Verdict |
|---|---|---|---|---|---|---|---|---|
| Revenue Assurance Agent | GPT-4.1 / o-series | Revenue Assurance | $3.4M | 41.2B | 18,400 | 88% | 6.2x | Keep and scale |
| Customer Care Agent | GPT-4o mini | Customer Experience | $2.6M | 96.4B | 9,100 | 74% | 4.1x | Optimize and scale |
| Experimental Campaign Agent | Claude / Bedrock mix | Brand Innovation | $2.2M | 12.7B | 6,400 | 6% | 0.4x | Retire |
| Network Engineer Copilot | GPT-4o + SLM router | Network Operations | $1.9M | 28.9B | 7,300 | 62% | 3.8x | Expand |
| HR TalentScout Agent | Phi-4 | People & Culture | $0.4M | 5.1B | 900 | 33% | 1.7x | Monitor |
Cost per AI workflow
Blended cost of a completed business outcome
Care interaction resolved
$0.34
Revenue leakage case investigated
$4.10
Network incident triaged
$1.86
Campaign asset generated
$12.40