ExampleCo audit workspace
AI Spend Dashboard
Demo view for a metadata-only audit across provider invoices, usage exports, token volume, latency, errors, retries, and API-key labels.
Monthly spend$42.6k+18% vs prior month
Projected opportunity$11.9k28% estimated range
Requests reviewed8.4Mmetadata only
High-confidence actions52 ready for benchmark
Daily AI Spend
Top Findings
Premium model overuseShort classification calls use the same model as customer-facing chat.
Retry amplificationTwo API keys show repeated failed calls during the same 11-minute window.
Batchable summary workNightly document summarization is priced like realtime interactive traffic.
Missing labels19% of spend lacks app or workflow tags, lowering confidence.
Routing Opportunities
| Workflow | Current Provider/Model | Monthly Spend | Suggested Action | Confidence | Risk |
|---|---|---|---|---|---|
| Support tagging | OpenAI premium chat model | $8,940 | Shadow smaller model on 2k samples | High | Low |
| RAG preprocessing | Mixed frontier summaries | $9,720 | Move nightly work to batch route test | High | Low |
| Embeddings | Single provider embedding route | $6,880 | Compare retrieval quality against cheaper embeddings | Medium | Medium |
| Customer chat | Primary production model | $12,440 | Observe only; do not reroute in phase one | Low | Hold |