Return on AI Spend
Every dollar of AI spend, tied to work that shipped
Arcline scores every unit of delivered work on outcome, complexity, and business consequence — so you can see exactly which people, teams, and models return more than they cost, and reallocate spend with confidence.
How the engine scores spend
Every unit of delivered work, plotted: model token spend against attributed business value. Anything above the break-even line returned more than it cost.
1 · Meter
Token spend per work item, per model — Claude, GPT, Gemini, and open weights
2 · Gate
Outcome decides if value exists: shipped and fixes count, rejected is zero, rollbacks go negative
3 · Weigh
Complexity × business consequence sets the value of what shipped
4 · Attribute
Value lands on the exact person + model pairing that produced it
Claude Opus 4.5$42/MTok
GPT-5.2 Codex$28/MTok
Claude Sonnet 4.5$12/MTok
Gemini 2.5 Pro$9/MTok
Qwen 3 Coder$3/MTok
Highest-value pairing
Rotimi Adeyemi+Claude Opus 4.5
$42K net value on 7 delivered work items at a 100% ship rate (13.4x ROI). This person-model pairing is the org's strongest spend-to-value converter — replicate their working patterns.
Weakest pairing
Jo Rivera+Claude Opus 4.5
$3K net on 2 work items at a 50% ship rate. Too much rejected and rolled-back work is burning spend without delivered value — recommend a model switch or targeted coaching before next budget cycle.
People × model pairings
Who paired with which model returned the most above cost
| Person | Model | Tokens | Value | |||||
|---|---|---|---|---|---|---|---|---|
Rotimi AdeyemiInfra | Claude Opus 4.5 | 7 | 100% | 80.6M | $3K | $46K | $42K | 13.4x |
Alex ChenPlatform | Claude Opus 4.5 | 5 | 100% | 58.0M | $2K | $44K | $41K | 17.9x |
Tove LarsenML Systems | GPT-5.2 Codex | 6 | 100% | 80.6M | $2K | $42K | $40K | 18.6x |
Priya GuptaML Systems | GPT-5.2 Codex | 5 | 100% | 44.8M | $1K | $33K | $31K | 26.1x |
Kei TanakaInfra | Claude Sonnet 4.5 | 4 | 100% | 44.1M | $529 | $27K | $27K | 51.4x |
Dana OkaforPlatform | Gemini 2.5 Pro | 5 | 80% | 82.6M | $743 | $27K | $26K | 35.8x |
Sun ParkFrontend | Claude Sonnet 4.5 | 7 | 71% | 88.0M | $1K | $26K | $25K | 24.7x |
Kei TanakaInfra | GPT-5.2 Codex | 8 | 63% | 116.7M | $3K | $28K | $25K | 8.5x |
Dana OkaforPlatform | Claude Sonnet 4.5 | 6 | 67% | 78.2M | $939 | $24K | $23K | 25.3x |
Jo RiveraFrontend | GPT-5.2 Codex | 5 | 100% | 66.0M | $2K | $24K | $22K | 13x |
Evelyn MarshInfra | Gemini 2.5 Pro | 5 | 80% | 102.9M | $927 | $23K | $22K | 24.7x |
Priya GuptaML Systems | Claude Sonnet 4.5 | 3 | 100% | 26.9M | $323 | $21K | $21K | 65.5x |
Dana OkaforPlatform | Qwen 3 Coder | 2 | 100% | 40.4M | $121 | $21K | $21K | 170.6x |
Rotimi AdeyemiInfra | Claude Sonnet 4.5 | 3 | 100% | 33.0M | $396 | $20K | $19K | 49.4x |
Priya GuptaML Systems | Claude Opus 4.5 | 5 | 80% | 68.1M | $3K | $21K | $18K | 7.3x |
Tove LarsenML Systems | Claude Sonnet 4.5 | 8 | 50% | 112.5M | $1K | $19K | $18K | 14.3x |
Kei TanakaInfra | Gemini 2.5 Pro | 2 | 100% | 30.6M | $276 | $17K | $17K | 60.8x |
Linh NguyenFrontend | Gemini 2.5 Pro | 3 | 100% | 52.9M | $476 | $16K | $16K | 34.3x |
Jo RiveraFrontend | Gemini 2.5 Pro | 4 | 75% | 79.2M | $714 | $16K | $16K | 23x |
Misha IvanovPlatform | GPT-5.2 Codex | 4 | 75% | 76.3M | $2K | $17K | $15K | 7.8x |
Alex ChenPlatform | Claude Sonnet 4.5 | 4 | 50% | 53.1M | $637 | $15K | $14K | 23.2x |
Tove LarsenML Systems | Gemini 2.5 Pro | 3 | 67% | 49.1M | $442 | $14K | $14K | 31.7x |
Alex ChenPlatform | GPT-5.2 Codex | 2 | 100% | 16.8M | $471 | $13K | $13K | 28.3x |
Dana OkaforPlatform | Claude Opus 4.5 | 3 | 67% | 42.6M | $2K | $15K | $13K | 8.2x |
Cam BrooksML Systems | Qwen 3 Coder | 6 | 50% | 188.8M | $566 | $13K | $13K | 23.2x |
Linh NguyenFrontend | Claude Sonnet 4.5 | 6 | 50% | 118.6M | $1K | $14K | $12K | 9.6x |
Misha IvanovPlatform | Qwen 3 Coder | 3 | 100% | 64.7M | $194 | $12K | $11K | 59.5x |
Cam BrooksML Systems | Gemini 2.5 Pro | 4 | 75% | 90.5M | $815 | $12K | $11K | 14.8x |
Kei TanakaInfra | Claude Opus 4.5 | 2 | 100% | 26.7M | $1K | $12K | $11K | 11x |
Sun ParkFrontend | GPT-5.2 Codex | 5 | 60% | 70.9M | $2K | $13K | $11K | 6.5x |
Sun ParkFrontend | Gemini 2.5 Pro | 3 | 67% | 45.1M | $406 | $10K | $10K | 25.4x |
Cam BrooksML Systems | Claude Sonnet 4.5 | 3 | 67% | 51.4M | $617 | $11K | $10K | 17x |
Rotimi AdeyemiInfra | GPT-5.2 Codex | 2 | 100% | 21.2M | $594 | $10K | $9K | 16.1x |
Misha IvanovPlatform | Gemini 2.5 Pro | 5 | 40% | 114.0M | $1K | $10K | $9K | 9.6x |
Kei TanakaInfra | Qwen 3 Coder | 3 | 67% | 59.2M | $178 | $9K | $9K | 50.6x |
Evelyn MarshInfra | Qwen 3 Coder | 2 | 100% | 38.3M | $115 | $9K | $9K | 77.2x |
Sun ParkFrontend | Qwen 3 Coder | 2 | 50% | 58.0M | $174 | $8K | $7K | 43.4x |
Linh NguyenFrontend | Qwen 3 Coder | 3 | 67% | 66.0M | $198 | $6K | $6K | 32.4x |
Jo RiveraFrontend | Claude Sonnet 4.5 | 4 | 25% | 65.7M | $788 | $6K | $6K | 8.2x |
Evelyn MarshInfra | Claude Sonnet 4.5 | 8 | 25% | 130.9M | $2K | $6K | $4K | 3.5x |
Misha IvanovPlatform | Claude Sonnet 4.5 | 2 | 50% | 34.1M | $409 | $3K | $3K | 7.9x |
Jo RiveraFrontend | Claude Opus 4.5 | 2 | 50% | 42.0M | $2K | $4K | $3K | 2.4x |
Delivered work ledger
Every unit of shipped work, scored · value = complexity × business consequence, weighted by outcome
| Work item | Person | Model | Outcome | Cmplx | Cnsq | Token cost | Net value |
|---|---|---|---|---|---|---|---|
Batch inference scheduler rewrite#4745 · W12 | Priya Gupta | GPT-5.2 Codex | Shipped | 10 | 7 | $316 | $8K |
Circuit breaker on attribution service#4561 · W12 | Rotimi Adeyemi | Claude Opus 4.5 | Fix shipped · 1 bug | 8 | 7 | $483 | $7K |
Deprecate legacy auth middleware#4267 · W12 | Dana Okafor | Claude Sonnet 4.5 | Shipped | 9 | 7 | $148 | $7K |
Circuit breaker on attribution service#4671 · W12 | Evelyn Marsh | Gemini 2.5 Pro | Shipped · 1 bug | 8 | 5 | $195 | $5K |
Circuit breaker on attribution service#4597 · W12 | Rotimi Adeyemi | GPT-5.2 Codex | Shipped · 1 bug | 8 | 5 | $328 | $5K |
Connection pool rightsizing for peak load#4248 · W12 | Dana Okafor | Claude Sonnet 4.5 | Shipped | 8 | 5 | $131 | $5K |
Optimistic updates for playbook adoption#4457 · W12 | Jo Rivera | Gemini 2.5 Pro | Shipped | 6 | 6 | $129 | $4K |
Accessible focus management in detail sheet#4494 · W12 | Jo Rivera | Claude Opus 4.5 | Shipped | 6 | 6 | $710 | $4K |
Streaming SSR for dashboard shell#4498 · W12 | Jo Rivera | Gemini 2.5 Pro | Shipped | 5 | 6 | $148 | $3K |
Blue/green deploys for ingestion workers#4615 · W12 | Kei Tanaka | GPT-5.2 Codex | Rejected | 7 | 6 | $552 | $-552 |
Blue/green deploys for ingestion workers#4722 · W12 | Evelyn Marsh | Claude Opus 4.5 | Rejected | 6 | 9 | $853 | $-853 |
Optimistic updates for playbook adoption#4408 · W11 | Sun Park | Claude Sonnet 4.5 | Shipped | 9 | 5 | $133 | $5K |