Mappings
Connect AI usage to business margin
Raw AI usage is just rows of tokens. Mapped to AcmeOps AI tenants, plans, features, and models, it becomes margin risk Costara can name and act on.
Tenant
Plan, revenue, AI budget, target margin.
Feature
Owner team and business function.
Model
Provider and seeded input/output prices.
Tenant mapping
5 tenantsEach tenant carries a plan, monthly revenue, AI budget, and target gross margin. This is what turns AI cost into a margin signal.
- Total managed revenue
- $41,500
- Total AI budget
- $15,250
| Tenant | Plan | Monthly revenue | Monthly AI budget | Target margin |
|---|---|---|---|---|
| Northstar Logistics | Enterprise | $14,000 | $5,600 | 60% |
| Atlas Health | Enterprise | $14,000 | $5,600 | 60% |
| Redwood Finance | Growth | $6,500 | $1,950 | 70% |
| BrightCart | Growth | $4,200 | $1,260 | 70% |
| Apex Legal | Starter | $2,800 | $840 | 70% |
Feature mapping
5 featuresEach product feature is owned by a team and serves a business function. Costara routes margin risk back to the people who can act on it.
- Customer-facing revenue feature
- 4 features
- Customer support automation
- 1 feature
| Feature | Owner team | Business function |
|---|---|---|
| Contract Analyzer | Legal Tech Team | Customer-facing revenue feature |
| Document Q&A | Knowledge Team | Customer-facing revenue feature |
| Research Assistant | Research Team | Customer-facing revenue feature |
| Summarization | Content Team | Customer-facing revenue feature |
| AI Support Copilot | Customer Success | Customer support automation |
Model mapping
6 modelsEvery model has explicit input and output list prices per million tokens. Costara never defaults a model to zero cost.
- Most expensive
- GPT-5.5
- Cheapest
- Llama 3.3 70B
List price per million tokens
- GPT-5.5Input $5 · Output $30Input$5Output$30
- Claude Opus 4.7Input $5 · Output $25Input$5Output$25
- Claude Sonnet 4.6Input $2 · Output $12Input$2Output$12
- GPT-5.4Input $2.50 · Output $11.25Input$2.50Output$11.25
- Gemini 2.5 ProInput $1.25 · Output $10Input$1.25Output$10
- Llama 3.3 70BInput $1 · Output $2Input$1Output$2
| Model | Provider | Input / 1M tokens | Output / 1M tokens | Pricing source |
|---|---|---|---|---|
| GPT-5.5 | OpenAI | $5 | $30 | OpenAI API pricing (Seeded demo value) |
| GPT-5.4 | OpenAI | $2.50 | $11.25 | OpenAI API pricing, long-context tier (Seeded demo value) |
| Claude Opus 4.7 | Anthropic | $5 | $25 | Anthropic list pricing (Seeded demo value) |
| Claude Sonnet 4.6 | Anthropic | $2 | $12 | Anthropic list pricing (Seeded demo value) |
| Gemini 2.5 Pro | $1.25 | $10 | Google Gemini API pricing, up to 200K prompt tier (Seeded demo value) | |
| Llama 3.3 70B | Meta | $1 | $2 | Together AI serverless pricing (Seeded demo value) |
All prices in this table are seeded demo values. Costara does not fetch live provider pricing in this build. AI cost is calculated as input_tokens × input_price + output_tokens × output_price for each row. GPT-5.5 and GPT-5.4 follow OpenAI API pricing. Claude Opus 4.7 and Claude Sonnet 4.6 follow Anthropic list pricing. Gemini 2.5 Pro follows Google Gemini API pricing for prompts up to 200K tokens. Llama 3.3 70B follows Together AI serverless pricing. Verify provider pricing before production use because rates, context tiers, cache discounts, and batch discounts can change.
Why mappings matter
Mappings are how Costara turns AI usage into margin risk.
Without these mappings, AI usage is just rows of tokens. With them, Costara names the unprofitable tenant, the risky feature, and the cost driver model, and recommends the action that protects gross margin.