06Measure
What your AI costs, and what it delivers.
Every call is priced from your own model prices. Record the business outcomes your workflows produce, and Worksi puts value next to cost in language leadership understands.
Outcomes, ROI and exports are included in Pro, Business and Enterprise, and in the trial.
- value generated
- outcomes × value per outcome
- total cost
- AI spend + platform + human review
- value multiple
- value ÷ total cost
- net value
- value − total cost
Prices come from your own model registry. Platform cost is prorated to the period.
Outcomes turn activity into value.
Define what a successful run is worth, a resolved ticket or a qualified lead, and record outcomes from code or the API. Without a value, the definition’s unit value is used.
- Outcome definitions with unit values, per workflow.
- Baselines from before AI: volume, spend, failure rate, human intervention and cost per outcome.
- Success rate and cost per outcome, per workflow, agent or team.
SDK packages are distributed by Worksi. Request access. Any OpenAI or Anthropic SDK works with the gateway without them.
with aiw.run("support-ticket-resolution", agent="triage-agent") as run: ... # uses the outcome definition's unit value run.outcome("ticket_resolved", success=True) # an outcome that arrives outside a runaiw.outcome( "lead_qualified", workflow="lead-qualification", success=score >= 70,)await aiw.run( { workflow: "support-ticket-resolution", agent: "triage-agent", }, async (run) => { // … // uses the outcome definition's unit value run.outcome("ticket_resolved", { success: true }) },)Value next to cost, in one view.
Performance → ROI shows what the period’s outcomes were worth beside what they cost. Every call in that cost is priced from your model registry, cached tokens included.
1. Value multiple: value divided by total cost.
2. Value generated by the outcomes recorded in the period.
3. Outcomes and success rate, successful out of recorded.
Before and after, with baselines.
Compare today’s failure rate, human intervention rate and cost per outcome with the baselines you measured before AI, per workflow.
1. Baseline period: each metric is compared with the period before AI.
2. Cost per outcome, before and after.
The ROI counts every cost.
ROI isn’t value divided by the model bill. Worksi’s cost model includes the platform and the people who review AI work.
- Value
- Successful outcomes × their value.
- AI spend
- Priced model and tool calls in the period.
- Platform cost
- Your monthly platform cost, prorated to the period.
- Human review cost
- Runs that needed an approval or an operator action × your cost per review.
- Net value and ROI
- Value minus total cost, and as a percentage of cost.
- Value multiple
- Value divided by total cost.
For finance
- Cost of every call in USD, from list prices in USD or CNY, rolled up by model, agent, workflow, team, provider or environment.
- Budget burn-down from your spend-limit policies, and a month‑end forecast from month-to-date spend.
- Cost allocation to teams, agents, workflows or environments, with a chargeback CSV.
- Provider invoice reconciliation against the spend Worksi computed.
- CSV and JSON exports of cost and ROI.
Put a price on every AI call.
Prices come from your own model registry, imported from the provider catalog or set by you, so cost per call, per agent and per outcome is yours from the first request.
SDK packages are distributed by Worksi. Request access. Any OpenAI or Anthropic SDK works with the gateway without them.
from aiworkforce import AIW aiw = AIW(service="triage-agent")with aiw.run("support-ticket-resolution", agent="triage-agent") as run: ... # uses the outcome definition's unit value run.outcome("ticket_resolved", success=True)