Forecast AI workflow execution economics
Forecast multi-step AI workflow demand, retry and loop depth, success, p95 latency, full cost, failure loss and net business value with tenant-local empirical Bayes, log-normal attempt economics, shared operating scenarios and common-control failure VaR/CVaR.
What it's for
Shows CTOs what an agent workflow will really cost and deliver when loops, tool failures, latency and shared controls—not just the happy path—are included.
What you give it
Inputs split into evidence read from your connected systems, calibration your team owns, and numerical controls that affect precision but never the result's meaning.
| Field | Type | Role | Required |
|---|---|---|---|
| current_workflows | array of objects (11 fields) | Evidence | Yes |
| historical_workflow_periods | array of objects (12 fields) | Evidence | Yes |
| horizon_periods | integer ≥ 1, ≤ 120 | Your calibration | Optional |
| max_detail_rows | integer ≥ 1, ≤ 500 | Numerical control | Optional |
| minimum_active_periods | integer ≥ 3 | Your calibration | Optional |
| minimum_historical_periods | integer ≥ 3 | Your calibration | Optional |
| prior_strength | number > 0 | Your calibration | Optional |
| scenarios | array of objects (9 fields) | Evidence | Yes |
| seed | integer ≥ 0 | Numerical control | Optional |
| simulations | integer ≥ 100, ≤ 100000 | Numerical control | Optional |
| tail_probability | number ≥ 0.001, ≤ 0.5 | Your calibration | Optional |
Each historical_workflow_periods
record
| Field | Type | Required |
|---|---|---|
| abandoned_execution_count | integer (≥ 0) | Yes |
| evidence_verified | boolean | Yes |
| failed_execution_count | integer (≥ 0) | Yes |
| id | string (non-empty) | Yes |
| initiated_execution_count | integer (≥ 0) | Yes |
| period | integer (≥ 0) | Yes |
| retry_step_attempt_count | integer (≥ 0) | Yes |
| succeeded_execution_count | integer (≥ 0) | Yes |
| total_cost | number (≥ 0) | Yes |
| total_step_attempt_count | integer (≥ 0) | Yes |
| total_trace_latency_ms | number (≥ 0) | Yes |
| workflow_class | string (non-empty) | Yes |
{
"current_workflows": [
{
"evidence_verified": true,
"fixed_cost_per_period": 10,
"forecast_execution_count_per_period": 1000,
"id": "support-current",
"loss_per_failure": 2,
"maximum_cost_per_execution": 1,
"maximum_p95_trace_latency_ms": 1000,
"maximum_step_attempts_per_execution": 4,
"minimum_success_probability": 0.8,
"value_per_success": 1,
"workflow_class": "support-resolution"
}
],
"historical_workflow_periods": [
{
"abandoned_execution_count": 2,
"evidence_verified": true,
"failed_execution_count": 8,
"id": "support-00",
"initiated_execution_count": 100,
"period": 0,
"retry_step_attempt_count": 40,
"succeeded_execution_count": 90,
"total_cost": 14,
"total_step_attempt_count": 140,
"total_trace_latency_ms": 14000,
"workflow_class": "support-resolution"
},
{
"abandoned_execution_count": 2,
"evidence_verified": true,
"failed_execution_count": 8,
"id": "support-01",
"initiated_execution_count": 100,
"period": 1,
"retry_step_attempt_count": 40,
"succeeded_execution_count": 90,
"total_cost": 14.1,
"total_step_attempt_count": 140,
"total_trace_latency_ms": 14020,
"workflow_class": "support-resolution"Truncated for display — the full payload is 213 lines.
What you get back
This is the actual output of running the example above — computed by the same function the platform calls, not an illustration.
{
"assumptions": [
"History retains every initiated execution, terminal outcome, retry/loop attempt, trace latency and reconciled cost in consecutive periods including zeros under a stable workflow class.",
"Beta success, Gamma retry pressure and log-normal per-attempt cost/latency are tenant-local predictive approximations; common control failure and management scenarios are shared across workflows.",
"Business value and loss are finance-owned non-overlapping horizon values. The forecast is conditional, not causal incremental ROI or an SLA.",
"Breach probabilities support aggregate workflow governance and never rank providers, teams or people."
],
"configuration": {
"common_scenario_rule": "one_scenario_and_common_control_state_is_shared_across_the_portfolio_per_simulation",
"horizon_periods": 6,
"minimum_active_periods": 8,
"minimum_historical_periods": 12,
"prior_strength": 2,
"seed": 43,
"simulations": 200,
"tail_probability": 0.1
},
"decision": "ai_workflow_execution_economics_forecast_supported",
"failed_gates": [],
"method": "empirical_bayes_ai_workflow_loop_cost_latency_value_forecast_v1",
"scenario_diagnostics": [
{
"probability": 0.8,
"scenario_id": "base"
},
{
"probability": 0.2,
"scenario_id": "control-stress"
}
],
"summary": {
"economic_loss_conditional_value_at_risk": 16588.6661,
"economic_loss_value_at_risk": 15466.2608,
"expected_failure_loss": 2838.78,
"expected_net_business_value": 871.357,
"expected_total_cost": 1147.493,
"probability_common_control_failure": 0.115,
"probability_of_negative_net_business_value": 0.115,
"unsupported_workflow_count": 0,
"workflow_count": 1
},
"truncated_scenario_count": 0,
"truncated_workflow_count": 0,
"unsupported_workflow_ids": [],Truncated for display — the full payload is 64 lines.
How it works
Forecasting & survival — Estimate when something completes or fails, with censoring and unresolved work handled honestly rather than dropped.
- 1 Retain consecutive workflow periods including zeros and reconcile initiated executions to success, failure and abandonment plus nested total/retry step attempts.
- 2 Learn tenant-local Beta success, Gamma retry pressure and log-normal cost/latency per attempt, then draw one coherent management scenario and common-control state for the whole portfolio.
- 3 Simulate bounded loop attempts, terminal outcomes, total cost, failure loss and net value; expose local success/cost/latency breaches and portfolio economic-loss VaR/CVaR with sparse-history abstention.
Before you trust it
Every tool in the catalog ships with the conditions under which its answer is meaningful — and the conditions under which it should abstain instead of guessing.
Assumptions & guardrails
- Training examples precede their outcomes, censoring and unresolved work are represented, and deployment populations remain comparable to validation cohorts.
- Workflow definitions, terminal states, attempts, cost, latency and outcome value are stable over the selected epoch; zero periods and unfinished work are retained and scenarios share one horizon.
- A predictive interval or risk estimate is not a deadline promise, causal explanation, or individual-performance judgment.
- This is a conditional operational forecast, not causal ROI, an SLA, a safety certificate or a prediction about a provider, team or person.
Minimum evidence
- historical_workflow_periods: required and organization-defined
- current_workflows: required and organization-defined
- scenarios: required and organization-defined
How to validate it
Use chronological train/calibration/test windows, compare proper scores and decision value with a simple baseline, and recalibrate only from outcomes resolved after prediction time.
Calibrating it to your org
Same for everyone
The mathematical kernel, validation rules, method version, and JSON output semantics are organization-independent; no tenant-trained coefficients or company benchmark is embedded in the function.
Specific to you
- stable workflow-class period projection from complete execution traces and mature terminal states, joined to effective cost and shared-control incident epochs without dropping unfinished or zero periods
- workflow epoch and terminal-state maturity, zero and retry policy, step-attempt and latency semantics, cost allocation, forecast horizon/demand, loop cap, success/cost/latency thresholds, value/loss, common-control scenarios, priors, support and tail appetite
Calibration workflow
- 1 Define the management decision, target outcome, aggregate unit, privacy boundary, cadence, and prediction/intervention horizon for this organization.
- 2 Build a tenant-scoped historical cohort using only information available before each prediction or decision; preserve zero periods, censoring, assignment probabilities, and unresolved outcomes when the method requires them.
- 3 Estimate statistical parameters on training history, but obtain costs, utilities, risk tolerance, practical-effect thresholds, capacity, and policy constraints from accountable decision owners.
- 4 Validate on later time windows or held-out aggregate units at the deployment grain, against a simple baseline and the function-specific validation strategy.
- 5 Deploy only if the returned decision clears evidence, overlap, calibration, robustness, and guardrail checks; warning, unsupported, schema-gap, and fallback decisions are abstentions.
- 6 Monitor realized outcomes, data drift, coverage, and decision regret; recalibrate at a governed cadence or after a detected regime/definition change, never merely because a stakeholder dislikes the result.
Call it from your AI
You don't wire up 388 tools in your MCP client. The GitRevio MCP server exposes 18 tools, three of which let an agent search the catalog, read a tool's schema, and run it — so the assistant finds this one on its own.
gitrevio_capabilities_search
{ "q": "forecast multistep ai workflow demand retry" }
→ finds "forecast_ai_workflow_execution_economics"
gitrevio_capability_describe
{ "capability_id": "forecast_ai_workflow_execution_economics" }
→ returns the input schema and agent guidance shown on this page
gitrevio_capability_run
{ "capability_id": "forecast_ai_workflow_execution_economics", "arguments": { ... } }
→ returns the result shown aboveWorks in Claude Desktop, Claude Code, Cursor, Cline, Continue.dev, Goose and Aider. See the MCP server.
Related tools
Audit AI workflow trace value integrity
Audit every AI workflow execution from root trace through model, tool, cache, review and control steps to one mature business outcome, reconciling parent lineage, retries, wall-clock latency, direct cost and uniquely attributed net value while retaining unfinished work.
Optimize AI workflow design portfolio
Select one governed AI workflow graph per use case, maximizing risk-adjusted business value under hard control, success, latency and scenario-availability gates plus shared model/tool/review capacity, dependencies, implementation budget and economic-regret CVaR.
Audit AI capability fallback integrity
Prove that every aggregate capability required when AI is unavailable has a current approved runbook and a sufficiently large, timely, successful, independently observed exercise conducted with AI actually disabled.
Audit AI code change evidence integrity
Prove that aggregate AI-assisted coding evidence comes from prospectively registered, nonoverlapping treatment/control studies with immutable assignment, configuration, trace and mature-outcome denominators before anyone estimates an effect.
Audit AI inference cost allocation integrity
Reconcile provider AI invoices bottom-up to workload and route usage, price terms, cached requests, retries, fixed charges and credits without combining currencies or silently allocating unexplained spend.
Audit AI knowledge grounding integrity
Audit the complete AI knowledge supply chain from immutable source versions through indexed chunks and effective access policy to retrieved evidence, claim-level citations and honestly mature grounding outcomes, without treating unresolved answers as failures.