Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.flowx.ai/llms.txt

Use this file to discover all available pages before exploring further.

Financial ROI is the bottom of the stack — the actual numbers your CFO needs. Per-agent payback, annual ROI %, 5-year Net Present Value, with risk adjustment.

Configuring agent baselines

The ROI calculation hinges on the baseline — what manual effort the agent replaces. Without a baseline, an agent has no measurable upside. Each agent baseline carries:
FieldMeaning
Manual hours per caseHow long a human takes to do what the agent now does.
Hourly rateLoaded cost — salary, benefits, overhead.
Volume per monthRealistic monthly throughput.
Replacement ratio1.0 means the agent fully replaces the human; lower values reflect partial replacement.
The monthly labour saved is hours × rate × volume × replacement_ratio.

The four numbers

NumberFormulaWhat it tells you
Monthly savingsLabour saved − LLM cost − ops overheadRun-rate impact.
Payback periodImplementation cost ÷ monthly savingsHow fast the investment pays for itself.
Annual ROI %(Annual savings − annual cost) ÷ implementation costPercentage return on the up-front investment.
5-year NPVDiscounted sum of net cashflows over 5 yearsThe standard finance comparison number.
Default discount rate is 10% — configurable per org.

Risk adjustment

A favourable headline ROI that ignores compliance risk is a fiction. Observatory adjusts:
risk_adjusted_npv = npv × (1 − risk_penalty)
The risk penalty is derived from the app’s Risk Dashboard score. High-risk apps see a meaningful haircut on their NPV; low-risk apps see ~none. This means an agent with low LLM cost and high compliance gaps may rank below an agent with higher LLM cost but a clean governance posture.

Per-project view

Project-level ROI is the sum of per-agent ROI for agents in the project, with shared implementation costs amortised across them. The project view shows:
  • Aggregate payback period
  • Per-agent contribution stacked
  • Sensitivity bands from Monte Carlo

Worked example

A claims-approval agent in production:
InputValue
Manual hours per case0.6
Hourly rate$52
Volume / month4,200
Replacement ratio0.85
LLM cost / month$1,180
Implementation cost$48,000
Risk penalty0.08
Computed:
  • Monthly labour saved: 0.6 × 52 × 4,200 × 0.85 = $111,384
  • Monthly savings: 111,384 − 1,180 = $110,204
  • Payback: 48,000 ÷ 110,204 ≈ 0.4 months
  • Annual ROI %: ≈ 2,650%
  • 5-year NPV (10% discount): ≈ $5.0M
  • Risk-adjusted NPV: ≈ $4.6M

Compliance ROI

The audit-savings half of the calculation.

Sensitivity analysis

Replace point estimates with confidence ranges.
Last modified on June 2, 2026