# Tracking the Drivers — Estimating AI Automation ROI

Source: https://www.skillbyai.com/en/ai-automation-roi/p-track

> Measure the inputs of the model, not just the outcome.

## A driver dashboard

Track each driver of the business case monthly: volume, adoption (share of items processed with AI), minutes per item, review and exception rates, error rate, realisation (hires avoided, overtime cut), and running costs. When the outcome disappoints, the drivers show why. Assign an owner to each and agree how it is measured; drivers that cannot be measured should not be in the case.

## A monthly driver report

One row per driver, plan versus actual.

```text
driver               plan     actual   source                owner
volume / month       12,000   10,400   ERP                   AP lead
adoption             85%      72%      tool logs             AP lead
minutes saved/item   3.81     2.6      timing sample + logs  ops analyst
realisation          60%      40%      HR / overtime data    finance
error rate drop      2.5 pt   1.8 pt   audit sample          controller
monthly cost         $3,016   $3,680   invoices + cloud      AI team
```

## Report drivers, not just ROI

A single ROI number cannot tell you what to fix; driver gaps can.

**Quiz:** Why track each driver of the business case separately?

- [ ] To hide the outcome
- [x] To see which assumption caused a gap between plan and actual
- [ ] Because ROI cannot be computed
- [ ] It is only needed for audits

*Answer:* To see which assumption caused a gap between plan and actual. Drivers explain outcomes.
