# Staged Investment and Pilots — Estimating AI Automation ROI

Source: https://www.skillbyai.com/en/ai-automation-roi/r-stage

> Buy information before buying scale.

## Gate funding on evidence

Instead of approving the whole budget upfront, fund in **stages**: a small pilot to measure the key assumptions (automatable share, review time, error rates, adoption), then a limited rollout, then full scale. Define in advance the results that justify the next stage. This caps losses if assumptions fail and makes the remaining estimate far more reliable. The tornado chart tells you which assumptions the pilot must test.

## A staged funding plan

Each stage has a budget and a decision rule.

```text
stage 1  pilot (1 team, 6 weeks)         $12,000   proceed if minutes saved >= 2.5 and error rate <= 2.5%
stage 2  limited rollout (3 teams)       $25,000   proceed if adoption >= 60% by week 8 and realisation plan agreed
stage 3  full rollout                   $16,000   track monthly vs forecast; reforecast quarterly
```

## Pre-agree the kill criteria

Decide before the pilot what result would stop the project; it prevents sunk-cost decisions later.

**Quiz:** What is the main benefit of staged investment?

- [ ] It removes the need for measurement
- [ ] It hides costs
- [ ] It guarantees success
- [x] It limits losses and tests key assumptions before full spending

*Answer:* It limits losses and tests key assumptions before full spending. Learn cheaply, then scale.